Patent Application
20120110185
The present disclosure relates to cloud computing systems and to techniques for rendering and provisioning cloud services in a cloud computing system.
In a cloud computing environment, numerous cloud service requests are serviced in relatively short periods of time. In such an environment, it is highly beneficial to automate placement, rendering, and provisioning of cloud services within and between data centers, so that cloud service requests can be accommodated dynamically with minimal (and preferably no) human intervention.
Examples of cloud services include: compute services, network services, and storage services. Examples of network services include Layer 2 (L2) virtual local area network (VLAN) or Layer 3 (L3) virtual routing and forwarding (VRF) connectivity between various physical and logical elements in a data center, Layer 4 (L4) to Layer 7 (L7) services including firewalls and load balancers, Quality of Service (QoS), access control lists (ACLs), and accounting.
Network management of cloud computing systems currently use orchestration tools which learn about all cloud elements within the data center, make all placement decisions for all cloud elements within the data center, and render and provision the cloud service request by communicating directly with each cloud element in the data center. Due to the wide variety of functionality and implementations within the data center, such orchestration tools need to be highly customizable and flexible so that they can utilize the wide array of management interfaces provided by various cloud elements. Considerable effort is required to make the orchestration tools properly interface with new implementations of cloud elements within the data center, leading to significant time delays before new features provided by the cloud elements can be managed by the orchestration tools. These orchestration tools have great difficulty scaling to large cloud environments.
Overview
Techniques are provided herein for distributed hierarchical rendering and provisioning of cloud services. At a device in a cloud computing system comprising a plurality of hierarchical levels, a cloud service request is received for rendering and provisioning of a virtual data center. A determination is made as to which aspects of the cloud service request are to be satisfied by devices in a first hierarchical level and which aspects of the cloud service request are to be satisfied by devices in a second hierarchical level and in subsequent hierarchical levels. Rendering and provisioning commands are provided to one or more devices in the first hierarchical level that are selected to satisfy aspects of the cloud service request in the first hierarchical level. A subset cloud service request for aspects of the cloud service request that are to be satisfied by devices in the second hierarchical level is sent to a device in the second hierarchical level.
Referring first to
The policy-based cloud service requests progress through a set of distributed hierarchical service rendering engines (SREs). The network level 10 connects multiple different data centers at the data center level 20, e.g., data center 20(1) labeled as DC 1 and data center 20(2) labeled as DC 2, and subsets of the data centers called “PODs” that are centered on aggregation switches within the data center. Again, the number of levels shown in
At each level of the hierarchy, there is at least one service rendering engine. In the network level 10, there are Provider Edge (PE) devices that perform routing and switching functions.
At the POD level 30, there are core/aggregation switches, firewalls, load balancers and web/application servers in each POD. The functions of the firewalls, load balancers, etc., may be hosted in a physical chassis or they may be hosted in a virtual machine executed on a computing element in the POD level 30. PODs 30(1)-30(n), labeled “POD 1.1”-“POD 1.n”, are connected to data center 20(1) and POD 40 is connected to data center 20(2). Thus, PODs 30(1)-30(n) may be viewed as different processing domains with respect to the data center 20(1), and the data center service rendering engine 200 in the edge switch 22(2) may select which one (or more) of a plurality of processing domains in the POD level to be used for aspects of a cloud service request that the data center service rendering engine 200 receives. Data center 20(2) cannot select one of the PODs 30(1)-30(n) because they are in different processing domains, but data center 20(2) can select POD 40. In each of PODs 30(1)-30(n), there are core/aggregation switches 32(1) and 32(2), one or more firewall (FW) devices 34, one or more load balancer (LB) devices 36, access switches 38(1) and 38(2) and servers 39(1)-39(m). The firewall and load balancers are not shown in POD 30(n) for simplicity. The servers 39(1)-39(m) each runs one or more virtual machine processes, i.e., virtual servers. There is a POD SRE 300 in core/aggregation switch 32(2) in each of PODs 30(1)-30(n). The POD SRE 300 may be hosted in other networking elements in the POD or in one or more virtual machines running on servers in the POD. In another form, the POD SRE functionality may be distributed across multiple devices in the POD. Similarly, in POD 40 there are core/aggregation switches 42(1) and 42(2), access switches 48(1) and 48(2) and servers 49(1)-49(m). There is a POD service rendering engine 300 in core/aggregation switch 42(2). POD 40 also includes one or more firewalls and load balancers but they are omitted in
Network management stations/orchestration tools external to the cloud computing system 5 transmit policy-based cloud service requests to one or more service rendering engines. That service rendering engine is responsible for all interactions with the external network management stations/orchestration tools regarding that cloud service request. An example of an external network management station (NMS) is shown at 50 in
The service rendering engine for any given hierarchical level is responsible for making placement decisions, rendering, and provisioning within its domain in its hierarchical level. This includes selection of one or more child domains in the next child (lower) hierarchical level for placement of cloud services or selection of one or more parent domains in the next parent (higher) hierarchical level for placement of cloud services, but does not include (knowledge of) details of placement, rendering, and provisioning within the child domain (or in the parent domain). Additional service rendering engines may be enabled in order to provide high availability and possibly in order to distribute processing load. In an active/standby approach, there is more than one service rendering engine for a particular domain and one of the service rendering engines is considered active and receives and responds to policy-based cloud service requests. The active service rendering engine checkpoints, i.e., registers, its state to any standby service rendering engines for the same domain. If the active service rendering engine goes down, then a standby service rendering engine will statefully transition to the active role and quickly resume processing of policy-based cloud service requests. In another approach, multiple service rendering engines for the same domain share a common distributed database containing all state information related to policy-based cloud service requests. This approach allows for policy-based cloud service requests to be load balanced across the multiple service rendering engines.
The cloud computing system 5 may be configured to employ recursive abstraction and normalization with respect to service rendering and placement mechanisms. At a bottom level of the hierarchy, there is collection of devices that comprises the POD. At the next level up, the POD is treated as if it were one device, like one large computer with numerous capabilities.
There are a plurality of layers of abstraction, a placement mechanism that takes place in each layer, policy elements, and interpretation of policy. Each layer provides abstraction and normalization with respect to service rendering and placement mechanisms. In other words, the determination as to where to place a cloud service request is based on normalized information representing capabilities of devices in a given hierarchical (e.g., first level) and another (e.g., second) hierarchical level.
Each layer can be non-homogeneous. For example, each layer can have its own policy rendering mechanism and its own placement mechanisms that apply to the level of abstraction provided by the immediate underlying layer. At each layer, the placement decisions are based on normalized information from a layer below. Once placed at the layer below, similar operations are performed at the layer below. Thus, this process is recursive from layer to layer, but within a given layer, the rendering and provisioning mechanisms may be different from each other.
The network view from the perspective of a data center service rendering engine is shown in
The terminology “aspects of the cloud service request” refers to features or functions of one or more cloud elements to support or implement the virtual data center functionality of the cloud service request. A virtual data center refers to functions of the cloud computing system invoked to simulate or virtualize a data center on behalf of a requesting entity.
The network view from the perspective of the POD service rendering engine 300 of POD 30(1) is shown in
Referring now to
The memory 62 and memory 92 shown in
The operations of processors 60 and 90 may be implemented by logic encoded in one or more tangible media (e.g., embedded logic such as an application specific integrated circuit, digital signal processor instructions, software that is executed by a processor, etc). The service rendering engine 100 may take any of a variety of forms, so as to be encoded in one or more tangible media for execution, such as fixed logic or programmable logic (e.g. software/computer instructions executed by a processor) and the processor 60 may be an application specific integrated circuit (ASIC) that comprises fixed digital logic, or a combination thereof. For example, the processor 60 may be embodied by digital logic gates in a fixed or programmable digital logic integrated circuit, which digital logic gates are configured to perform the operations of the service rendering engine 100. In one form, the service rendering engine 100 is embodied in a processor or computer-readable memory medium (memory 62) that is encoded with instructions for execution by a processor (e.g. a processor 60) that, when executed by the processor, are operable to cause the processor to perform the operations described herein in connection with service rendering engine 100. Similar configurations are also possible for the processor 90 and the policy agent 400 stored in memory 92.
Policy rendering and provisioning is accomplished by communicating applicable subsets of the policy-based cloud service request from service rendering engines to cloud elements. Within each cloud element, the policy agent is responsible for receiving the policy subsets and converting the policy-based information into that cloud element's internal configuration to perform the desired function in that cloud element. This takes the place of traditional network management configuration interfaces, relieving the service rendering engines from the task of policy transformation into various configuration interfaces and models.
Reference is now made to
The resource manager 80 makes the placement decision and triggers the policy agent in a cloud element or the resource manager 80 in a child service rendering engine, informing them that they have been selected to satisfy a policy-based cloud service request. The trigger message, shown at 82 in
The policy subset represents rendering and provisioning commands or information for a given policy agent in a given cloud element. When a service rendering engine sends rendering and provisioning commands to a policy agent, it sends rendering and provisioning subset information derived from a received cloud service request. The rendering and provisioning subset information comprises information that enables a policy agent to configure one or more functions of the device that receives the information. The rendering and provisioning subset information is different depending on the nature and general function of the device, i.e., firewall, load balancer, access switch, server, edge switch, core/aggregation switch, etc., and the function that the device is to perform in order to satisfy the overall virtual data center to be established according to the cloud service request.
Reference is now made to
At 210, a policy-based cloud service request is received by the data center service rendering engine 200. This request may originate from the NMS 50 shown in
At 215(1) and 215(2), the data center service rendering engine 200 communicates with the abstract device brokers 400 for all selected cloud elements, e.g., edge switch 22(1) and firewall device 24, within the data center domain in order to provide these cloud elements with the applicable policy rendering and provisioning subset information. At 217(1) and 217(2), the abstract device brokers 400 within the data center domain acknowledge receipt of the policy subsets to the data center service rendering engine 200.
At 220, the data center service rendering engine 200 progresses the policy-based cloud service request to the POD service rendering engine 300 for the selected POD, shown as POD 30(1) in this example. Again, the data center service rendering engine 200 may have multiple processing domains (e.g., PODs) to choose from and it selects one or more of these processing domains depending on the nature of the cloud service request. The cloud service request sent at 220 from the data center service rendering engine 200 to the POD service rendering engine 300 is a subset of the cloud service request received at 210 at the data center service rendering engine 200 because the cloud service request sent at 220 consists of those aspect of the cloud service request at 210 that the data center service rendering engine 200 determines to be satisfied by cloud elements in the POD hierarchical level 30 rather than in the data center hierarchical level 20. The subset cloud service request may be a filtered or marked version of the original cloud service request, as explained above.
Reference is now made to
At 310(1)-310(3), the POD service rendering engine communicates with the abstract device brokers for all selected cloud elements within the POD domain in order to provide each such cloud element with the applicable rendering and provisioning commands, e.g., rendering and provisioning policy subset information derived from the subset cloud service request at 220. At 312(1)-312(3), the abstract device brokers within the POD domain acknowledge (ACK) receipt of the policy subsets to the POD service rendering engine 300.
In response to receiving acknowledgments from the POD elements 36, 38(1) and 39(1), at 314 the POD service rendering engine 300 sends an acknowledgement to the data center service rendering engine 200.
Referring back to
Reference is now made to
At 410, a policy agent in a server in POD 30(1), e.g., server 39(m), detects that a virtual machine 35 comes up (goes active) on the server 39(m). The activation of the virtual server in the server 39(m) triggers a notification to the policy agent 400 in that server 39(m). At 412, the policy agent 400 in server 39(m) requests a policy from the POD service rendering engine 300. Again, the activation of the virtual machine 35 is in response to a policy-based cloud service request for establishing a virtual data center, but the activation of the virtual machine 35 occurs some time after receipt of the policy-based cloud service request. The cloud service request may still be received at a service rendering engine at any hierarchical level.
In response to receiving the policy request from the policy agent 400 in the server 39(m) where the virtual machine 35 went active, the POD service rendering engine 300 makes a placement decision, determining which cloud elements within the POD domain should be used to establish the requested virtual data center. In this case, the POD service rendering engine 300 selects a load balancer 46 and an access switch 38(2).
At 320(1) and 320(2), the POD service rendering engine communicates with the abstract device brokers 400 for all selected cloud elements within the POD domain, e.g., load balancer 36 and access switch 38(2), in order to provide each such cloud element with the applicable policy subset for rendering and provisioning. At 322(1) and 322(2), the abstract device brokers 400 within the POD domain acknowledge receipt of the policy subsets to the POD service rendering engine.
At 330, the POD service rendering engine 300 provides the applicable policy subset to the policy agent 400 that triggered POD rendering and provisioning.
At 340, the POD service rendering engine 300 triggers the data center service rendering engine 200. Referring now to
At 230(1) and 230(2), the data center service rendering engine 200 communicates with the abstract device brokers for all selected cloud elements within the data center domain in order to provide each such cloud element with the applicable policy subset for rendering and provisioning. At 232(1) and 232(2), the abstract device brokers 400 within the data center domain acknowledge receipt of the policy subsets to the data center service rendering engine 200.
At 240, the data center service rendering engine 200 sends an acknowledgement in response to the initial policy-based cloud service request.
In one implementation, operations 320(1)-320(3) and operations 230(1) and 230(2) involve a handshake rather than a unidirectional message containing the applicable policy subset.
Reference is now made to
At 510, at a device in a cloud computing system comprising a plurality of hierarchical levels, receiving a cloud service request for rendering and provisioning of a virtual data center. For example, at a device that executes a first service rendering engine in a device in a first hierarchical level, a cloud service request for rendering and provisioning of a virtual data center is received. In one example, the cloud service request may be received from an outside entity, e.g., NMS 50 (
At 520, a determination is made as to which aspects of the cloud service request are to be satisfied by devices in a first hierarchical level and which aspects of the cloud service request are to be satisfied by devices in a second hierarchical level and in subsequent hierarchical levels. For example, the first service rendering engine determines which aspects of the cloud service request are to be satisfied by cloud elements (devices) in the first hierarchical level and which aspects of the cloud service request are to be satisfied by cloud elements (devices) in a second hierarchical level and in subsequent (other) hierarchical levels (either up or down in the hierarchy). When the first service rendering engine determines at 520 which aspects of the cloud service request are to be satisfied by devices in the second hierarchical level it does so without knowledge of details of placement, rendering and provisioning of devices in the second hierarchical level. In the case of the “bottom up” flow described above in connection with
At 530, rendering and provisioning commands are provided to one or more devices in the first hierarchical level that are selected to satisfy aspects of the cloud service request in the first hierarchical level. For example, the first service rendering engine provides rendering and provisioning commands to one or more devices in the first hierarchical level that are selected by the first service rendering engine to satisfy aspects of the cloud service request in the first hierarchical level.
At 540, a subset cloud service request is sent to a device in the second hierarchical level for aspects of the cloud service request that are to be satisfied by devices in the second hierarchical level. For example, the first service rendering engine sends a subset cloud service request for aspects of the cloud service request that are to be satisfied by devices in the second hierarchical level to a second service rendering engine in a device in the second hierarchical level. As explained above, in one example the subset cloud service request is a forwarded version of the original cloud service request that has been filtered or otherwise portions of which are marked to indicate those aspects to be rendered by the second service rendering engine.
In sum, a mechanism is provided for distributed hierarchical rendering and provisioning of policy-based cloud service requests within and between data centers. Cloud service requests are satisfied by following the hierarchy in a top down or bottom up fashion, with a service rendering engine at each layer of the hierarchy responsible for placement, rendering, and provisioning at that level, then handing off to the next lower or higher level service rendering engine. Service rendering engines hand off subsets of the policy-based cloud service request to abstract device brokers resident on each cloud element for rendering and provisioning within each of those could elements. Each service rendering engine need only be capable of configuring and interfacing with cloud elements in its own processing domain of its hierarchical level and with the service rendering engines of its adjacent hierarchical levels.
The hierarchical and distributed control plane techniques describe herein provides vastly improved scalability. These techniques are useful for Cloud-Centric Networking (CCN) computing environments comprising numerous data centers with hundreds of thousands of servers per data center. Although one implementation involves three levels of hierarchy as described herein (POD level, data center level, and network level), these techniques may be employed for an arbitrary number of hierarchical levels, allowing customers to control the tradeoff between accuracy and scalability. These techniques render and provision policy on individual cloud elements by passing subsets of the policy-based cloud service request to abstract device brokers resident on each cloud element. This significantly simplifies service rendering engines and network management functionality by offloading policy transformation into various configuration interfaces and models, eliminating long implementation time lags between availability of new capabilities at network elements, and the ability of service rendering engines and network management stations to provision those new capabilities. Moreover, these techniques significantly reduce operational complexity by reducing the multiplicity of management interfaces and models required to render and provision cloud services.
The distributed hierarchical cloud services rendering and provisioning techniques described herein may be embodied in a method, an apparatus comprising a network interface unit and a processor configured to perform operations described herein, and/or computer-readable medium storing instructions that, when executed cause a processor to perform operations described herein.
The techniques described herein may be applied to failure management and disaster recovery to automatically relocate services from failed devices/PODs/Data Centers. When a device in one of the plurality of hierarchical levels fails or otherwise become inoperable, a determination is made as to whether there are one or more other available devices within that hierarchical level to satisfy the operations of the failed device and if so rendering and provisioning commands are provided to those one or more other devices. Otherwise, a cloud service request is sent to a device in another hierarchical level (e.g., in a next higher or next lower level) which determines whether there are one or more available devices in that hierarchical level to satisfy the operations of the failed device.
The techniques described herein may be in various forms, including a method, apparatus and computer-readable memory medium. In the latter case, instructions are stored in a tangible computer-readable memory medium (e.g., a memory device), which instructions are operable (or cause a processor, when executed) to: at a device in a cloud computing system comprising a plurality of hierarchical levels, receive a cloud service request for rendering and provisioning of a virtual data center; determine which aspects of the cloud service request are to be satisfied by devices in the first hierarchical level and which aspects of the cloud service request are to be satisfied by devices in a second hierarchical level and in subsequent hierarchical levels; provide rendering and provisioning commands to one or more devices in the first hierarchical level that are selected to satisfy aspects of the cloud service request in the first hierarchical level; and send to a device in the second hierarchical level a subset cloud service request for aspects of the cloud service request that are to be satisfied by devices in the second hierarchical level.
In apparatus form, an apparatus is provided that comprises a network interface unit configured to perform communications over a network; and a processor configured to be coupled to the network interface unit. The processor is configured to: receive a cloud service request for rendering and provisioning of a virtual data center in a cloud computing system comprising a plurality of hierarchical levels; determine which aspects of the cloud service request are to be satisfied by devices in a first hierarchical level and which aspects of the cloud service request are to be satisfied by devices in a second hierarchical level and in subsequent hierarchical levels; provide rendering and provisioning commands to one or more devices in the first hierarchical level that are selected to satisfy aspects of the cloud service request in the first hierarchical level; and send to a device in the second hierarchical level a subset cloud service request for aspects of the cloud service request that are to be satisfied by devices in the second hierarchical level.
In addition, localized rendering of a cloud element may be embodied in an apparatus comprising a network interface unit configured to perform communications over a network; a processor configured to be coupled to the network interface unit. The processor is configured to: receive a rendering and provisioning information associated with a cloud service request in a cloud computing system; and convert the rendering and provisioning information into internal configuration information to perform one or more functions associated with the cloud service request.
The above description is intended by way of example only.
