US20170111478A1 - Load balancing utilizing adaptive thresholding - Google Patents

Abstract

Methods and systems for processing data requests in a client-server computing environment, such as the Web, are disclosed. A client device initially transmits a data request to a randomly selected first server device. The first server device may reject the data request if its request queue threshold is exceeded. The client device retransmits the data request, including the request queue threshold, to a randomly selected second server device. The second server device may reject the data request if the request queue threshold of the first server device is smaller than a request queue threshold of the second server device. The client device transmits the data request back to the first server device, including the request queue threshold of the second server device. The first server device processes the data request and adjusts its request queue threshold based on the request queue thresholds of the first and second server devices.

Description

INCORPORATION BY REFERENCE TO ANY PRIORITY APPLICATIONS
• Any and all applications for which a foreign or domestic priority claim is made are identified in the Application Data Sheet filed with the present application and are incorporated by reference under 37 CFR 1.57 and made a part of this specification.
BACKGROUND
• The ubiquity of computers in business, government, and private homes has resulted in availability of massive amounts of information from network-connected sources, such as data stores accessible through communication networks, such as the Internet. In recent years, computer communication and search tools have become widely available to facilitate the location and availability of information to users. Most computer communication and search tools implement a client-server architecture where a user client computer communicates with a remote server computer over a communication network. In order to achieve better system performance and throughput in the client-server architecture, large communication network bandwidths are needed as the number of client computers communicating with server computers increases.
• One approach to increasing communication bandwidths relates to employing multiple networked server computers offering the same services. These server computers may be arranged in server farms, in which a single server from the server farm receives and processes a particular request from a client computer. Typically, server farms implement some type of load balancing algorithm to distribute requests from client computers among the multiple servers. Generally described, in a typical client-server computing environment, client devices generally issue requests to server devices for some kind of service and/or processing and the server devices process those requests and return suitable results to the client devices. In an environment where multiple clients send requests to multiple servers, workload distribution among the servers significantly affects the quality of service that the client devices receive from the servers. In many modern client-server environments, client devices number in the hundreds of thousands or millions, while the servers number in the hundreds or thousands. In such environments server load balancing becomes particularly important to system performance.
• One approach to increase the effectiveness of load balancing and the resulting system performance and throughput, is to efficiently find the servers which have lower load levels than other servers, and assign new client requests to these servers. Finding and distributing workload to overloaded and under-utilized servers may be done in a central or a distributed manner. Central control of load balancing requires a dedicated controller, such as a master server, to keep track of all servers and their respective loads at all times, incurring certain administrative costs associated with keeping lists of servers and connections up-to-date. Additionally, such a master server constitutes a single point of failure in the system, requiring multiple mirrored master servers for more reliable operation. Still further, the reliability and scalability of the number of servers in the server farm can be dependent on the ability and efficiency of the dedicated controller to handle the increased number of servers.
• Other approaches to finding and distributing workloads in a multi-server environment exist that relate to distributed, software-based approaches in which the client computers implement some type of load balancing software components. In one such approach, the client computer randomly selects a server. For example, a pseudo-random number generator may be utilized to select one of N servers. However, random selection of servers does not take the actual server loads into consideration and, thus, cannot avoid occasionally loading a particular server. Random server selection algorithms improve the average performance for request handling. This means such algorithms improve request handling for about 50% of the requests, but not for the majority of the requests. In another approach, the client computing device can implement a weighted probability selection algorithm in which the selection of a server is determined, at least in part, on the reported load/resources of each server. This approach must contend with the problem of information distribution among client devices. That is, server load information must be updated periodically at each client device to make optimal server selection based on server loads. The server load may be indicated by a length of a request queue at each server, a request processing latency, or other similar indicators. In yet another approach, a round-robin algorithm for server assignment may be used where each request is sent to a next server according to a number indicated by a counter maintained at the client device. Although simple to implement, this approach does not distribute the load optimally among servers because the round-robin cycle in different clients could coincide, causing multiple clients to call the same server at the same time. In yet another approach, servers may be assigned to individual clients on a priority basis. For example, a client may be assigned a number of servers according to a prioritized list of servers where the client sends a request to the server with the highest priority first, and next re-sends the request to the server with the next highest priority, if needed, and so on. As noted above, each of these approaches for server load distribution suffer from a particular problem that make server selection and load distribution sub-optimal, causing low levels of performance.
SUMMARY
• This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This summary is not intended to identify key features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.
• In one aspect of the invention a computer-implemented method for processing data requests is provided that includes obtaining a data request for a document and/or service available from a server, such as a Web server. A first request queue threshold, associated with a first server, and a second request queue threshold, associated with a second server, are compared to determine whether to process the data request. Based on the comparison of the two thresholds, the first request queue threshold is increased and the data request is processed.
• According to another aspect of the invention, a system for load balancing is provided including a first server, coupled with a network, for obtaining and processing a data request. A first data store coupled with the first server is provided for storing information associated with the data request. The system includes a second server, coupled with the network, for obtaining and processing the data request. A second data store coupled with the second server is also provided for storing information associated with the data request. The system also includes a first request queue associated with the first server, having a first threshold and a second request queue associated with the second server, having a second threshold. The first server increases the first threshold and processes the data request based on a comparison of the first threshold and the second threshold.
• According to yet another aspect of the invention, a system for load balancing is provided including a client component operating within a client computing device for transmitting a data request to a server. The system includes a first server, coupled with a network, for obtaining and processing a data request. A first data store coupled with the first server is provided for storing information associated with the data request. The system includes a second server, coupled with the network, for obtaining and processing the data request. A second data store coupled with the second server is also provided for storing information associated with the data request. The system also includes a first request queue associated with the first server, having a first threshold and a second request queue associated with the second server, having a second threshold. The first server increases the first threshold and processes the data request based on a comparison of the first threshold and the second threshold.
• According to yet another aspect of the invention, a computer-implemented method for processing data requests including transmitting a first data request to a first server computing device is provided. The first data request is rejected if a first request queue threshold associated with the first server computing device is exceeded. A second data request is formed by adding the first request queue threshold to the first data request, and transmitting the second data request to a second server computing device. The second data request is also rejected if a second request queue threshold associated with the second server computing device is greater than the first request queue threshold included in the second data request. A third data request is then formed, by adding the second request queue threshold to the first data request, and transmitted to the first server computing device.
• Other aspects and advantages of the present invention will become apparent from the detailed description that follows including the use of adaptive thresholds for balancing server loads.
DESCRIPTION OF THE DRAWINGS
• The foregoing aspects and many of the attendant advantages of this invention will become more readily appreciated as the same become better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein:
• FIG. 1 is a block diagram depicting an illustrative client-server operating environment suitable for distributed load balancing, including a number of client devices and a number of server devices having a client request queue;
• FIG. 2A is a block diagram of the client-server operating environment ofFIG. 1 illustrating the initiation of a request from a client device to a first server computing device;
• FIG. 2B is a block diagram of the client-server operating environment ofFIG. 1 illustrating a first alternative of a first server computing device accepting the request from the client computing device for processing;
• FIG. 2C is a block diagram of the client-server operating environment ofFIG. 1 illustrating a second alternative first server computing device rejecting the request from the client computing device and the client computing device sending the request to a second server for processing;
• FIG. 2D is a block diagram of the client-server operating environment ofFIG. 1 illustrating a first alternative second server computing device rejecting the request from the client computing device and the client computing device resending the request to the first server computing device;
• FIG. 2E is a block diagram of the client-server operating environment ofFIG. 1 illustrating a second alternative second server computing device accepting the request from the client computing device and adjusting its queue threshold; and
• FIG. 3 is a flow diagram depicting an illustrative method for accepting requests and adjusting server computing device queue thresholds.
DETAILED DESCRIPTION
• Generally described, the invention relates to load balancing in a client-server computing environment. Specifically, the invention relates to the balancing of server load using distributed routing of client requests. In accordance with an embodiment of the invention, a client device initially transmits a data request to a selected first server device using any one of a variety of methods for selecting the server. The first server device processes the request and may reject the data request if its request queue threshold is exceeded. On rejection, the first server device includes its request queue threshold in a rejection message to the client device. The client device retransmits the data request, including the request queue threshold, to a second server device, selected in a similar manner. The second server device may reject the data request if the request queue threshold of the first server device is smaller than a request queue threshold of the second server device. In a second rejection message to the client device, the second server device includes its request queue threshold. The client device transmits the data request back to the first server device, including the request queue threshold of the second server device. The first server device processes the data request and adjusts its request queue threshold based on the request queue threshold of the first and the second server devices.
• The following detailed description describes illustrative embodiments of the invention. Although specific operating environments, system configurations, user interfaces, and flow diagrams may be illustrated and/or described, it should be understood that the examples provided are not exhaustive and do not limit the invention to the precise forms and embodiments disclosed. Persons skilled in the field of computer programming will recognize that the components and process elements described herein may be interchangeable with other components or elements or combinations of components or elements and still achieve the benefits and advantages of the invention. Although the present description may refer to the Internet, persons skilled in the art will recognize that other network environments that include local area networks, wide area networks, and/or wired or wireless networks, as well as standalone computing environments, such as personal computers, may also be suitable. In addition, although the below description describes a client-server architecture, those skilled in the art will recognize that the invention may be implemented in a peer-to-peer network as well.
• Prior to discussing the details of the invention, it will be appreciated by those skilled in the art that the following description is presented largely in terms of logic operations that may be performed by conventional computer components. These computer components, which may be grouped in a single location or distributed over a wide area, generally include computer processors, memory storage devices, display devices, input devices, etc. In circumstances where the computer components are distributed, the computer components are accessible to each other via communication links.
• In the following descriptions, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to one skilled in the art that the invention may be practiced without some or all of these specific details. In other instances, well-known process elements have not been described in detail in order not to unnecessarily obscure the invention.
• FIG. 1 is a block diagram depicting a sample client-server operating environment100 suitable for distributed load balancing.Client devices102 are coupled toserver devices106 via anetwork102. In one illustrative embodiment, thenetwork104 is the Internet and theclient devices102 communicate with the severdevices106 via Web protocols such as the HTTP (Hyper Text Transport Protocol). In this embodiment, theservers106 may be Web servers arranged in a server farm accessible through the same URI (Uniform Resource Identifier). In a Web environment, theclient devices102 generally search for documents using a query statement and theserver devices106 find documents that match the query and return Web pages to theclient devices102, which are displayed in a Web browser on theclient device102. In another illustrative embodiment, for example, in a corporate environment, thenetwork104 may be a LAN. Theservers106 may offer a number of services to theclient devices102, such as FTP (File Transfer Protocol), database access services, file access services, application services, etc. In one embodiment, the client request may be for data, such as Web pages, to be returned to theclient102 by theserver106. In another embodiment, the client request may indicate a request to perform some process or task at theserver106, such as a registration or a data update at theserver106, without returning any data. In all cases, however, theserver106 processes the request from theclient102. Client devices may include, but are not limited to, a personal computer, a personal digital assistant (PDA), a mobile phone device, etc. In one illustrative embodiment, theclient device102 may include an independent component for interacting with thequeue processing application112 for routing requests. In another illustrative embodiment, theclient device102 may include a software component, integrated with another software component running on theclient device102, for interacting with thequeue processing application112 for routing requests. For example, theclient device102 may include a plug-in component integrated with a Web browser running on theclient device102. Such plug-in component may be specifically used for interactions with thequeue processing application112.
• With continued reference toFIG. 1, theserver device106 may include aserver queue108 having aqueue threshold110 and aqueue processing application112. In one illustrative embodiment, thequeue processing application112 may be an independent software component and determines whether accepting the request causes the length of theserver queue108 to exceed thequeue threshold110. In another illustrative embodiment, thequeue processing application112 may be an integral part of another application, such as a search engine, running on theserver device106. In one embodiment, the documents queried may be stored in adata store114.Data store114 may be local, such as a disk drive or disk farm, or may be remote such as a remote database.
• In an illustrative embodiment, the client-server environment comprises client computing devices and server computing devices coupled together through the Internet. In another illustrative embodiment, the client-server environment comprises client computing devices and server computing devices coupled together through a local area network (LAN) such as Ethernet. In yet another illustrative embodiment, the clients and servers may be virtual applications running on the same physical machine. Those skilled in the art will appreciate that the client and server components may take other forms comprising any combination of hardware and software without departing from the essence of a client-server architecture including requests from client components and processing of those requests by the server components.
• Although the above descriptions and the detailed descriptions that follow may refer to a single client and two servers, it will be appreciated by those skilled in the art that the present invention is not limited to a single client or two servers, but is equally applicable to any number of client and server machines/components. Additionally, even though the following descriptions may refer to the Web and Web-based protocols, those skilled in the art will appreciate that the techniques and systems described are equally applicable to other kinds of computing environments, such as LANs and multiple virtual servers embodied in a single machine. Still further, although the present invention will be described with regard to network-based client-server communications, the present invention may be applicable to either network-based client-severs, virtual client-severs or a combination thereof
• FIGS. 2A-2E illustrate distributed request routing in the client-server operating environment ofFIG. 1. With reference toFIG. 2A, when a request arrives from theclient device102, the request is queued in theserver queue108 subject to thethreshold110, as more fully described below. In one illustrative embodiment, theserver queue108 comprises fixed-size cells for storing fixed-size requests. In another embodiment, theserver queue108 comprises an array of pointers to requests. In yet another embodiment, theserver queue108 comprises an indexed table. In yet another embodiment, theserver queue108 comprises a linked-list data structure. Those skilled in the art will appreciate that a queue may be implemented in many other ways while still maintaining the essential properties of a queue data structure. Thequeue processing application112 takes the requests from the head of theserver queue108 and processes each request, possibly using data stored indata store114. Depending on the request, theserver device106 may return some data to theclient device102. In some cases the data returned to theclient102 may be an acknowledgement that the request has been processed. The requests from theclients102 may be routed to any one of themultiple servers106, which offer the same services based on the same data and information stored in correspondingdata stores114. In one illustrative embodiment, thedata stores114 coupled with the correspondingserver device106 include the same data synchronized periodically to stay consistent. In another illustrative embodiment, themultiple servers106 may be coupled to thesame data store114. In one illustrative embodiment, thedata store114 comprises a search index suitable for use by search engines. In another illustrative embodiment, thedata store114 may be a local or remote database. In yet another illustrative embodiment, thedata store114 may be a file server or an application server. The routing of client requests to the servers is based on a distributed request routing algorithm with adaptive thresholding for dynamically adjusting theserver queue thresholds110 for eachserver device106. As noted above, in a distributed request routing algorithm, no central control system exists for request routing. Rather, the request routing algorithm is implemented using all clients and servers in a cooperative and distributed manner, as more fully described below.
• Thequeue threshold110 is used by eachrespective server device106 to determine whether to accept a request sent by aclient102 for service. Thequeue processing application112 compares the current queue load with thequeue threshold110. If, upon processing the request, thequeue threshold110 is exceeded, then the request is rejected, otherwise, the request is accepted for further processing by theserver device106. In one illustrative embodiment, to route a request, theclient device102 selects afirst server device106 using any of a variety of methods for selecting the server. As described above, the selection methods can include random selection, probabilistic selection, weighted probabilistic server selection, server assignment, and the like. For example, in a weighted probabilistic server selection algorithm, a probability of selection to eachserver device106 based on a server load is calculated based on reported server loads/resources. The probability is inversely proportional to the server load. So, the server assigned the highest probability is the server with the lightest load. In one embodiment, the server load is characterized by a length of theserver queue108. The longer the length of theserver queue108, the more load theserver device106 has. In this embodiment, the server with the shortest queue length is selected. In another embodiment,server106 may be selected randomly, for example, using a pseudo-random number generator. In yet another embodiment, theserver106 may be selected according to a pre-assigned order. For example, a prioritized list of servers may be used by each client from which to select the next server for transmission of data requests. In yet another embodiment, theserver106 may be selected according to a round-robin scheme. For purposes of illustration, in one embodiment, if thefirst server device106 receives the request and thefirst queue threshold110 is exceeded, the server device rejects the request and returns a rejection message to theclient device102. The rejection message includes thefirst queue threshold110.
• With reference now toFIG. 2B, in a first alternative, theserver106 can accept the request fromclient102. The acceptance of the request from theclient102 is based on server load as represented by a length of theserver queue108. If the threshold is not exceeded, then the request is accepted by theserver106 for further processing. Theclient device102 may be notified of the acceptance of the request. Acceptance of the request by theserver106 includes placing the request at the back of therequest queue108 for later processing by theserver106.
• Referring toFIG. 2C, if theserver106 rejects the request, theclient device102 is notified of the rejection via a rejection message through thenetwork104. In one illustrative embodiment, thefirst queue threshold110 of theserver queue108 of thefirst server106 is included in the rejection message. In another illustrative embodiment, the rejection message is the request originally sent by theclient device102 with thefirst queue threshold110 appended to the request. In another illustrative embodiment, the rejection message may be a simple message including only thefirst queue threshold110 and the sever ID of thefirst server device106. Those skilled in the art will appreciate that other configurations of rejection messages may be used.
• Theclient device102 selects a second server to which the request is to be sent. In one illustrative embodiment, theclient device102 includes thefirst queue threshold110 in the request sent to thesecond server device116. For example, thefirst queue threshold110 may be included in a URI as a parameter for thesecond server116. Thesecond server116 receives the request including thefirst queue threshold110. Thesecond server device116 treats the request the same way as did thefirst server device106, namely, thequeue processing application122 determines whether accepting the request causes the length of theserver queue118 to exceed asecond queue threshold120.
• As noted above, thesecond server116 may determine that accepting the request will cause theserver queue118 to exceed thecorresponding threshold120. If such determination is made, then thesecond server116 compares thefirst queue threshold110 with thesecond queue threshold120. If thefirst queue threshold110 is less than thesecond queue threshold120, thesecond server116 also rejects the request, as illustrated inFIG. 2D. Thesecond server116 rejects the request and sends a rejection message to theclient device102, via thenetwork104, indicating the rejection. The rejection message includes thequeue threshold120 of thesecond server116. The client device receives the rejection message and resends the request to thefirst server device106. The request resent to thefirst server device106 includes thequeue threshold120 of thesecond server device116. Thefirst server device106 receives the resent request and processes the request in a manner similar to thesecond server device116, as described with respect toFIG. 2C. At this point, thequeue processing application112 compares thefirst queue threshold110 of thefirst server device106 to thesecond queue threshold120 included in the resent request. Thesecond queue threshold120 is necessarily greater than thefirst queue threshold110 because the same comparison with the same two queue thresholds was done at thesecond server device116 resulting in the rejection and resending of the request to thefirst server device106. Thequeue processing application112 adjusts thefirst queue threshold110 to be equal to thesecond queue threshold120, equalizing the queue thresholds in the first and thesecond server devices106 and116, respectively.
• As noted above, the request sent to thesecond server116 includes thefirst queue threshold110. If thefirst queue threshold110 is greater than thesecond queue threshold120, thesecond server116 accepts the request and adjusts thesecond queue threshold120 to the same value as thefirst queue threshold110, equalizing the two queue thresholds. This way, the queue thresholds are equalized dynamically. Queue threshold equalization provides uniform load distribution acrossservers106 and116 by synchronizing queue thresholds.FIG. 2E is a block diagram of the operating environment ofFIG. 1 illustrating the second server computing device accepting the request from the client computing device and adjusting its queue threshold. Thesecond server device116 accepts the request if thefirst queue threshold110 is greater than thesecond queue threshold120, and adjusts thesecond queue threshold120, as discussed above. In one illustrative embodiment, thesecond server device116 processes the request using the data stored in thedata store124. Thesecond server device116 may also accept the request if thefirst queue threshold110 is equal to thesecond queue threshold120. In this case, the second queue threshold is increased by a constant amount to enable thesecond queue118 to accept the request. Theclient device102 is notified of the acceptance of the request by thesecond server device116 and continues its interactions, such as receiving results of its request, with thesecond server device116.
• The distributed request routing algorithm described above can be partly implemented at theserver device106, and partly at theclient device102. The processing of the client requests and adjustment ofserver queue thresholds110 is done at theserver devices106 using the threshold data exchanged between the first and the second server devices through theclient device102 via rejection messages and requests.
• FIG. 3 is a flow diagram depicting an illustrative method for accepting requests and adjusting server computing device queue thresholds. The request processing routine starts atblock300 and proceeds to block305 where theserver device106 obtains a request sent from theclient device102. Atdecision block310, thequeue processing application112 determines whether adding the request to theserver queue108 causes the pendingqueue threshold110 to be exceeded. If it is determined that the pendingqueue threshold110 will not be exceeded, the routine proceeds to block340 where the request is processed by theserver device106. If it is determined that the pendingqueue threshold110 will be exceeded by accepting the request, the routine proceeds to decision block315 where it is determined whether the request is an alternate request rejected by analternate server device116 and rerouted by theclient device102 to another severdevice106.
• As discussed above, thealternate queue threshold120 is included in the rerouted request for access by anotherserver device106. With continued reference toFIG. 3, in one illustrative embodiment, thealternate queue threshold120 is appended to a URI as a parameter which can be retrieved and used by thequeue processing application112 for queue threshold comparisons. In another illustrative embodiment, thequeue threshold120 may be included in a header field of the request sent by theclient device102. If the request is an alternate request, the routine proceeds todecision box320 where it is determined whether thealternate queue threshold120 is greater than or equal to thepending queue threshold110. If so, atblock330, if thealternate queue threshold120 is greater than the pendingqueue threshold110, the pendingqueue threshold110 is set to equal thealternate queue threshold120. If thealternate queue threshold120 is equal to thepending queue threshold110, the pendingqueue threshold110 is increased by a constant amount to enable it to accept the request and the request is placed at the back of thepending queue108. Atblock335, the request is processed and the routine terminates atblock360. Back atdecision block320, if it is determined that thealternate queue threshold120 is less than the pendingqueue threshold110, the request is rejected atblock325. The rejection of the request is communicated to theclient device102 by sending a rejection message to theclient device102 including thepending queue threshold110. As discussed above, after receiving the rejection message, theclient device102 will reroute the request, including thepending queue threshold110, to another server device selected randomly. Atdecision block315, if it is determined that the request is not from analternate server device116, the request is rejected atblock325 because accepting the request would cause thepending queue threshold110 to be exceeded. The routine terminates atblock360.
• Returning to block340, if the pending queue threshold has not been exceeded atdecision block310, the routine300 proceeds to decision block345 where it is determined whether to decrease thepending queue threshold110. As discussed above, smaller queue lengths result in less request processing delay and increased overall system performance Decreasing queue threshold decreases the average queue length. The determination to decrease thepending queue threshold110 is based on the length of thepending queue108. In one illustrative embodiment, thequeue processing application112 continuously polls the length of thepending queue108 to determine whether the length is less than a predetermined fraction of thepending queue threshold110. If so, then thepending queue threshold110 is decreased. In one illustrative embodiment, the pendingqueue threshold110 is reduced by a fixed amount. In another illustrative embodiment, the pendingqueue threshold110 is reduced by an amount which is a percentage of the current value, such as ten percent. In yet another illustrative embodiment, thequeue processing application112 may be notified, via a system message generated by theserver device106, that an event associated with the pending queue length has taken place. The event may be specified based on the queue length being less than the pendingqueue threshold110 for a predetermined length of time or a predetermined number of requests. If it is determined that the pendingqueue threshold110 should be decreased, the routine proceeds to block350 where the pendingqueue threshold110 is decreased by an appropriate amount, as discussed above and the routine terminates atblock360. If it is determined that the pendingqueue threshold110 should not be decreased, the routine proceeds to block360, and the routine300 terminates atblock360.
• Test and simulation results indicate that the embodiments of the present invention improve request handling performance, in a client-server computing environment, at 99.9th percentile for different loads. This means that request handling performance is improved for almost all requests under various load conditions. Such performance improvements are very close to those achieved by hardware-based, central load distribution methods without the drawbacks of such methods discussed above. More specifically, the request handling performance is improved by lowering latency and queue thresholds.
• While illustrative embodiments have been illustrated and described, it will be appreciated that various changes can be made therein without departing from the spirit and scope of the invention.

Claims (1)

What is claimed is:

1. A computer-implemented method of operating a server computing device to process data requests, the method comprising:

receiving a first data request from a client computing device at a first server computing device, wherein the first data request includes a second queue threshold obtained from a second server computing device;

comparing, by the first server computing device, a first queue threshold associated with the first server computing device and the second queue threshold obtained from the second server computing device;

determining, by the first server computing device, whether to process the data request at the first server computing device based on the comparison of the first queue threshold and the second queue threshold; and

managing, by the first server computing device, the first queue threshold based on the comparison of the first and second queue thresholds; and

processing the data request, at the first server computing device, based on the comparison of the first and second queue thresholds.

Images