Patent Application
20200120065
Networked computing devices communicate with one another over Transmission Control Protocol/Internet Protocol (TCP/IP). A networked device must be associated with a network address (e.g., an IP address) in order to communicate with other networked devices over a TCP/IP network. Dynamic Host Configuration Protocol (DHCP) is a network management protocol for dynamically assigning network addresses and other configuration parameters to networked devices. Generally, a DHCP server receives a DHCP request from a networked device and assigns an IP address and configuration parameters to the requesting device.
Distributed computing networks may include several hardware servers, each of which requires a dedicated IP address. Hardware servers may host one or more virtual machines, each of which also requires a dedicated IP address to independently communicate with other network components. Conventionally, IP addresses are assigned to hardware servers and their virtual machines by a centralized DHCP server. The centralized DHCP server is accessible over a physical network to which the hardware servers are connected.
A centralized DHCP server may present reliability, scalability and debuggability challenges. If a centralized DHCP server malfunctions or is otherwise unavailable, all hardware servers and virtual machines served by this DHCP server are impacted. With increased server cluster sizes and virtual machine densities, a single DHCP server may quickly become overwhelmed with DHCP traffic and its corresponding processing requirements. Moreover, DHCP discover and DHCP request packets require broadcast functionality in the physical network, which complicates physical network configuration (e.g., often requiring a DHCP relay) and end-to-end debugging of the DHCP channel.
DHCP implementations which address one or more of the foregoing issues are desired.
The following description is provided to enable any person in the art to make and use the described embodiments. Various modifications, however, will remain readily-apparent to those in the art.
Some embodiments provide a technical solution to the above-described technical problems posed by a centralized DHCP server implementation in a distributed environment. Technical features according to some embodiments include caching of a predetermined networking state on a network node, intercepting, before layer-2 routing, a DHCP packet transmitted by a virtual machine hosted on the node, generating a DHCP response to the DHCP packet based on the cached networking state, and providing the DHCP response to the virtual machine.
By implementing DHCP support for a virtual machine on the node hosting the virtual machine, the overall DHCP load may be distributed and scalability may be improved. Some embodiments may improve the reliability of DHCP support and the ability to troubleshoot DHCP faults by shortening the DHCP communication channel and removing dependencies on the physical network and on broadcast support.
Virtual machine 112 issues DHCP discover and request packets as is known in order to obtain an IP address via DHCP. These packets are intercepted by virtual machine switch 114 and node agent 116 generates appropriate DHCP offer/ack packets based on the DHCP response-related information received from controller 120. The generated DHCP offer/ack packets are reinjected back to virtual machine 112 through virtual machine switch 114.
Server node 110 may comprise a traditional standalone computing device, a server blade mounted in a rack of several server blades, or any other set of hardware providing networking, compute, and storage resources. Node 110 executes a host operating system which supports the execution of one or more virtual machines. The one or more virtual machines may access and use allocated portions of the networking, compute, and storage resources of server node 110. Accordingly, virtual machine 112 is an abstraction of the networking, compute, and storage resources of server node 110. Virtual machine 112 may execute its own operating system, on which one or more applications may execute.
Virtual machine switch 114 provides a connection between virtual machine 112 (and any other virtual machines executing on server node 110) and external networks on different subnets. Virtual machine switch 114 may comprise a software-based layer-2 Ethernet network switch component of the host operating system of server node 110. Virtual machine switch 114 may support callout drivers for providing extensible functionality as will be described below. Node agent 116 may also be a component of the host operating system of server node 110.
Controller 120 may manage all virtual machines and nodes in a node cluster of one or more nodes. Accordingly, controller 120 may provide DHCP response-related information for a plurality of virtual machines being hosted on each of a plurality of nodes such as node 110. Controller 120 may communicate with network devices, services and components to discover network devices, detect service configurations and transmit configuration information to the network. Controller 120 may provide an interface to a management application to allow a user to configure, monitor, troubleshoot and deploy devices within the network.
Initially, at S210, DHCP information is transmitted from a node controller to a node. With reference to
The DHCP response-related information specifies a networking configuration associated with virtual machine 112. The information may include information based on which a DHCP response packet may be generated, for example, the Media Access Control (MAC) address of virtual machine 112, the IP address of virtual machine 112, a DHCP server IP address, the IP address of the default gateway, a subnet mask, Domain Name Service server settings, a domain name, and routes. According to some embodiments, the transmitted DHCP response-related information includes a binary-formatted DHCP response packet. Any other information may be included in the DHCP response-related information, such as but not limited to an interface type, virtual network information (e.g., an IP address of virtual machine 112 and a subnet mask in a virtual address space), and network Access Control Lists.
Next, at S220, a virtual machine switch located on the node receives a DHCP request packet from a virtual machine executing on the node. For example, virtual machine 112 of node 110 includes a virtual Network Interface Controller (NIC) which is connected to virtual machine switch 114 through a switch port. Virtual machine 112 may issue a DHCP discover packet in order to initiate a DHCP session, and virtual machine switch 114 receives the packet at S220.
Prior to S220, and according to some embodiments, node agent 116 registers an interception handler with virtual machine switch 114 for DHCP discover and request packets. Virtual machine switch 114 calls this interception handler upon receiving a DHCP discover or request packet. In some embodiments, virtual machine switch 114 provides a filter engine that supports the registration of callout kernel-mode drivers such as the interception handler. Accordingly, the filter engine of virtual machine switch 114 calls the interception handler in response to receiving the DHCP discover packet at S220, resulting in transmission of the packet to the interception handler of node agent 116 at S230.
Node agent 116 generates an appropriate DHCP response to the DHCP discover packet at S240. The response is based on the DHCP response-related information received from controller 120 at S210. In the present example, the response to the DHCP discover packet is a DHCP offer packet and includes a MAC address, the offered IP address, the subnet mask, the lease duration, and the IP address of the DHCP server. As described above, all of this information has been previously received from controller 120 at S210 and is available on node 110. Node agent 116 therefore has all of the information required to generate the DHCP response.
The generated DHCP offer packet is transmitted from node agent 116 to virtual machine switch 114 at S250, by virtue of the callback mechanism. This reinjected packet is then transmitted back to virtual machine 112 using normal TCP/IP routing protocols at S260. S230 through S260 may repeat for a subsequent DHCP request packet sent from virtual machine 112 and corresponding DHCP acknowledge packet generated by node agent 116, in order to complete the DHCP session. Accordingly, a complete DHCP session may be supported without requiring layer-2 forwarding of any DHCP packets out of node 110.
In some embodiments, controller 120 transmits DHCP information to node 110 in response to each change to the network configuration.
Node 310 of system 300 hosts virtual machine 311 associated with a virtual network and virtual machine 312 which is not associated with a virtual network. More specifically, virtual machine 311 is associated with a virtual IP address space that is independent of the IP address space of the physical network to which node 310 is connected. The virtual IP address space is defined within the configuration of virtual machine 311.
DHCP information is transmitted from a node controller to a node at S405 as described above with respect to S210. Referring to
A virtual machine switch located on the node receives a network packet from a virtual machine hosted on the node at S410. For example, virtual machine switch 315 may receive a network packet from virtual NIC 313 associated with virtual machine 311. At S415, virtual machine switch 315 determines if the received network packet is a DHCP discover or request packet. The determination may be based on a callback registered with filter engine 316 which is to be invoked upon receipt of such packets. If the received network packet is not a DHCP discover or request packet, virtual machine switch 315 forwards the packet at S420 to the physical network via a physical NIC of node 310.
Flow proceeds to S425 if it is determined at S415 that the network packet is a DHCP packet. At S425, it is determined whether the packet is associated with a virtual network. For example, filter engine 316 may determine that the packet was received from virtual NIC 313, which is associated with a virtual network. In this case, filter engine 316 calls virtual network interception handler 318 of node agent 317 at S430.
In the illustrated embodiment, node agent 317 includes virtual network interception handler 318 for handling intercepted DHCP packets which were issued from a virtual machine associated with a virtual network, and interception handler 319 for handling intercepted DHCP packets issued from a virtual machine which is not associated with a virtual network. Virtual network interception handler 318 translates the network packet from its virtual network address space to a physical network address space at S435. A mapping or other information for performing this translation may be included in the information received from controller 320 at S405.
DHCP responder 330 generates an appropriate DHCP response to the address-translated DHCP packet at S440. The response is generated based on the DHCP response-related information received from controller 320 at S405 and reflects the IP address space of the physical network. Virtual network interception handler 318 then translates the generated DHCP response from the physical network address space to the virtual network address space at S445.
The DHCP response packet is transmitted from node agent 317 to virtual machine switch 315 at S460 based on the original callback. This reinjected packet is then transmitted back to virtual machine 311 in the virtual address space using normal TCP/IP routing protocols at S465.
If it is determined at S425 that the packet is not associated with a virtual network (e.g., the packet was received from virtual NIC 314 of virtual machine 312), filter engine 316 calls interception handler 319 of node agent 317 at S450. Handler 319 does not perform an address translation, and DHCP responder 330 generates an appropriate DHCP response to the non-translated DHCP packet at S450 based on the DHCP response-related information received from controller 320. Flow then continues through 460 and 465 as described above to return the DHCP response to virtual machine 312.
More particularly, a virtual machine located on the node transmits a DHCP request packet including a DDNS host name. As described above, node agent 510 intercepts the packet and forwards it to DHCP responder 512. Since the packet includes DDNS registration information, DHCP responder 512 calls DDNS registration cache 514 to cache the DDNS registration information. DDNS registration cache 514 asynchronously executes a background thread to persist the DDNS registration information to disk 530 without blocking DHCP responder 512 from reinjecting the DHCP response as described above. DDNS registration cache 514 will load the persisted DDNS registration information from disk 530 whenever node agent 510 restarts.
Node agent client 522 of controller 520 periodically calls a synchronization routine (e.g., via a GetStateRequest API), and DDNS registration cache 514 of node agent 510 responds with the DDNS registration information and DHCP status. Node agent client 522 stores the DDNS registration information into DDNS registration monitor 524. DDNS registration monitor 524 wakes up based on the event, and creates a DDNS programming task in network allocation data model 526. DDNS registration monitor 524 relieves node agent client 522 from having to write a transaction. According to alternative embodiments, node agent 510 publishes the DDNS registration information to PubSub and controller 520 (or the DNS server) subscribes thereto.
Nodes 610, 612 and 614 may be traditional standalone computing devices and/or individual blades in a rack of many blade servers. Each of nodes 610, 612 and 614 includes a NIC that manages communication with an external physical network. One or more node processors on each node executes a host operating system that supports multiple virtual machines. Although the virtual machine switch of each node is illustrated separately, the host operating system may include a virtual machine switch and a node agent to operate as described herein. Controller 620 may provide node agents on each of nodes 610, 612 and 614 with DHCP response-related information as described above.
Each virtual machine may execute its own operating system which may be the same or different than the operating system executed on any other virtual machine. Each virtual machine may run one or more applications on its operating system. Each of nodes 610, 612 and 614 also includes storage (e.g., hard disk drives) and memory (e.g., RAM) that can be accessed and used by the node processors and virtual machines. In other architectures, some nodes may be dedicated to storage and while other nodes are dedicated to executing virtual machines.
According to some embodiments, nodes 610, 612 and 614 belong to a single cluster. Grouping nodes into a cluster may allow for workload sharing via a load balancer, enable a single point of operation/management via a controller such as controller 620, and/or provide a path for scaling to meet increased demand. Routers and/or network switches may support data traffic between nodes 610, 612 and 614, and between infrastructure 600 and external resources and users.
According to some embodiments, customer computing system 720 requests resources from resource service provider 730 located in public cloud 740. The request may specify datacenter resources and architectures, and may include configuration information of any level of specificity. Based on the request, resource service provider 730 transmits service configuration information to datacenter 710.
Datacenter 710 configures its resources based on the configuration information to provide the requested resources. For example, datacenter manager 711 may transmit information to controllers 712, 713 and 714 to instantiate a number of virtual machines having various characteristics (including but not limited to virtual network address spaces) within clusters 715, 716 and/or 717.
The elements of datacenter 710 may be distributed and/or mirrored to multiple locations to provide redundancy and disaster recovery. For example, the datacenter may designate one virtual machine on one node as the primary location for an application and a second virtual machine on the same or different node as a standby or back-up in case the primary virtual machine or node fails. Datacenter manager 711 may automatically shift incoming user requests from the primary virtual machine to the back-up virtual machine without requiring intervention.
Each functional component described herein may be implemented in computer hardware (integrated and/or discrete circuit components), in program code and/or in one or more computing systems executing such program code as is known in the art. Such a computing system may include one or more processing units which execute processor-executable program code stored in a memory system.
The above-described diagrams represent logical architectures for describing processes according to some embodiments, and actual implementations may include more or different components arranged in other manners. Other topologies may be used in conjunction with other embodiments. Moreover, each component or device described herein may be implemented by any number of devices in communication via any number of other public and/or private networks. Two or more of such computing devices may be located remote from one another and may communicate with one another via any known manner of network(s) and/or a dedicated connection. Each component or device may comprise any number of hardware and/or software elements suitable to provide the functions described herein as well as any other functions.
Embodiments described herein are solely for the purpose of illustration. Those in the art will recognize other embodiments may be practiced with modifications and alterations to that described above.
This application is a continuation of U.S. patent application Ser. No. 15/966,612, filed on Apr. 30, 2018, the entire contents of which are incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15966612 | Apr 2018 | US |
| Child | 16707252 | US |
