ROUTING FOR LARGE SERVER DEPLOYMENTS

Abstract

In one aspect, the present disclosure relates to a method comprising: receiving, at a client device, information from a node manager about a plurality of nodes in a computer cluster, the information comprising a network address associated each of the plurality of nodes and sending, by the client device, a request to a load balancer to access a first node from the plurality of nodes, the request comprising a first URL including an encoded representation of the network address associated with the first node. The load balancer is configured to determine the request should be routed to a first network address based on decoding the URL, the first network address associated with a first node from the plurality of nodes and forward the request to the first node in response to the determining.

Claims

1-20. (canceled)

21. A method, comprising: receiving, by a computing system, a request from a client device for access to a first node of a plurality of nodes of a computer cluster, the request comprising a first uniform resource locator (URL), the first URL comprising an encoded representation of a network address associated with the node; decoding, by the computing system, the first URL to determine a location to which the request should be routed; and routing, by the computing system, the request to the determined location based on the decoding and without having prior knowledge about state information associated with the plurality of nodes.

22. The method of claim 21, further comprising: receiving, by the computing system, a second request to access another node of the computer cluster from a second client device determining, by the computing system, that the second request does not include a URL of a desired node; and routing, by the computing system, the second request to a second node of the plurality of nodes without having prior knowledge about the state information associated with the plurality of nodes.

23. The method of claim 21, further comprising: generating, by the computing system, a plurality of URLs, each of the plurality of URLs comprising an encoded representation of a network address associated with a respective node from the plurality of nodes in the computer cluster.

24. The method of claim 21, further comprising: identifying, by the computing system, state information associated with the plurality of nodes; determining, by the computing system, that a load on the computer cluster exceeds a threshold amount; and based on the determining, allocating, by the computing system, new nodes to the computer cluster.

25. The method of claim 21, wherein a path portion of the first URL comprises the encoded representation.

26. The method of claim 21, wherein a subdomain portion of the first URL comprises the encoded representation.

27. The method of claim 21, wherein the first URL comprises a hash of an Internet Protocol (IP) address associated with the first node.

28. A non-transitory computer readable medium comprising one or more sequences of instructions, which, when executed by a processor, causes a computing system to perform operations comprising: receiving, by the computing system, a request from a client device for access to a first node of a plurality of nodes of a computer cluster, the request comprising a first uniform resource locator (URL), the first URL comprising an encoded representation of a network address associated with the node; decoding, by the computing system, the first URL to determine a location to which the request should be routed; and routing, by the computing system, the request to the determined location based on the decoding and without having prior knowledge about state information associated with the plurality of nodes.

29. The non-transitory computer readable medium of claim 28, further comprising: receiving, by the computing system, a second request to access another node of the computer cluster from a second client device determining, by the computing system, that the second request does not include a URL of a desired node; and routing, by the computing system, the second request to a second node of the plurality of nodes without having prior knowledge about the state information associated with the plurality of nodes.

30. The non-transitory computer readable medium of claim 28, further comprising: generating, by the computing system, a plurality of URLs, each of the plurality of URLs comprising an encoded representation of a network address associated with a respective node from the plurality of nodes in the computer cluster.

31. The non-transitory computer readable medium of claim 28, further comprising: identifying, by the computing system, state information associated with the plurality of nodes; determining, by the computing system, that a load on the computer cluster exceeds a threshold amount; and based on the determining, allocating, by the computing system, new nodes to the computer cluster.

32. The non-transitory computer readable medium of claim 28, wherein a path portion of the first URL comprises the encoded representation.

33. The non-transitory computer readable medium of claim 28, wherein a subdomain portion of the first URL comprises the encoded representation.

34. The non-transitory computer readable medium of claim 28, wherein the first URL comprises a hash of an Internet Protocol (IP) address associated with the first node.

35. A system comprising: a processor; and a memory comprising one or more sequences of instructions, which, when executed by the processor, causes the system to perform operations comprising: receiving a request from a client device for access to a first node of a plurality of nodes of a computer cluster, the request comprising a first uniform resource locator (URL), the first URL comprising an encoded representation of a network address associated with the node; decoding the first URL to determine a location to which the request should be routed; and routing the request to the determined location based on the decoding and without having prior knowledge about state information associated with the plurality of nodes.

36. The system of claim 35, wherein the operations further comprise: receiving a second request to access another node of the computer cluster from a second client device determining that the second request does not include a URL of a desired node; and routing the second request to a second node of the plurality of nodes without having prior knowledge about the state information associated with the plurality of nodes.

37. The system of claim 35, wherein the operations further comprise: generating a plurality of URLs, each of the plurality of URLs comprising an encoded representation of a network address associated with a respective node from the plurality of nodes in the computer cluster.

38. The system of claim 35, wherein the operations further comprise: identifying state information associated with the plurality of nodes; determining that a load on the computer cluster exceeds a threshold amount; and based on the determining, allocating new nodes to the computer cluster.

39. The system of claim 35, wherein a path portion of the first URL comprises the encoded representation.

40. The system of claim 35, wherein a subdomain portion of the first URL comprises the encoded representation.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Various objectives, features, and advantages of the disclosed subject matter can be more fully appreciated with reference to the following detailed description of the disclosed subject matter when considered in connection with the following drawings.

[0012] FIG. 1 is a diagram of a cluster computing system, according to some embodiments of the present disclosure.

[0013] FIG. 2 is a flow diagram showing processing that may occur within the system of FIG. 1, according to some embodiments of the present disclosure.

[0014] FIG. 3 is another flow diagram showing processing that may occur within the system of FIG. 1, according to some embodiments of the present disclosure.

[0015] The drawings are not necessarily to scale, or inclusive of all elements of a system, emphasis instead generally being placed upon illustrating the concepts, structures, and techniques sought to be protected herein.

DETAILED DESCRIPTION

[0016] According to various embodiments of the present disclosure, a load balancer may route network traffic to particular backend servers using network address information encoded within incoming requests. An incoming request may include the network address of a backend server encoded within a Uniform Resource Locator (URL), for example within the subdomain or path portion of the URL. The load balancer may decode the network address from the subdomain/path portion of the URL, and then forward the request to the server at that address. Using this approach, the load balancer may be able to route traffic to a large number of backend servers without having to maintain state information about those servers.

[0017] FIG. 1 illustrates a cluster computing system 100, according to an embodiment of the present disclosure. The system 100 includes a cluster node manager 102, a database 104, a cluster load balancer 106, and a computer cluster 108. Users can interact with the system 100 using one or more client devices 112, such as a desktop, laptop, tablet, or mobile computing device. In some embodiments, users can interact with the system 100 using a web browser executing on the client device 112. The various components of the system 100 and client devices 112 may be connected as illustrated in FIG. 1 or in any other suitable manner. The illustrated components may be connected by one or more wireless or wireline computer networks.

[0018] In the embodiment of FIG. 1, the computer cluster (or “cluster”) 108 includes one or more nodes 110a, 110b, . . . , 110n (110 generally). Each node 110 may correspond to a physical or virtual computer. In some embodiments, all of the nodes 110 have substantially the same hardware and/or software configuration. For example, each node 110 may comprise a Hypertext Transfer Protocol (HTTP) server and a copy of application software responsive to HTTP requests received by the HTTP server.

[0019] In particular embodiments, the system 100 may be used for machine intelligence (MI) tasks. In such embodiments, each node 110 may comprise an HTTP server, visualization tools, and an MI processor (none of which are shown in FIG. 1). The MI processor may comprise TENSORFLOW™, an open-source software library for MI, and the visualization tools may comprise TENSORBOARD™, an open-source tool to visualize neural network models (or other MI models) defined as “graphs” within a TENSORFLOW™ application. In other embodiments, the visualization tools and/or MI processor may utilize other open-source or proprietary MI software libraries known to one of ordinary skill in the art. In certain embodiments, the visualization tools may include a web-based graphical user interface, and a client device 112 can connect to the HTTP server of a particular node 110 in order to access the node's visualization tools.

[0020] Referring again to the embodiment of FIG. 1, the cluster node manager (or “node manager”) 102 may be configured to manage nodes within the cluster 108. For example, the node manager 102 may instantiate and de-instantiate nodes 110 based on system load or other criteria. In particular embodiments, the node manager 102 may allocate nodes 110 based on load information obtained from individual nodes 110 and use this load information to dynamically adjust the number of nodes 110. In particular embodiments, the load information may include one or more metrics associated with the training of a MI model, such as a deep neural network model.

[0021] In some embodiments, the cluster computing system 100 may be hosted within a cloud computing environment. In such embodiments, the node manager 102 may instantiate and de-instantiate nodes 110 using an Application Programming Interface (API) provided by the cloud computing environment.

[0022] Referring to the embodiment of FIG. 1, the node manager 102 may maintain information about the state of the cluster 108. In some embodiments, the node manager 102 maintains a lookup table including a network address for each allocated node 110. In certain embodiments, the network addresses are Internet Protocol (IP) addresses. In some embodiments, the IP addresses are internal to the system 100, meaning that client devices 112 cannot directly connect to the nodes 110 using those IP addresses. Instead, the client devices 112 may connect to the nodes indirectly through the cluster load balancer 106. In certain embodiments, the node manager 102 stores state information within the database 104.

[0023] Referring again to the embodiment of FIG. 1, the node manager 102 may include an HTTP server or other type of server that can process requests from the client device 112. In some embodiments, the client device 112 can request, from the node manager 102, information about nodes 110 currently allocated within the cluster 108. In certain embodiments, the node manager 102 may generate a Uniform Resource Locator (URLs) for each node 110 in the cluster and return a list of node URLs to the client device 112. In some embodiments, the node manager 102 may return the list of URLs as a web page which is rendered in a web browser of the client device 112. In some embodiments, the URL generated for a particular node 110 may include an encoded representation of that node's network address within the system 110. For example, assuming a first node 110a has IP address 10.0.1.1, the URL generated for node 110a may include an encoded representation of this address within the subdomain portion of the URL as shown here:

http://inst-0a000101.example-url.com

[0024] As another example, the node's address may be encoded with the path portion of the URL, as shown here:

http://example-url.com/inst-0a000101/

[0025] In particular embodiments, the node's network address may be encoded using hexadecimal encoding. In some embodiments, the network address may be encrypted within the URL to avoid exposing network address information outside of the system 100 and/or to prevent URL tampering.

[0026] In some embodiments, the cluster load balancer 106 manages network traffic between the client devices 112 and the cluster nodes 110. In particular, the cluster load balancer 106 may receive a request from a client device 112, determine which cluster node 110 should handle the request, and then forward the client request to that node 110. In turn, the node 110 may process the client request, return a response to the cluster load balancer 106, and the cluster load balancer 106 may forward the response to the client device.

[0027] In some embodiments, cluster load balancer 106 may be configured to route client requests to particular nodes 110 specified by the request. For example, a client request may include a URL having an encoded representation of a node's network address. The URL may correspond to a URL generated by the cluster node manager 102 and provided to the client device 112, as described above. In such embodiments, the cluster load balancer 106 may be configured to decode the URL to determine the node's network address (e.g., IP address) and then forward the client request to the node using the decoded address. For example, if the client request includes the following URL:

http://inst-0a000101.example-url.com

[0028] the cluster load balancer 106 may decode this URL to determine node address 10.0.1.1 and then forward the client request to that address.

[0029] Whereas traditional load balancers may need to keep track of each backend server, the cluster load balancer 106 can route client requests to particular nodes 110 without having to maintain any information about the nodes. Moreover, unlike existing load balancers, the techniques described herein do not require iterating through rules lists or using lookup tables and, thus, scale to an arbitrary number of nodes 110 without affecting the processing or storage requirements within the load balancer 106.

[0030] In one embodiment, the cluster load balancer 106 utilizes a NGINX™, an open source web server and load balancer, along with OpenResty® to route network traffic to the cluster nodes 110.

[0031] Referring to FIG. 2, a method 200 can be used for stateless routing in a cluster computing system, according to an embodiment of the present disclosure. At block 202, a URL may be generated for each node in a computer cluster (e.g., each node 110 in cluster 108 of FIG. 1). Each URL may include an encoded representation of the network address for the corresponding node. In some embodiments, the URLs have a format similar to one of the URL formats described above in conjunction with FIG. 1. At block 204, the URLs may be sent to a client device, such as client device 112 of FIG. 1. In certain embodiments, blocks 202 and 204 may be performed by a cluster node manager (e.g., cluster node manager 102 of FIG. 1). In some embodiments, the URLs are sent to the client device as a web page.

[0032] At block 206, a request is received from a client device, the request including a first URL. The first URL includes an encoded representation of a network address of a first node of the cluster (e.g., node 110a in FIG. 1). The first URL may be one of the URLs generated and sent to the client at blocks 202 and 204. At block 208, the first URL is decoded to obtain the network address of the first node. At block 210, the request is forwarded to the first node using the decoded network address. At block 212, a response is received from the first node, e.g., after the first node has completed processing of the request. At block 214, the response is forwarded to the client device. In some embodiments, blocks 206-214 may be performed by a cluster load balancer (e.g., cluster load balancer 106 of FIG. 1).

[0033] Referring to FIG. 3, a method 300 may be performed by a client device (e.g., client device 112 of FIG. 1) to send a request to particular node within a computer cluster (e.g., node 110a in cluster 108 of FIG. 1), according to an embodiment of the present disclosure. At block 302, a request is sent to a cluster node manager (e.g., cluster node manager 102 of FIG. 1). At block 304, the cluster node manager responds with a list of URLs, each URL is associated with a node of the cluster and includes an encoded representation of that node's network address. In particular embodiments, the URLs have a format similar to one of the URL formats described above in conjunction with FIG. 1. In some embodiments, the cluster node manager responses with a web page including the list of URLs.

[0034] At block 306, a request is sent to a cluster load balancer (e.g., cluster load balancer 106 of FIG. 1), the request including a first URL associated with the first node. In some embodiments, the cluster load balancer is configured to decode the first URL to obtain the first node's network address and then forward the request to the first node using the decoded network address. At block 308, a response is received from the cluster load balancer, the response being generated by the first node. In some embodiments, the cluster load balancer is configured to forward the response from the first node to the client device.

[0035] Although embodiments of the present disclosure have been described for use with cluster computing systems, the concepts sought to be protected herein can be utilized in any multi-server computing system.

[0036] The subject matter described herein can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them. The subject matter described herein can be implemented as one or more computer program products, such as one or more computer programs tangibly embodied in an information carrier (e.g., in a machine readable storage device), or embodied in a propagated signal, for execution by, or to control the operation of, data processing apparatus (e.g., a programmable processor, a computer, or multiple computers). A computer program (also known as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.

[0037] The processes and logic flows described in this specification, including the method steps of the subject matter described herein, can be performed by one or more programmable processors executing one or more computer programs to perform functions of the subject matter described herein by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus of the subject matter described herein can be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).

[0038] Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processor of any kind of digital computer. Generally, a processor will receive instructions and data from a read only memory or a random access memory or both. The essential elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto optical disks, or optical disks. Information carriers suitable for embodying computer program instructions and data include all forms of nonvolatile memory, including by way of example semiconductor memory devices, such as EPROM, EEPROM, flash memory device, or magnetic disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.

[0039] It is to be understood that the disclosed subject matter is not limited in its application to the details of construction and to the arrangements of the components set forth in the following description or illustrated in the drawings. The disclosed subject matter is capable of other embodiments and of being practiced and carried out in various ways. Also, it is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting. As such, those skilled in the art will appreciate that the conception, upon which this disclosure is based, may readily be utilized as a basis for the designing of other structures, methods, and systems for carrying out the several purposes of the disclosed subject matter. It is important, therefore, that the claims be regarded as including such equivalent constructions insofar as they do not depart from the spirit and scope of the disclosed subject matter.

[0040] Although the disclosed subject matter has been described and illustrated in the foregoing exemplary embodiments, it is understood that the present disclosure has been made only by way of example, and that numerous changes in the details of implementation of the disclosed subject matter may be made without departing from the spirit and scope of the disclosed subject matter.