Patent Application
20180351788
The present technology pertains to network configuration and troubleshooting, and more specifically to automatically localize faults in network policies deployed within a network.
Network configurations for large data center networks are often specified at a centralized controller. The controller can program switches, routers, servers, and elements in the network according to the specified network configurations. Network configurations are inherently very complex, and involve low level as well as high level configurations of several layers of the network such as access policies, forwarding policies, routing policies, security policies, QoS policies, etc. Given such complexity, the network configuration process is error prone. In many cases, the configurations defined on a controller, which can reflect an intent specification for the network, can contain errors and inconsistencies that are often extremely difficult to identify and may create significant problems in the network.
In order to describe the manner in which the above-recited and other advantages and features of the disclosure can be obtained, a more particular description of the principles briefly described above will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings. Understanding that these drawings depict only exemplary embodiments of the disclosure and are not therefore to be considered to be limiting of its scope, the principles herein are described and explained with additional specificity and detail through the use of the accompanying drawings in which:
Various embodiments of the disclosure are discussed in detail below. While specific implementations are discussed, it should be understood that this is done for illustration purposes only. A person skilled in the relevant art will recognize that other components and configurations may be used without parting from the spirit and scope of the disclosure.
Various embodiments of the disclosure are discussed in detail below. While specific implementations are discussed, it should be understood that this is done for illustration purposes only. A person skilled in the relevant art will recognize that other components and configurations may be used without parting from the spirit and scope of the disclosure. Thus, the following description and drawings are illustrative and are not to be construed as limiting. Numerous specific details are described to provide a thorough understanding of the disclosure. However, in certain instances, well-known or conventional details are not described in order to avoid obscuring the description. References to one or an embodiment in the present disclosure can be references to the same embodiment or any embodiment; and, such references mean at least one of the embodiments.
Reference to “one embodiment” or “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosure. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Moreover, various features are described which may be exhibited by some embodiments and not by others.
The terms used in this specification generally have their ordinary meanings in the art, within the context of the disclosure, and in the specific context where each term is used. Alternative language and synonyms may be used for any one or more of the terms discussed herein, and no special significance should be placed upon whether or not a term is elaborated or discussed herein. In some cases, synonyms for certain terms are provided. A recital of one or more synonyms does not exclude the use of other synonyms. The use of examples anywhere in this specification including examples of any terms discussed herein is illustrative only, and is not intended to further limit the scope and meaning of the disclosure or of any example term. Likewise, the disclosure is not limited to various embodiments given in this specification.
Without intent to limit the scope of the disclosure, examples of instruments, apparatus, methods and their related results according to the embodiments of the present disclosure are given below. Note that titles or subtitles may be used in the examples for convenience of a reader, which in no way should limit the scope of the disclosure. Unless otherwise defined, technical and scientific terms used herein have the meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. In the case of conflict, the present document, including definitions will control.
Additional features and advantages of the disclosure will be set forth in the description which follows, and in part will be obvious from the description, or can be learned by practice of the herein disclosed principles. The features and advantages of the disclosure can be realized and obtained by means of the instruments and combinations particularly pointed out in the appended claims. These and other features of the disclosure will become more fully apparent from the following description and appended claims, or can be learned by the practice of the principles set forth herein.
Disclosed herein are systems, methods, and computer-readable media for localizing faults. The systems, methods, and computer-readable media include identifying, using at least one processing device, a failure in a network comprising a plurality of network devices and a controller for managing a network policy deployed throughout the network, the network policy associated with a plurality of objects, wherein each object of the plurality of objects is a manageable component that impacts how network traffic is treated by at least one network device of the plurality of network devices. The systems, methods, and computer-readable media further include determining from the plurality of objects, using the at least one processing device, a set of objects that are indicative of the failure in the network.
The disclosed technology addresses the need in the art for identifying faults within the implementation of a network policy across a network. The present technology involves system, methods, and computer-readable media for identifying faults within network policies implemented in a network. The present technology will be described in the following disclosure as follows. The discussion begins with an introductory discussion of network assurance and a description of example computing environments, as illustrated in
Network assurance is the guarantee or determination that the network is behaving as intended by the network operator and has been configured properly (e.g., the network is doing what it is intended to do). Intent can encompass various network operations, such as bridging, routing, security, service chaining, endpoints, compliance, QoS (Quality of Service), audits, etc. Intent can be embodied in one or more policies, settings, configurations, etc., defined for the network and individual network elements (e.g., switches, routers, applications, resources, etc.). However, often times, the configurations, policies, etc., defined by a network operator are incorrect or not accurately reflected in the actual behavior of the network. For example, a network operator specifies a configuration A for one or more types of traffic but later finds out that the network is actually applying configuration B to that traffic or otherwise processing that traffic in a manner that is inconsistent with configuration A. This can be a result of many different causes, such as hardware errors, software bugs, varying priorities, configuration conflicts, misconfiguration of one or more settings, improper rule rendering by devices, unexpected errors or events, software upgrades, configuration changes, failures, etc. As another example, a network operator implements configuration C but one or more other configurations result in the network behaving in a manner that is inconsistent with the intent reflected by the implementation of configuration C. For example, such a situation can result when configuration C conflicts with other configurations in the network.
The approaches herein can provide network assurance by modeling various aspects of the network and/or performing consistency checks as well as other network assurance checks. The network assurance approaches herein can be implemented in various types of networks, including a private network, such as a local area network (LAN); an enterprise network; a standalone or traditional network, such as a data center network; a network including a physical or underlay layer and a logical or overlay layer, such as a VXLAN or software-defined network (SDN) (e.g., Application Centric Infrastructure (ACI) or VMware NSX networks); etc.
Network models can be constructed for a network and implemented for network assurance. A network model can provide a representation of one or more aspects of a network, including, without limitation the network's policies, configurations, requirements, security, routing, topology, applications, hardware, filters, contracts, access control lists, infrastructure, etc. As will be further explained below, different types of models can be generated for a network.
Such models can be implemented to ensure that the behavior of the network will be consistent (or is consistent) with the intended behavior reflected through specific configurations (e.g., policies, settings, definitions, etc.) implemented by the network operator. Unlike traditional network monitoring, which involves sending and analyzing data packets and observing network behavior, network assurance can be performed through modeling without necessarily ingesting packet data or monitoring traffic or network behavior. This can result in foresight, insight, and hindsight: problems can be prevented before they occur, identified when they occur, and fixed immediately after they occur.
Thus, network assurance can involve modeling properties of the network to deterministically predict the behavior of the network. The network can be determined to be healthy if the model(s) indicate proper behavior (e.g., no inconsistencies, conflicts, errors, etc.). The network can be determined to be functional, but not fully healthy, if the modeling indicates proper behavior but some inconsistencies. The network can be determined to be non-functional and not healthy if the modeling indicates improper behavior and errors. If inconsistencies or errors are detected by the modeling, a detailed analysis of the corresponding model(s) can allow one or more underlying or root problems to be identified with great accuracy.
The modeling can consume numerous types of smart events which model a large amount of behavioral aspects of the network. Smart events can impact various aspects of the network, such as underlay services, overlay services, tenant connectivity, tenant security, tenant end point (EP) mobility, tenant policy, tenant routing, resources, etc.
Having described various aspects of network assurance, the disclosure now turns to a discussion of example network environments for network assurance.
Leafs 104 can be responsible for routing and/or bridging tenant or customer packets and applying network policies or rules. Network policies and rules can be driven by one or more Controllers 116, and/or implemented or enforced by one or more devices, such as Leafs 104. Leafs 104 can connect other elements to the Fabric 120. For example, Leafs 104 can connect Servers 106, Hypervisors 108, Virtual Machines (VMs) 110, Applications 112, Network Device 114, etc., with Fabric 120. Such elements can reside in one or more logical or virtual layers or networks, such as an overlay network. In some cases, Leafs 104 can encapsulate and decapsulate packets to and from such elements (e.g., Servers 106) in order to enable communications throughout Network Environment 100 and Fabric 120. Leafs 104 can also provide any other devices, services, tenants, or workloads with access to Fabric 120. In some cases, Servers 106 connected to Leafs 104 can similarly encapsulate and decapsulate packets to and from Leafs 104. For example, Servers 106 can include one or more virtual switches or routers or tunnel endpoints for tunneling packets between an overlay or logical layer hosted by, or connected to, Servers 106 and an underlay layer represented by Fabric 120 and accessed via Leafs 104.
Applications 112 can include software applications, services, containers, appliances, functions, service chains, etc. For example, Applications 112 can include a firewall, a database, a CDN server, an IDS/IPS, a deep packet inspection service, a message router, a virtual switch, etc. An application from Applications 112 can be distributed, chained, or hosted by multiple endpoints (e.g., Servers 106, VMs 110, etc.), or may run or execute entirely from a single endpoint.
VMs 110 can be virtual machines hosted by Hypervisors 108 or virtual machine managers running on Servers 106. VMs 110 can include workloads running on a guest operating system on a respective server. Hypervisors 108 can provide a layer of software, firmware, and/or hardware that creates, manages, and/or runs the VMs 110. Hypervisors 108 can allow VMs 110 to share hardware resources on Servers 106, and the hardware resources on Servers 106 to appear as multiple, separate hardware platforms. Moreover, Hypervisors 108 on Servers 106 can host one or more VMs 110.
In some cases, VMs 110 and/or Hypervisors 108 can be migrated to other Servers 106. Servers 106 can similarly be migrated to other locations in Network Environment 100. For example, a server connected to a specific leaf can be changed to connect to a different or additional leaf. Such configuration or deployment changes can involve modifications to settings, configurations and policies that are applied to the resources being migrated as well as other network components.
In some cases, one or more Servers 106, Hypervisors 108, and/or VMs 110 can represent or reside in a tenant or customer space. Tenant space can include workloads, services, applications, devices, networks, and/or resources that are associated with one or more clients or subscribers. Accordingly, traffic in Network Environment 100 can be routed based on specific tenant policies, spaces, agreements, configurations, etc. Moreover, addressing can vary between one or more tenants. In some configurations, tenant spaces can be divided into logical segments and/or networks and separated from logical segments and/or networks associated with other tenants. Addressing, policy, security and configuration information between tenants can be managed by Controllers 116, Servers 106, Leafs 104, etc.
Configurations in Network Environment 100 can be implemented at a logical level, a hardware level (e.g., physical), and/or both. For example, configurations can be implemented at a logical and/or hardware level based on endpoint or resource attributes, such as endpoint types and/or application groups or profiles, through a software-defined network (SDN) framework (e.g., Application-Centric Infrastructure (ACI) or VMWARE NSX). To illustrate, one or more administrators can define configurations at a logical level (e.g., application or software level) through Controllers 116, which can implement or propagate such configurations through Network Environment 100. In some examples, Controllers 116 can be Application Policy Infrastructure Controllers (APICs) in an ACI framework. In other examples, Controllers 116 can be one or more management components for associated with other SDN solutions, such as NSX Managers.
Such configurations can define rules, policies, priorities, protocols, attributes, objects, etc., for routing and/or classifying traffic in Network Environment 100. For example, such configurations can define attributes and objects for classifying and processing traffic based on Endpoint Groups (EPGs), Security Groups (SGs), VM types, bridge domains (BDs), virtual routing and forwarding instances (VRFs), tenants, priorities, firewall rules, etc. Other example network objects and configurations are further described below. Traffic policies and rules can be enforced based on tags, attributes, or other characteristics of the traffic, such as protocols associated with the traffic, EPGs associated with the traffic, SGs associated with the traffic, network address information associated with the traffic, etc. Such policies and rules can be enforced by one or more elements in Network Environment 100, such as Leafs 104, Servers 106, Hypervisors 108, Controllers 116, etc. As previously explained, Network Environment 100 can be configured according to one or more particular software-defined network (SDN) solutions, such as CISCO ACI or VMWARE NSX. These example SDN solutions are briefly described below.
ACI can provide an application-centric or policy-based solution through scalable distributed enforcement. ACI supports integration of physical and virtual environments under a declarative configuration model for networks, servers, services, security, requirements, etc. For example, the ACI framework implements EPGs, which can include a collection of endpoints or applications that share common configuration requirements, such as security, QoS, services, etc. Endpoints can be virtual/logical or physical devices, such as VMs, containers, hosts, or physical servers that are connected to Network Environment 100. Endpoints can have one or more attributes such as a VM name, guest OS name, a security tag, application profile, etc. Application configurations can be applied between EPGs, instead of endpoints directly, in the form of contracts. Leafs 104 can classify incoming traffic into different EPGs. The classification can be based on, for example, a network segment identifier such as a VLAN ID, VXLAN Network Identifier (VNID), NVGRE Virtual Subnet Identifier (VSID), MAC address, IP address, etc.
In some cases, classification in the ACI infrastructure can be implemented by Application Virtual Switches (AVS), which can run on a host, such as a server or switch. For example, an AVS can classify traffic based on specified attributes, and tag packets of different attribute EPGs with different identifiers, such as network segment identifiers (e.g., VLAN ID). Finally, Leafs 104 can tie packets with their attribute EPGs based on their identifiers and enforce policies, which can be implemented and/or managed by one or more Controllers 116. Leaf 104 can classify to which EPG the traffic from a host belongs and enforce policies accordingly.
Another example SDN solution is based on VMWARE NSX. With VMWARE NSX, hosts can run a distributed firewall (DFW) which can classify and process traffic. Consider a case where three types of VMs, namely, application, database and web VMs, are put into a single layer-2 network segment. Traffic protection can be provided within the network segment based on the VM type. For example, HTTP traffic can be allowed among web VMs, and disallowed between a web VM and an application or database VM. To classify traffic and implement policies, VMWARE NSX can implement security groups, which can be used to group the specific VMs (e.g., web VMs, application VMs, database VMs). DFW rules can be configured to implement policies for the specific security groups. To illustrate, in the context of the previous example, DFW rules can be configured to block HTTP traffic between web, application, and database security groups.
Returning now to
Controllers 116 can provide centralized access to fabric information, application configuration, resource configuration, application-level configuration modeling for a software-defined network (SDN) infrastructure, integration with management systems or servers, etc. Controllers 116 can form a control plane that interfaces with an application plane via northbound APIs and a data plane via southbound APIs.
As previously noted, Controllers 116 can define and manage application-level model(s) for configurations in Network Environment 100. In some cases, application or device configurations can also be managed and/or defined by other components in the network. For example, a hypervisor or virtual appliance, such as a VM or container, can run a server or management tool to manage software and services in Network Environment 100, including configurations and settings for virtual appliances.
As illustrated above, Network Environment 100 can include one or more different types of SDN solutions, hosts, etc. For the sake of clarity and explanation purposes, various examples in the disclosure will be described with reference to an ACI framework, and Controllers 116 may be interchangeably referenced as controllers, APICs, or APIC controllers. However, it should be noted that the technologies and concepts herein are not limited to ACI solutions and may be implemented in other architectures and scenarios, including other SDN solutions as well as other types of networks which may not deploy an SDN solution.
Further, as referenced herein, the term “hosts” can refer to Servers 106 (e.g., physical or logical), Hypervisors 108, VMs 110, containers (e.g., Applications 112), etc., and can run or include any type of server or application solution. Non-limiting examples of “hosts” can include virtual switches or routers, such as distributed virtual switches (DVS), application virtual switches (AVS), vector packet processing (VPP) switches; VCENTER and NSX MANAGERS; bare metal physical hosts; HYPER-V hosts; VMs; DOCKER Containers; etc.
Endpoints 122 can be associated with respective Logical Groups 118. Logical Groups 118 can be logical entities containing endpoints (physical and/or logical or virtual) grouped together according to one or more attributes, such as endpoint type (e.g., VM type, workload type, application type, etc.), one or more requirements (e.g., policy requirements, security requirements, QoS requirements, customer requirements, resource requirements, etc.), a resource name (e.g., VM name, application name, etc.), a profile, platform or operating system (OS) characteristics (e.g., OS type or name including guest and/or host OS, etc.), an associated network or tenant, one or more policies, a tag, etc. For example, a logical group can be an object representing a collection of endpoints grouped together. To illustrate, Logical Group 1 can contain client endpoints, Logical Group 2 can contain web server endpoints, Logical Group 3 can contain application server endpoints, Logical Group N can contain database server endpoints, etc. In some examples, Logical Groups 118 are EPGs in an ACI environment and/or other logical groups (e.g., SGs) in another SDN environment.
Traffic to and/or from Endpoints 122 can be classified, processed, managed, etc., based Logical Groups 118. For example, Logical Groups 118 can be used to classify traffic to or from Endpoints 122, apply policies to traffic to or from Endpoints 122, define relationships between Endpoints 122, define roles of Endpoints 122 (e.g., whether an endpoint consumes or provides a service, etc.), apply rules to traffic to or from Endpoints 122, apply filters or access control lists (ACLs) to traffic to or from Endpoints 122, define communication paths for traffic to or from Endpoints 122, enforce requirements associated with Endpoints 122, implement security and other configurations associated with Endpoints 122, etc.
In an ACI environment, Logical Groups 118 can be EPGs used to define contracts in the ACI. Contracts can include rules specifying what and how communications between EPGs take place. For example, a contract can define what provides a service, what consumes a service, and what policy objects are related to that consumption relationship. A contract can include a policy that defines the communication path and all related elements of a communication or relationship between endpoints or EPGs. For example, a Web EPG can provide a service that a Client EPG consumes, and that consumption can be subject to a filter (ACL) and a service graph that includes one or more services, such as firewall inspection services and server load balancing.
As used herein, an “Alias” can refer to a changeable name for a given object. Thus, even if the name of an object, once created, cannot be changed, the Alias can be a field that can be changed.
As used herein, the term “Aliasing” can refer to a rule (e.g., contracts, policies, configurations, etc.) that overlaps one or more other rules. For example, Contract 1 defined in a logical model of a network can be said to be aliasing Contract 2 defined in the logical model of the network if Contract 1 overlaps Contract 1. In this example, by aliasing Contract 2, Contract 1 may render Contract 2 redundant or inoperable. For example, if Contract 1 has a higher priority than Contract 2, such aliasing can render Contract 2 redundant based on Contract 1's overlapping and higher priority characteristics.
As used herein, the term “APIC” can refer to one or more controllers (e.g., Controllers 116) in an ACI framework. The APIC can provide a unified point of automation and management, policy programming, application deployment, health monitoring for an ACI multitenant fabric. The APIC can be implemented as a single controller, a distributed controller, or a replicated, synchronized, and/or clustered controller.
As used herein, the term “BDD” can refer to a binary decision tree. A binary decision tree can be a data structure representing functions, such as Boolean functions.
As used herein, the term “BD” can refer to a bridge domain. A bridge domain can be a set of logical ports that share the same flooding or broadcast characteristics. Like a virtual LAN (VLAN), bridge domains can span multiple devices. A bridge domain can be a L2 (Layer 2) construct.
As used herein, a “Consumer” can refer to an endpoint, resource, and/or EPG that consumes a service.
As used herein, a “Context” can refer to an L3 (Layer 3) address domain that allows multiple instances of a routing table to exist and work simultaneously. This increases functionality by allowing network paths to be segmented without using multiple devices. Non-limiting examples of a context or L3 address domain can include a Virtual Routing and Forwarding (VRF) instance, a private network, and so forth.
As used herein, the term “Contract” can refer to rules or configurations that specify what and how communications in a network are conducted (e.g., allowed, denied, filtered, processed, etc.). In an ACI network, contracts can specify how communications between endpoints and/or EPGs take place. In some examples, a contract can provide rules and configurations akin to an Access Control List (ACL).
As used herein, the term “Distinguished Name” (DN) can refer to a unique name that describes an object, such as an MO, and locates its place in Management Information Model 200. In some cases, the DN can be (or equate to) a Fully Qualified Domain Name (FQDN).
As used herein, the term “Endpoint Group” (EPG) can refer to a logical entity or object associated with a collection or group of endoints as previously described with reference to
As used herein, the term “Filter” can refer to a parameter or configuration for allowing communications. For example, in a whitelist model where all communications are blocked by default, a communication must be given explicit permission to prevent such communication from being blocked. A filter can define permission(s) for one or more communications or packets. A filter can thus function similar to an ACL or Firewall rule. In some examples, a filter can be implemented in a packet (e.g., TCP/IP) header field, such as L3 protocol type, L4 (Layer 4) ports, and so on, which is used to allow inbound or outbound communications between endpoints or EPGs, for example.
As used herein, the term “L2 Out” can refer to a bridged connection. A bridged connection can connect two or more segments of the same network so that they can communicate. In an ACI framework, an L2 out can be a bridged (Layer 2) connection between an ACI fabric (e.g., Fabric 120) and an outside Layer 2 network, such as a switch.
As used herein, the term “L3 Out” can refer to a routed connection. A routed Layer 3 connection uses a set of protocols that determine the path that data follows in order to travel across networks from its source to its destination. Routed connections can perform forwarding (e.g., IP forwarding) according to a protocol selected, such as BGP (border gateway protocol), OSPF (Open Shortest Path First), EIGRP (Enhanced Interior Gateway Routing Protocol), etc.
As used herein, the term “Managed Object” (MO) can refer to an abstract representation of objects that are managed in a network (e.g., Network Environment 100). The objects can be concrete objects (e.g., a switch, server, adapter, etc.), or logical objects (e.g., an application profile, an EPG, a fault, etc.). The MOs can be network resources or elements that are managed in the network. For example, in an ACI environment, an MO can include an abstraction of an ACI fabric (e.g., Fabric 120) resource.
As used herein, the term “Management Information Tree” (MIT) can refer to a hierarchical management information tree containing the MOs of a system. For example, in ACI, the MIT contains the MOs of the ACI fabric (e.g., Fabric 120). The MIT can also be referred to as a Management Information Model (MIM), such as Management Information Model 200.
As used herein, the term “Policy” can refer to one or more specifications for controlling some aspect of system or network behavior. For example, a policy can include a named entity that contains specifications for controlling some aspect of system behavior. To illustrate, a Layer 3 Outside Network Policy can contain the BGP protocol to enable BGP routing functions when connecting Fabric 120 to an outside Layer 3 network.
As used herein, the term “Profile” can refer to the configuration details associated with a policy. For example, a profile can include a named entity that contains the configuration details for implementing one or more instances of a policy. To illustrate, a switch node profile for a routing policy can contain the switch-specific configuration details to implement the BGP routing protocol.
As used herein, the term “Provider” refers to an object or entity providing a service. For example, a provider can be an EPG that provides a service.
As used herein, the term “Subject” refers to one or more parameters in a contract for defining communications. For example, in ACI, subjects in a contract can specify what information can be communicated and how. Subjects can function similar to ACLs.
As used herein, the term “Tenant” refers to a unit of isolation in a network. For example, a tenant can be a secure and exclusive virtual computing environment. In ACI, a tenant can be a unit of isolation from a policy perspective, but does not necessarily represent a private network. Indeed, ACI tenants can contain multiple private networks (e.g., VRFs). Tenants can represent a customer in a service provider setting, an organization or domain in an enterprise setting, or just a grouping of policies.
As used herein, the term “VRF” refers to a virtual routing and forwarding instance. The VRF can define a Layer 3 address domain that allows multiple instances of a routing table to exist and work simultaneously. This increases functionality by allowing network paths to be segmented without using multiple devices. Also known as a context or private network.
Having described various terms used herein, the disclosure now returns to a discussion of Management Information Model (MIM) 200 in
The hierarchical structure of MIM 200 starts with Policy Universe 202 at the top (Root) and contains parent and child nodes 116, 204, 206, 208, 210, 212. Nodes 116, 202, 204, 206, 208, 210, 212 in the tree represent the managed objects (MOs) or groups of objects. Each object in the fabric (e.g., Fabric 120) has a unique distinguished name (DN) that describes the object and locates its place in the tree. The Nodes 116, 202, 204, 206, 208, 210, 212 can include the various MOs, as described below, which contain policies that govern the operation of the system.
Controllers 116 (e.g., APIC controllers) can provide management, policy programming, application deployment, and health monitoring for Fabric 120.
Node 204 includes a tenant container for policies that enable an administrator to exercise domain-based access control. Non-limiting examples of tenants can include:
User tenants defined by the administrator according to the needs of users. They contain policies that govern the operation of resources such as applications, databases, web servers, network-attached storage, virtual machines, and so on.
The common tenant is provided by the system but can be configured by the administrator. It contains policies that govern the operation of resources accessible to all tenants, such as firewalls, load balancers, Layer 4 to Layer 7 services, intrusion detection appliances, and so on.
The infrastructure tenant is provided by the system but can be configured by the administrator. It contains policies that govern the operation of infrastructure resources such as the fabric overlay (e.g., VXLAN). It also enables a fabric provider to selectively deploy resources to one or more user tenants. Infrastructure tenant polices can be configurable by the administrator.
The management tenant is provided by the system but can be configured by the administrator. It contains policies that govern the operation of fabric management functions used for in-band and out-of-band configuration of fabric nodes. The management tenant contains a private out-of-bound address space for the Controller/Fabric internal communications that is outside the fabric data path that provides access through the management port of the switches. The management tenant enables discovery and automation of communications with virtual machine controllers.
Node 206 can contain access policies that govern the operation of switch access ports that provide connectivity to resources such as storage, compute, Layer 2 and Layer 3 (bridged and routed) connectivity, virtual machine hypervisors, Layer 4 to Layer 7 devices, and so on. If a tenant requires interface configurations other than those provided in the default link, Cisco Discovery Protocol (CDP), Link Layer Discovery Protocol (LLDP), Link Aggregation Control Protocol (LACP), or Spanning Tree Protocol (STP), an administrator can configure access policies to enable such configurations on the access ports of Leafs 104.
Node 206 can contain fabric policies that govern the operation of the switch fabric ports, including such functions as Network Time Protocol (NTP) server synchronization, Intermediate System-to-Intermediate System Protocol (IS-IS), Border Gateway Protocol (BGP) route reflectors, Domain Name System (DNS) and so on. The fabric MO contains objects such as power supplies, fans, chassis, and so on.
Node 208 can contain VM domains that group VM controllers with similar networking policy requirements. VM controllers can share virtual space (e.g., VLAN or VXLAN space) and application EPGs. Controllers 116 communicate with the VM controller to publish network configurations such as port groups that are then applied to the virtual workloads.
Node 210 can contain Layer 4 to Layer 7 service integration life cycle automation framework that enables the system to dynamically respond when a service comes online or goes offline. Policies can provide service device package and inventory management functions.
Node 212 can contain access, authentication, and accounting (AAA) policies that govern user privileges, roles, and security domains of Fabric 120.
The hierarchical policy model can fit well with an API, such as a REST API interface. When invoked, the API can read from or write to objects in the MIT. URLs can map directly into distinguished names that identify objects in the MIT. Data in the MIT can be described as a self-contained structured tree text document encoded in XML or JSON, for example.
Tenant portion 204A of MIM 200 can include various entities, and the entities in Tenant Portion 204A can inherit policies from parent entities. Non-limiting examples of entities in Tenant Portion 204A can include Filters 240, Contracts 236, Outside Networks 222, Bridge Domains 230, VRF Instances 234, and Application Profiles 224.
Bridge Domains 230 can include Subnets 232. Contracts 236 can include Subjects 238. Application Profiles 224 can contain one or more EPGs 226. Some applications can contain multiple components. For example, an e-commerce application could require a web server, a database server, data located in a storage area network, and access to outside resources that enable financial transactions. Application Profile 224 contains as many (or as few) EPGs as necessary that are logically related to providing the capabilities of an application.
EPG 226 can be organized in various ways, such as based on the application they provide, the function they provide (such as infrastructure), where they are in the structure of the data center (such as DMZ), or whatever organizing principle that a fabric or tenant administrator chooses to use.
EPGs in the fabric can contain various types of EPGs, such as application EPGs, Layer 2 external outside network instance EPGs, Layer 3 external outside network instance EPGs, management EPGs for out-of-band or in-band access, etc. EPGs 226 can also contain Attributes 228, such as encapsulation-based EPGs, IP-based EPGs, or MAC-based EPGs.
As previously mentioned, EPGs can contain endpoints (e.g., EPs 122) that have common characteristics or attributes, such as common policy requirements (e.g., security, virtual machine mobility (VMM), QoS, or Layer 4 to Layer 7 services). Rather than configure and manage endpoints individually, they can be placed in an EPG and managed as a group.
Policies apply to EPGs, including the endpoints they contain. An EPG can be statically configured by an administrator in Controllers 116, or dynamically configured by an automated system such as VCENTER or OPENSTACK.
To activate tenant policies in Tenant Portion 204A, fabric access policies should be configured and associated with tenant policies. Access policies enable an administrator to configure other network configurations, such as port channels and virtual port channels, protocols such as LLDP, CDP, or LACP, and features such as monitoring or diagnostics.
Access Portion 206A can contain fabric and infrastructure access policies. Typically, in a policy model, EPGs are coupled with VLANs. For traffic to flow, an EPG is deployed on a leaf port with a VLAN in a physical, VMM, L2 out, L3 out, or Fiber Channel domain, for example.
Access Portion 206A thus contains Domain Profile 236 which can define a physical, VMM, L2 out, L3 out, or Fiber Channel domain, for example, to be associated to the EPGs. Domain Profile 236 contains VLAN Instance Profile 238 (e.g., VLAN pool) and Attacheable Access Entity Profile (AEP) 240, which are associated directly with application EPGs. The AEP 240 deploys the associated application EPGs to the ports to which it is attached, and automates the task of assigning VLANs. While a large data center can have thousands of active VMs provisioned on hundreds of VLANs, Fabric 120 can automatically assign VLAN IDs from VLAN pools. This saves time compared with trunking down VLANs in a traditional data center.
L_Model 270A is the logical representation of the objects and their relationships in MIM 200. L_Model 270A can be generated by Controllers 116 based on configurations entered in Controllers 116 for the network, and thus represents the configurations of the network at Controllers 116. This is the declaration of the “end-state” expression that is desired when the elements of the network entities (e.g., applications) are connected and Fabric 120 is provisioned by Controllers 116. In other words, because L_Model 270A represents the configurations entered in Controllers 116, including the objects and relationships in MIM 200, it can also reflect the “intent” of the administrator: how the administrator wants the network and network elements to behave.
LR_Model 270B is the abstract model expression that Controllers 116 (e.g., APICs in ACI) resolve from L_Model 270A. LR_Model 270B can thus provide the elemental configuration components that would be delivered to the physical infrastructure (e.g., Fabric 120) to execute one or more policies. For example, LR_Model 270B can be delivered to Leafs 104 in Fabric 120 to configure Leafs 104 for communication with attached Endpoints 122.
Li_Model 272 is a switch-level or switch-specific model obtained from Logical Model 270A and/or Resolved Model 270B. For example, Li_Model 272 can represent the portion of L_Model 270A and/or LR_Model 270B pertaining to a specific switch or router i. To illustrate, Li_Model 272 L1 can represent the portion of L_Model 270A and/or LR_Model 270B pertaining to Leaf 1 (104). Thus, Li_Model 272 can be generated from L_Model 270A and/or LR_Model 270B for one or more switch or routers (e.g., Leafs 104 and/or Spines 102) on Fabric 120.
Ci_Model 274 is the actual in-state configuration at the individual fabric member i (e.g., switch i). In other words, Ci_Model 274 is a switch-level or switch-specific model that is based on Li_Model 272. For example, Controllers 116 can deliver Li_Model 272 to Leaf 1 (104). Leaf 1 (104) can take Li_Model 272, which can be specific to Leaf 1 (104), and render the policies in Li_Model 272 into a concrete model, Ci_Model 274, that runs on Leaf 1 (104). Leaf 1 (104) can render Li_Model 272 via the OS on Leaf 1 (104), for example. Thus, Ci_Model 274 can be analogous to compiled software, as it is the form of Li_Model 272 that the switch OS at Leaf 1 (104) can execute.
Hi_Model 276 is also a switch-level or switch-specific model for switch i, but is based on Ci_Model 274 for switch i. Hi_Model 276 is the actual configuration (e.g., rules) stored or rendered on the hardware or memory (e.g., TCAM memory) at the individual fabric member i (e.g., switch i). For example, Hi_Model 276 can represent the configurations (e.g., rules) which Leaf 1 (104) stores or renders on the hardware (e.g., TCAM memory) of Leaf 1 (104) based on Ci_Model 274 at Leaf 1 (104). The switch OS at Leaf 1 (104) can render or execute Ci_Model 274, and Leaf 1 (104) can store or render the configurations from Ci_Model in storage, such as the memory or TCAM at Leaf 1 (104). The configurations from Hi_Model 276 stored or rendered by Leaf 1 (104) represent the configurations that will be implemented by Leaf 1 (104) when processing traffic.
While Models 272, 274, 276 are shown as device-specific models, similar models can be generated or aggregated for a collection of fabric members (e.g., Leafs 104 and/or Spines 102) in Fabric 120. When combined, device-specific models, such as Model 272, Model 274, and/or Model 276, can provide a representation of Fabric 120 that extends beyond a particular device. For example, in some cases, Li_Model 272, Ci_Model 272, and/or Hi_Model 272 associated with some or all individual fabric members (e.g., Leafs 104 and Spines 102) can be combined or aggregated to generate one or more aggregated models based on the individual fabric members.
As referenced herein, the terms H Model, T Model, and TCAM Model can be used interchangeably to refer to a hardware model, such as Hi_Model 276. For example, Ti Model, Hi_Model and TCAMi Model may be used interchangeably to refer to Hi_Model 276.
Models 270A, 270B, 272, 274, 276 can provide representations of various aspects of the network or various configuration stages for MIM 200. For example, one or more of Models 270A, 270B, 272, 274, 276 can be used to generate Underlay Model 278 representing one or more aspects of Fabric 120 (e.g., underlay topology, routing, etc.), Overlay Model 280 representing one or more aspects of the overlay or logical segment(s) of Network Environment 100 (e.g., COOP, MPBGP, tenants, VRFs, VLANs, VXLANs, virtual applications, VMs, hypervisors, virtual switching, etc.), Tenant Model 282 representing one or more aspects of Tenant portion 204A in MIM 200 (e.g., security, forwarding, service chaining, QoS, VRFs, BDs, Contracts, Filters, EPGs, subnets, etc.), Resources Model 284 representing one or more resources in Network Environment 100 (e.g., storage, computing, VMs, port channels, physical elements, etc.), etc.
In general, L_Model 270A can be the high-level expression of what exists in the LR_Model 270B, which should be present on the concrete devices as Ci_Model 274 and Hi_Model 276 expression. If there is any gap between the models, there may be inconsistent configurations or problems.
Assurance Appliance 300 can run on one or more Servers 106, VMs 110, Hypervisors 108, EPs 122, Leafs 104, Controllers 116, or any other system or resource. For example, Assurance Appliance 300 can be a logical service or application running on one or more VMs 110 in Network Environment 100.
The Assurance Appliance 300 can include Data Framework 308, which can be based on, for example, APACHE APEX and HADOOP. In some cases, assurance checks can be written as individual operators that reside in Data Framework 308. This enables a natively horizontal scale-out architecture that can scale to arbitrary number of switches in Fabric 120 (e.g., ACI fabric).
Assurance Appliance 300 can poll Fabric 120 at a configurable periodicity (e.g., an epoch). The analysis workflow can be setup as a DAG (Directed Acyclic Graph) of Operators 310, where data flows from one operator to another and eventually results are generated and persisted to Database 302 for each interval (e.g., each epoch).
The north-tier implements API Server (e.g., APACHE Tomcat and Spring framework) 304 and Web Server 306. A graphical user interface (GUI) interacts via the APIs exposed to the customer. These APIs can also be used by the customer to collect data from Assurance Appliance 300 for further integration into other tools.
Operators 310 in Data Framework 308 (e.g., APEX/Hadoop) can together support assurance operations. Below are non-limiting examples of assurance operations that can be performed by Assurance Appliance 300 via Operators 310.
Assurance Appliance 300 can check to make sure the configurations or specification from L_Model 270A, which may reflect the user's intent for the network, including for example the security policies and customer-configured contracts, are correctly implemented and/or rendered in Li_Model 272, Ci_Model 274, and Hi_Model 276, and thus properly implemented and rendered by the fabric members (e.g., Leafs 104), and report any errors, contract violations, or irregularities found.
Assurance Appliance 300 can check for issues in the specification of the user's intent or intents (e.g., identify contradictory or conflicting policies in L_Model 270A).
TCAM is a scarce resource in the fabric (e.g., Fabric 120). However, Assurance Appliance 300 can analyze the TCAM utilization by the network data (e.g., Longest Prefix Match (LPM) tables, routing tables, VLAN tables, BGP updates, etc.), Contracts, Logical Groups 118 (e.g., EPGs), Tenants, Spines 102, Leafs 104, and other dimensions in Network Environment 100 and/or objects in MIM 200, to provide a network operator or user visibility into the utilization of this scarce resource. This can greatly help for planning and other optimization purposes.
Assurance Appliance 300 can validate that the fabric (e.g. fabric 120) has no inconsistencies in the Endpoint information registered (e.g., two leafs announcing the same endpoint, duplicate subnets, etc.), among other such checks.
Assurance Appliance 300 can validate that BDs, VRFs, subnets (both internal and external), VLANs, contracts, filters, applications, EPGs, etc., are correctly programmed.
Assurance Appliance 300 can validate that infrastructure routing (e.g., IS-IS protocol) has no convergence issues leading to black holes, loops, flaps, and other problems.
The network fabric (e.g., Fabric 120) can interface with other external networks and provide connectivity to them via one or more protocols, such as Border Gateway Protocol (BGP), Open Shortest Path First (OSPF), etc. The learned routes are advertised within the network fabric via, for example, MP-BGP. These checks can ensure that a route reflection service via, for example, MP-BGP (e.g., from Border Leaf) does not have health issues.
Assurance Appliance 300 can validate rules in the specification of the network (e.g., L_Model 270A) are complete and do not have inconsistencies or other problems. MOs in the MIM 200 can be checked by Assurance Appliance 300 through syntactic and semantic checks performed on L_Model 270A and/or the associated configurations of the MOs in MIM 200. Assurance Appliance 300 can also verify that unnecessary, stale, unused or redundant configurations, such as contracts, are removed.
Topology Explorer 312 is configured to discover nodes in Fabric 120, such as Controllers 116, Leafs 104, Spines 102, etc. Topology Explorer 312 can additionally detect a majority election performed amongst Controllers 116, and determine whether a quorum exists amongst Controllers 116. If no quorum or majority exists, Topology Explorer 312 can trigger an event and alert a user that a configuration or other error exists amongst Controllers 116 that is preventing a quorum or majority from being reached. Topology Explorer 312 can detect Leafs 104 and Spines 102 that are part of Fabric 120 and publish their corresponding out-of-band management network addresses (e.g., IP addresses) to downstream services. This can be part of the topological view that is published to the downstream services at the conclusion of Topology Explorer's 312 discovery epoch (e.g., 5 minutes, or some other specified interval).
Unified Collector 314 can receive the topological view from Topology Explorer 312 and use the topology information to collect information for network assurance from Fabric 120. Such information can include L_Model 270A and/or LR_Model 270B from Controllers 116, switch software configurations (e.g., Ci_Model 274) from Leafs 104 and/or Spines 102, hardware configurations (e.g., Hi_Model 276) from Leafs 104 and/or Spines 102, etc. Unified Collector 314 can collect Ci_Model 274 and Hi_Model 276 from individual fabric members (e.g., Leafs 104 and Spines 102).
Unified Collector 314 can poll the devices that Topology Explorer 312 discovers in order to collect data from Fabric 120 (e.g., from the constituent members of the fabric). Unified Collector 314 can collect the data using interfaces exposed by Controller 116 and/or switch software (e.g., switch OS), including, for example, a Representation State Transfer (REST) Interface and a Secure Shell (SSH) Interface.
In some cases, Unified Collector 314 collects L_Model 270A, LR_Model 270B, and/or Ci_Model 274 via a REST API, and the hardware information (e.g., configurations, tables, fabric card information, rules, routes, etc.) via SSH using utilities provided by the switch software, such as virtual shell (VSH or VSHELL) for accessing the switch command-line interface (CLI) or VSH_LC shell for accessing runtime state of the line card.
Unified Collector 314 can poll other information from Controllers 116, including: topology information, tenant forwarding/routing information, tenant security policies, contracts, interface policies, physical domain or VMM domain information, OOB (out-of-band) management IP's of nodes in the fabric, etc.
Unified Collector 314 can also poll other information from Leafs 104 and Spines 102, such as: Ci Models 274 for VLANs, BDs, security policies, Link Layer Discovery Protocol (LLDP) connectivity information of Leafs 104 and/or Spines 102, endpoint information from EPM/COOP, fabric card information from Spines 102, routing information base (RIB) tables, forwarding information base (FIB) tables from Leafs 104 and/or Spines 102, security group hardware tables (e.g., TCAM tables) from switches, etc.
Assurance Appliance 300 can run one or more instances of Unified Collector 314. For example, Assurance Appliance 300 can run one, two, three, or more instances of Unified Collector 314. The task of data collecting for each node in the topology (e.g., Fabric 120 including Spines 102, Leafs 104, Controllers 116, etc.) can be sharded or load balanced, to a unique instance of Unified Collector 314. Data collection across the nodes can thus be performed in parallel by one or more instances of Unified Collector 314. Within a given node, commands and data collection can be executed serially. Assurance Appliance 300 can control the number of threads used by each instance of Unified Collector 314 to poll data from Fabric 120.
Data collected by Unified Collector 314 can be compressed and sent to downstream services. In some examples, Unified Collector 314 can collect data in an online fashion or real-time fashion, and send the data downstream, as it is collected, for further analysis. In some examples, Unified Collector 314 can collect data in an offline fashion, and compile the data for later analysis or transmission.
Assurance Appliance 300 can contact Controllers 116, Spines 102, Leafs 104, and other nodes to collect various types of data. In some scenarios, Assurance Appliance 300 may experience a failure (e.g., connectivity problem, hardware or software error, etc.) that prevents it from being able to collect data for a period of time. Assurance Appliance 300 can handle such failures seamlessly, and generate events based on such failures.
Switch Logical Policy Generator 316 can receive L_Model 270A and/or LR_Model 270B from Unified Collector 314 and calculate Li_Model 272 for each network device i (e.g., switch i) in Fabric 120. For example, Switch Logical Policy Generator 316 can receive L_Model 270A and/or LR_Model 270B and generate Li_Model 272 by projecting a logical model for each individual node i (e.g., Spines 102 and/or Leafs 104) in Fabric 120. Switch Logical Policy Generator 316 can generate Li_Model 272 for each switch in Fabric 120, thus creating a switch logical model based on L_Model 270A for each switch.
Switch Logical Configuration Generator 316 can also perform change analysis and generate lint events or records for problems discovered in L_Model 270A and/or LR_Model 270B. The lint events or records can be used to generate alerts for a user or network operator.
Policy Operator 318 can receive Ci_Model 274 and Hi_Model 276 for each switch from Unified Collector 314, and Li_Model 272 for each switch from Switch Logical Policy Generator 316, and perform assurance checks and analysis (e.g., security adherence checks, TCAM utilization analysis, etc.) based on Ci_Model 274, Hi_Model 276, and Li_Model 272. Policy Operator 318 can perform assurance checks on a switch-by-switch basis by comparing one or more of the models.
Returning to Unified Collector 314, Unified Collector 314 can also send L_Model 270A and/or LR_Model 270B to Routing Policy Parser 320, and Ci_Model 274 and Hi_Model 276 to Routing Parser 326.
Routing Policy Parser 320 can receive L_Model 270A and/or LR_Model 270B and parse the model(s) for information that may be relevant to downstream operators, such as Endpoint Checker 322 and Tenant Routing Checker 324. Similarly, Routing Parser 326 can receive Ci_Model 274 and Hi_Model 276 and parse each model for information for downstream operators, Endpoint Checker 322 and Tenant Routing Checker 324.
After Ci_Model 274, Hi_Model 276, L_Model 270A and/or LR_Model 270B are parsed, Routing Policy Parser 320 and/or Routing Parser 326 can send cleaned-up protocol buffers (Proto Buffs) to the downstream operators, Endpoint Checker 322 and Tenant Routing Checker 324. Endpoint Checker 322 can then generate events related to Endpoint violations, such as duplicate IPs, APIPA, etc., and Tenant Routing Checker 324 can generate events related to the deployment of BDs, VRFs, subnets, routing table prefixes, etc.
Static Policy Analyzer 360 can include one or more of the Operators 310 executed or hosted in Assurance Appliance 300. However, in other configurations, Static Policy Analyzer 360 can run one or more operators or engines that are separate from Operators 310 and/or Assurance Appliance 300. For example, Static Policy Analyzer 360 can be a VM, a cluster of VMs, or a collection of endpoints in a service function chain.
Static Policy Analyzer 360 can receive as input L_Model 270A from Logical Model Collection Process 366 and Rules 368 defined for each feature (e.g., object) in L_Model 270A. Rules 368 can be based on objects, relationships, definitions, configurations, and any other features in MIM 200. Rules 368 can specify conditions, relationships, parameters, and/or any other information for identifying configuration violations or issues.
Moreover, Rules 368 can include information for identifying syntactic violations or issues. For example, Rules 368 can include one or more rules for performing syntactic checks. Syntactic checks can verify that the configuration of L_Model 270A is complete, and can help identify configurations or rules that are not being used. Syntactic checks can also verify that the configurations in the hierarchical MIM 200 are complete (have been defined) and identify any configurations that are defined but not used. To illustrate, Rules 368 can specify that every tenant in L_Model 270A should have a context configured configured; every contract in L_Model 270A should specify a provider EPG and a consumer EPG; every contract in L_Model 270A should specify a subject, filter, and/or port; etc.
Rules 368 can also include rules for performing semantic checks and identifying semantic violations or issues. Semantic checks can check conflicting rules or configurations. For example, Rule1 and Rule2 can have aliasing issues, Rule1 can be more specific than Rule2 and thereby create conflicts/issues, etc. Rules 368 can define conditions which may result in aliased rules, conflicting rules, etc. To illustrate, Rules 368 can specify that an allow policy for a specific communication between two objects can conflict with a deny policy for the same communication between two objects if the allow policy has a higher priority than the deny policy, or a rule for an object renders another rule unnecessary.
Static Policy Analyzer 360 can apply Rules 368 to L_Model 270A to check configurations in L_Model 270A and output Configuration Violation Events 370 (e.g., alerts, logs, notifications, etc.) based on any issues detected. Configuration Violation Events 370 can include semantic or semantic problems, such as incomplete configurations, conflicting configurations, aliased rules, unused configurations, errors, policy violations, misconfigured objects, incomplete configurations, incorrect contract scopes, improper object relationships, etc.
In some cases, Static Policy Analyzer 360 can iteratively traverse each node in a tree generated based on L_Model 270A and/or MIM 200, and apply Rules 368 at each node in the tree to determine if any nodes yield a violation (e.g., incomplete configuration, improper configuration, unused configuration, etc.). Static Policy Analyzer 360 can output Configuration Violation Events 370 when it detects any violations.
Each block shown in
At step 400, Assurance Appliance 300 can collect data and obtain models associated with Network Environment 100. The models can include Models 270A-B, 272, 274, 276. The data can include fabric data (e.g., topology, switch, interface policies, application policies, EPGs, etc.), network configurations (e.g., BDs, VRFs, L2 Outs, L3 Outs, protocol configurations, etc.), security configurations (e.g., contracts, filters, etc.), service chaining configurations, routing configurations, and so forth. Other information collected or obtained can include, for example, network data (e.g., RIB/FIB, VLAN, MAC, ISIS, DB, BGP, OSPF, ARP, VPC, LLDP, MTU, QoS, etc.), rules and tables (e.g., TCAM rules, ECMP tables, etc.), endpoint dynamics (e.g., EPM, COOP EP DB, etc.), statistics (e.g., TCAM rule hits, interface counters, bandwidth, etc.).
At step 402, Assurance Appliance 300 can analyze and model the received data and models. For example, Assurance Appliance 300 can perform formal modeling and analysis, which can involve determining equivalency between models, including configurations, policies, etc.
At step 404, Assurance Appliance 300 can generate one or more smart events. Assurance Appliance 300 can generate smart events using deep object hierarchy for detailed analysis, such as Tenants, switches, VRFs, rules, filters, routes, prefixes, ports, contracts, subjects, etc.
At step 406, Assurance Appliance 300 can visualize the smart events, analysis and/or models. Assurance Appliance 300 can display problems and alerts for analysis and debugging, in a user-friendly GUI.
Each policy object represents a reusable, collection of elements, which enables a particular functionality and/or otherwise defines a particular component or function, etc., of the network environment 100. Example policy objects include EPGs, Virtual Routing and Forwarding (VRF), contract(s), filter(s), etc. Any given object can be used by other objects and network policies. Thus, when a given policy object is referenced in a given network policy, the policy object may become integrated, meaning that if a change is made to the definition of the policy object, the change is reflected in all other policy objects and network policies that reference the changed object. Stated differently, the relationship among policy objects dictates how a network policy may be realized. So if an object is absent or ill-represented in any of the controller(s), switches and/or TCAM layers, all of the EPG pairs that rely on that policy object will be negatively impacted. Thus, these policy objects on which a set of EPG pairs rely are considered shared risks in the network policy deployment.
As noted above, a controller (e.g., controller 116) manages deployment of a network policy, by for example, compiling a network policy and distributing corresponding instructions to switches that directly connect to endpoints belonging to the EPGs in the network policy. Upon receipt of the instructions, the switch (e.g., via switch agent) updates its logical view and renders ACL rules into the TCAM. During the deployment process, there are two major locations that eventually cause the failure of a TCAM rule update: 1) when updating from controller to switch; and 2) when updating from switch to TCAM. The first may cause global faults in the network whereas the second is generally associated with local faults. For instance, if the controller cannot reach out to a large number of switches for some reason, the policy objects associated with the unreachable switches are not updated. Alternatively, when one switch is unreachable (e.g., a switch agent misbehaves or TCAM has hardware glitches) the scope of risk model should be restricted to a particular switch level.
Referring specifically to the steps of
At 504, the obtained TCAM rules are processed to conduct an equivalence check between the logical model and/or the network policy (e.g., at the controller 116) and the collected TCAM rules (i.e., the hardware model to TCAM model). The equivalence check resolves the relationships among policy objects in the network policy and generates per-switch logical TCAM rules (L-type). A logical TCAM rule is treated as missing, if it meets any of the following two conditions: (1) a corresponding switch TCAM rule is not present; and (2) the rule is present but violates the network policy. An example of a network policy violation is that a rule configured as “allow” is programmed as “deny” in a switch TCAM rules and vice versa. Thus, the equivalence check detects inconsistency between the desired state (i.e., the network policy) and actual state (i.e., the collected TCAM rules) and produces or otherwise outputs missing TCAM rules (the rules that should have been deployed in the TCAM but are absent from the TCAM and hence, violating the network policy) across switches.
At 506, one or more risk models are generated that represents a risk(s) associated with the deployment of the network policy throughout or across the various components of the network environment 100, such as the controller(s), switch(es) and TCAM(s).
In some instances, a Switch Risk Model may be generated. The Switch Risk Model identifies policy objects that violate a network policy or deteriorate communication between endpoints at a switch level. A Switch Risk Model consists of shared risks (i.e., policy objects) and the elements (i.e., EPG pair) that can be impacted by the shared risks on a per-switch basis. The model is built from a network policy and the physical locations of endpoints belonging to EPGs in the network.
Referring generally again to
In the Controller Risk Model, a switch ID and an EPG pair form a triplet (on the left-hand side) is illustrated. A triplet has edges to policy objects that the EPG pair relies on in that specified switch. Since the same policy object can be present in more than one switch, an EPG pair in multiple switches can have an edge to the object.
Referring again to
For example, an EPG pair (or switch-EPG tuples associated with the EPG pair across switches) will be marked as a fail, if it has at least one failed edge between the pair and a policy object. Otherwise, the EPG pair is considered a success. Each EPG pair node marked as fail is an observation. A set of observations is called a failure signature. Any policy object shared across multiple EPG pairs becomes a shared risk. If all edges to a policy object are marked as fail, it is highly likely that the failure of deploying that policy object explains the observations present in the failure signature, and such a policy object is added to a set called a hypothesis.
At 510, based on the augmented risk models (i.e., the Switch risk model and/or the Controller risk model), a subset of the policy objects are identified that are most likely to be responsible for or otherwise associated with faults or failures in the network. Initially, in some instances, a greedy algorithm may be executed to identify policy objects to maximize two utility values—(i) hit ratio and (ii) coverage ratio—computed for each policy object.
In such an algorithm, Let Gi be a set of EPG pairs that depend on a shared risk i, Oi be a subset of Gi in which EPG pairs are marked as fail (observations), and F be the failure signature, a set that contains all observations, i.e., F=u Oi for all i. For the given shared risk i, a hit ratio, hi is defined as:
hi=|Gi∩Oi|/|Gi|
Stated differently, a hit ratio is a fraction of EPG pairs that are observations out of all EPG pairs that depend on a shared risk. A hit ratio is 1 when all EPG pairs that depend on a shared risk are marked as fail. And a coverage ratio, ci is defined as:
ci=|Gi∩O
i
|/|F|
A coverage ratio denotes a fraction of failed EPG pairs associated with a shared risk from the failure signature.
The greedy algorithm first identifies a set of policy objects (i.e., shared risks) to examine based on the values of hit ratio. In an ideal case, those values should be 1. However, in practice, perfect failure information is usually unavailable to the algorithm, which makes the hit ratio less than 1. To cope with such a real constraint, the algorithm may inspect shared risks whose hit ratio is above some fixed threshold value.
Next, given the set of shared risks chosen from the previous step, the algorithm selects the shared risks (i.e., policy objects) with the highest coverage ratio values. In particular, the generated risk models (operation 504) may be provided as input to a Fault Location Algorithm to identify a subset of policy objects (e.g., from the pool of policy objects determined by the greedy algorithm) that are likely responsible for faults and/or failures in the network. Algorithm 1 and 2 included below provide an illustration of a Fault Location Algorithm according to one embodiment.
F: failure signature, R: risk model, C: change logs
P: unexplained set, Q: explained set, H: hypothesis
K: a set of shared risks
Algorithm 1 shows the core part of the Fault Localization Algorithm. The algorithm takes failure signature F and risk model R as input. F has a set of observations, e.g., EPG pairs marked as a fault or failure in the risk model, such as in the Switch Risk Model. For each observation in F, the algorithm obtains a list of policy objects with fail edges in the observation and computes utility values (i.e., hit and coverage ratios) for all those objects (lines 6-10). Then, based on the utility values of policy objects (i.e., shared risks) in the model, the Fault Localization Algorithm picks a subset of the policy objects (i.e., shared risks) and treats them as faulty (line 11 and Algorithm 2).
In Algorithm 2, if the hit ratio of a policy object (i.e., shared risk) is 1, the risk is included in a candidate risk set (lines 3-7); and then from the candidate set, the policy objects (i.e., shared risks) that have the highest coverage ratio values are chosen as a set of policy objects (i.e., shared risks) that covers a maximum number of unexplained observations (line 8). This group is referred to in Algorithm 1 as faultySet.
If faultySet is not empty, all EPG pairs that have an edge to any policy object (i.e., shared risk) in the faultySet are pruned from the model (lines 15-16), and failed EPG pairs (observations) are moved from unexplained to explained (line 17). Finally, all the policy object (i.e., shared risk) in the faultySet are added to the hypothesis set, H. This process repeats until either there are no more observations left unexplained or when faultySet is empty.
There are chances that some observations remain unexplained because the shared risks associated with those observations have a hit ratio less than 1 and thus are not selected during the above candidate selection procedure. To handle the remaining unexplained observations, the Fault Location Algorithm searches the change logs of the controllers (i.e., logs identifying changes made to objects), and selects the objects to which some actions are recently applied (lines 21-24 in Algorithm 1).
The lines 4-19 in Algorithm 1 cover the following: (i) filter F2 is identified as a candidate because it has the highest coverage ratio among the shared risks with a hit ratio of 1; (ii) all the EPG pairs that depend on F2 are pruned from the model; (iii) and F2 is added to hypothesis. The lines 21-24 ensure that the algorithm adds filter F3 (assuming F3 is lately modified) to hypothesis since there are no shared risks with a hit ratio of 1.
Various examples of fault localization will now be provided.
TCAM overflow: Due to TCAM overflow, several filters may not be deployed at TCAM. Assume a switch “A” is being tested for faults. Switch A generated fault logs that indicate that there is a TCAM overflow when TCAM utilization is beyond a certain level. To identify the fault, the system first localized the faulty filter objects with risk models, correlated them with the change logs for ‘add filter’ instruction, and then correlated the change logs with the fault logs. The system included data describing the fault signature of TCAM overflow, so the system can automatically match the fault logs with that signature and tag those failed filters accordingly.
Unresponsive switch: In this use case, assume switch A became unresponsive while the controller was sending the ‘add filter’ instructions to the switch. The equivalence checker reported that the rules associated with some filters are missing. Then, the SCOUT algorithm localized those filters as faulty objects. Using filter creation times from the change logs and the fault logs that indicate the switch was inactive (both maintained at the controller), the correlation engine was able to detect that filters were created when the switch was inactive.
Too many missing rules: As a variant of the above scenario, a policy with a large number of policy objects was pushed onto an unresponsive switch. In such a scenario, more than 300K missing rules were reported by the equivalence checker. Without fault localization, it is extremely challenging for network admins to correlate and identify the set of underlying objects that are fundamentally responsible for the problem. The algorithms discussed herein were able to automatically identify and report the unresponsive switch as the root cause behind these huge
The disclosure now turns to
The interfaces 702 are typically provided as modular interface cards (sometimes referred to as “line cards”). Generally, they control the sending and receiving of data packets over the network and sometimes support other peripherals used with the network device 700. Among the interfaces that may be provided are Ethernet interfaces, frame relay interfaces, cable interfaces, DSL interfaces, token ring interfaces, and the like. In addition, various very high-speed interfaces may be provided such as fast token ring interfaces, wireless interfaces, Ethernet interfaces, Gigabit Ethernet interfaces, ATM interfaces, HSSI interfaces, POS interfaces, FDDI interfaces, WIFI interfaces, 3G/4G/5G cellular interfaces, CAN BUS, LoRA, and the like. Generally, these interfaces may include ports appropriate for communication with the appropriate media. In some cases, they may also include an independent processor and, in some instances, volatile RAM. The independent processors may control such communications intensive tasks as packet switching, media control, signal processing, crypto processing, and management. By providing separate processors for the communications intensive tasks, these interfaces allow the master microprocessor 704 to efficiently perform routing computations, network diagnostics, security functions, etc.
Although the system shown in
Regardless of the network device's configuration, it may employ one or more memories or memory modules (including memory 706) configured to store program instructions for the general-purpose network operations and mechanisms for roaming, route optimization and routing functions described herein. The program instructions may control the operation of an operating system and/or one or more applications, for example. The memory or memories may also be configured to store tables such as mobility binding, registration, and association tables, etc. Memory 706 could also hold various software containers and virtualized execution environments and data.
The network device 700 can also include an application-specific integrated circuit (ASIC), which can be configured to perform routing and/or switching operations. The ASIC can communicate with other components in the network device 700 via the bus 710, to exchange data and signals and coordinate various types of operations by the network device 700, such as routing, switching, and/or data storage operations, for example.
To enable user interaction with the computing device 800, an input device 845 can represent any number of input mechanisms, such as a microphone for speech, a touch-sensitive screen for gesture or graphical input, keyboard, mouse, motion input, speech and so forth. An output device 835 can also be one or more of a number of output mechanisms known to those of skill in the art. In some instances, multimodal systems can enable a user to provide multiple types of input to communicate with the computing device 800. The communications interface 840 can generally govern and manage the user input and system output. There is no restriction on operating on any particular hardware arrangement and therefore the basic features here may easily be substituted for improved hardware or firmware arrangements as they are developed.
Storage device 830 is a non-volatile memory and can be a hard disk or other types of computer readable media which can store data that are accessible by a computer, such as magnetic cassettes, flash memory cards, solid state memory devices, digital versatile disks, cartridges, random access memories (RAMs) 825, read only memory (ROM) 820, and hybrids thereof.
The storage device 830 can include services 832, 834, 836 for controlling the processor 810. Other hardware or software modules are contemplated. The storage device 830 can be connected to the system connection 805. In one aspect, a hardware module that performs a particular function can include the software component stored in a computer-readable medium in connection with the necessary hardware components, such as the processor 810, connection 805, output device 835, and so forth, to carry out the function.
For clarity of explanation, in some instances the present technology may be presented as including individual functional blocks including functional blocks comprising devices, device components, steps or routines in a method embodied in software, or combinations of hardware and software.
In some embodiments the computer-readable storage devices, mediums, and memories can include a cable or wireless signal containing a bit stream and the like. However, when mentioned, non-transitory computer-readable storage media expressly exclude media such as energy, carrier signals, electromagnetic waves, and signals per se.
Methods according to the above-described examples can be implemented using computer-executable instructions that are stored or otherwise available from computer readable media. Such instructions can comprise, for example, instructions and data which cause or otherwise configure a general purpose computer, special purpose computer, or special purpose processing device to perform a certain function or group of functions. Portions of computer resources used can be accessible over a network. The computer executable instructions may be, for example, binaries, intermediate format instructions such as assembly language, firmware, or source code. Examples of computer-readable media that may be used to store instructions, information used, and/or information created during methods according to described examples include magnetic or optical disks, flash memory, USB devices provided with non-volatile memory, networked storage devices, and so on.
Devices implementing methods according to these disclosures can comprise hardware, firmware and/or software, and can take any of a variety of form factors. Typical examples of such form factors include laptops, smart phones, small form factor personal computers, personal digital assistants, rackmount devices, standalone devices, and so on. Functionality described herein also can be embodied in peripherals or add-in cards. Such functionality can also be implemented on a circuit board among different chips or different processes executing in a single device, by way of further example.
The instructions, media for conveying such instructions, computing resources for executing them, and other structures for supporting such computing resources are means for providing the functions described in these disclosures.
Although a variety of examples and other information was used to explain aspects within the scope of the appended claims, no limitation of the claims should be implied based on particular features or arrangements in such examples, as one of ordinary skill would be able to use these examples to derive a wide variety of implementations. Further and although some subject matter may have been described in language specific to examples of structural features and/or method steps, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to these described features or acts. For example, such functionality can be distributed differently or performed in components other than those identified herein. Rather, the described features and steps are disclosed as examples of components of systems and methods within the scope of the appended claims.
Claim language reciting “at least one of” refers to at least one of a set and indicates that one member of the set or multiple members of the set satisfy the claim. For example, claim language reciting “at least one of A and B” means A, B, or A and B.
This application claims priority to and the benefit of U.S. Provisional Patent Application No. 62/513,093, filed on May 31, 2017, which is incorporated herein by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 62513093 | May 2017 | US |
