Patent Application
20190342380
Within the field of computing, many scenarios involve a processing service that performs a set of workloads using a server set. For example, a database service may provide a distributed set of servers with various capabilities, such as a query intake server that receives a query; a query processing server that parses the query; and a storage server that applies the logical operations of the parsed query over a data set.
A large-scale, distributed server set may involve a significant number of servers that perform a large number of distinct workloads for a variety of applications and/or clients. Moreover, the workloads of various applications and clients may utilize different process paths through the server set. For example, a process path for a first workload may involve a first sequence of tasks to be performed by a corresponding first sequence of servers, such as a first intake server in a first region; a query processing server; and a first storage server that stores records involved in the first workload. A process path for a second workload may involve a different sequence of tasks to be performed by a corresponding second sequence of servers, such as a second intake server in a second region; the same query processing server; a second storage server that stores records involved in the second workload.
In such scenarios, workloads may be subject to various forms of performance sensitivities. As a first such example, a workload may be sensitive to latency (e.g., a realtime application in which users or devices have to receive a result of the workload within a limited time, and in which delays may be perceptible and/or problematic). As a second such example, a workload may be sensitive to scalability and throughput (e.g., demand for the workload may fluctuate over time, and the inability of the server set to scale up to handle an influx of volume may be problematic). As a third such example, a workload may be sensitive to consistency and/or concurrency issues (e.g., a strictly deterministic workload may have to receive the same result across multiple instances, where inconsistent results may be problematic). As a fourth such example, a workload may be sensitive to replication and/or resiliency (e.g., downtime, data loss, or the failure of the workload may be problematic). In view of such sensitivities, it may be desirable to enable the server set to provide a performance guarantee for a workload, e.g., a guarantee that the server set is capable of handling a surge of volume up to a particular amount while maintaining latency below a particular level.
In multitenant scenarios, workloads for different applications and/or clients may share a process path. Other workloads may take different process paths through the server set, but may both include one or more servers of the server set that are to be shared by the workloads. Servers may be shared among workloads that are associated with different clients and/or applications; in some scenarios, a server may concurrently handle workloads on behalf of hundreds or even thousands of different applications or clients.
A variety of multi-tenancy techniques may be utilized to ensure that a first workload on behalf of a first client or application does not interfere with a second workload on behalf of a second client or application. As a first such example, the server may utilize process and data isolation techniques to ensure that a first workload on behalf of a first client cannot achieve unauthorized access to a second workload on behalf of a second client, including accessing data owned by the second workload or even identifying the presence of the second workload, including the second client or application for which the second workload is processed.
As a second such example, the server may protect against resource overutilization. For instance, if a first workload through the server begins exhibiting a surge of volume that exceeds the share of computing resources allocated to the first workload, the use of a resource-sharing technique may enable the server to confine the consequences of the excessive volume to the first workload and to avoid impacting the processing of the second workload, such that a performance guarantee extended to the second workload remains fulfilled. In this manner, servers may allocate and regulate computing resource utilization to promote fulfillment of performance guarantees and allocate computational resources fairly over all of the workloads handled by the server.
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 factors or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
The processing of workloads through a large-scale server set, in view of performance guarantees, may encounter difficulties due to the sequential nature of the workloads and interactions of servers along various process paths. For example, a workload may involve a process path that involves a sequence of tasks to be performed by an intake server, a query processing server, and a storage server, and to fulfill a performance guarantee. While the respective servers may utilize computational resource allocation to handle the individual tasks of the workload, a problem may arise if the storage server begins to experience an excessive computational load. Such excessive computational load may arise, e.g., due to an over-allocation of tasks onto the storage server; a shortage of computational resources, such as a reduction of network bandwidth; or an unanticipated surge of processing, such as a failure of a process on the storage server that necessitates the invocation of a recovery process. In addition to slowing the processing of the workload through the storage server, the excessive computational load of the storage server may create a processing jam between the storage server and the query processing server that is upstream of the storage server in the process path. For example, the query processing server may continue to handle the query processing task of the workload at the same rate, but the rate at which the query processing server is able to pass the workload to the storage server may be diminished. The query processing server may address the discrepancy in various ways, such as utilizing an outbound queue for the completed workload of processed queries; however, if the reduced processing rate of the storage server persists, the outbound queue may overflow. Moreover, the additional processing burden placed upon the query processing server may propagate the processing jam upward, even to the point of potentially affecting other workloads that have a processing path through the query processing server that do not utilize the storage server. The resulting gridlock may cause a widespread failure of performance guarantees for a variety of workloads due the processing jam between the storage server and the upstream query processing server.
In view of such problems, it may be desirable to configure the server set to evaluate the processing paths of the various workloads, and to provide techniques for mitigating a processing jam that may arise between a particular server and a downstream server. In particular, a server of a process path may estimate, measure, and/or monitor its processing capabilities, and compare such processing capabilities with the performance guarantees of the workloads utilizing a process path through the server. If the server detects a risk of failing the performance guarantee (e.g., due to an overprovisioning of the server or a computational resource shortage), the server may transmit a performance capability alert to an upstream server of the server path, as an indication that the workload being passed to the server may be too large to ensure that the performance guarantees are met. The upstream server that receives the performance capability alert may respond by rate-limiting its processing of the workload, within the performance guarantee, thereby downscaling the processing rate of the upstream server upon the workload to match the diminished processing rate of the downstream server. In some scenarios, the upstream server may propagate the performance capability alert further upstream. If the performance capability alert reaches a first server of the server path, the first server may refuse or slow a workload acceptance rate. In this manner, the server set may adapt the acceptance of the workload for which the process path is capable of fulfilling the performance guarantee, and in response to fluctuating processing capabilities (including an unexpected loss of processing capability) of the servers of the server path.
A first embodiment of the presented techniques involves a server of a server set that performs workloads according to a performance guarantee. The server comprises a processor and a memory storing instructions that, when executed by the processor, cause the server to operate in accordance with the techniques presented herein. In particular, the server performs a task of the workload according to a performance guarantee, wherein the workload is processed through the server set according to a process path. On condition of receiving, from a downstream server of the process path, a performance capability alert indicating that a computational load of the downstream server risks failing the performance guarantee for the workload, the server may rate limit the task of the workload to reduce the computational load of the downstream server. After completing the task (to which the rate-limit may have been applied), the server delivers the workload to the downstream server.
A second embodiment of the presented techniques involves a method of configuring a server of a server set to participate in workloads. The method involves, executing, by a processor of the server, instructions that cause the server to operate in accordance with the techniques presented herein. In particular, the server receives a workload from an upstream server of a process path of the workload, wherein the workload is associated with a performance guarantee. The server performs a task on the workload, and further identifies a performance capability of the server and compares the performance capability with the performance guarantee of the workload. Responsive to determining that the performance capability risks failing the performance guarantee, the server transmits a performance capability alert to the upstream server. Additionally, responsive to receiving a performance capability alert from a downstream server of the process path, the server rate limits a receipt of additional workloads from the upstream server.
A third embodiment of the presented techniques involves a method of configuring a server set to perform a workload according to a performance guarantee. The method involves configuring a server within a process path of the workload through the server set to operate in accordance with the techniques presented herein. The method further involves configuring the server to perform a task on the workload according to the performance guarantee. The method further involves configuring the server to receive a performance capability alert from the downstream server, wherein the performance capability alert indicates that a computational load of the downstream server risks failing the performance guarantee for the workload. The method further involves configuring the server to rate-limit the task of the server to reduce the workload delivered to the downstream server. The method further involves configuring the server to, after performing the task on the workload, deliver the workload to a downstream server of the process path.
To the accomplishment of the foregoing and related ends, the following description and annexed drawings set forth certain illustrative aspects and implementations. These are indicative of but a few of the various ways in which one or more aspects may be employed. Other aspects, advantages, and novel features of the disclosure will become apparent from the following detailed description when considered in conjunction with the annexed drawings.
The claimed subject matter is now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the claimed subject matter. It may be evident, however, that the claimed subject matter may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing the claimed subject matter.
In this example scenario 100, the server set 106 comprises a set of server 108 that respectively perform a task 110. For example, a pair of intake servers 108 may serve as a front-end, client-facing interface that accepts queries to be processed on behalf of the clients 102; a query processing server 108 that parses queries, such as translating the query from a query language into a sequence of logical relationships to be applied over the data set; and a pair of storage servers 108 that store a replica or a portion of the data set over which the queries are to be applied. The servers 108 may be arranged such that the workloads 104 of the clients 102 are processed according to a process path 112, e.g., a sequence of servers 108 that respectively apply a task 110 to the workload 104 and deliver the partially processed workload 104 to the next, downstream server 108 in the process path 112. The process path 112 may enable a pipelined evaluation of workloads 104 that enable the servers 108 to apply the tasks 110 in the manner of an assembly line, thereby reducing idle processing capacity and promoting the scalability of the server set 106 to handle a significant volume of workloads 104 in a concurrent manner.
As further shown in the example scenario 100 of
As further shown in the example scenario 100 of
Due to the performance dependencies of the workloads 104, the server set 106 may extend to each client 102 a performance guarantee 114 of a performance capability of the server set 106 to handle the workload 104. For example, the server set 106 may extend to the first client 102 a performance guarantee 114 that processing of the majority of workloads 104 (e.g., 95% of workloads 104) will complete within 10 milliseconds. For the second client 102, the server set 106 may offer a performance guarantee 114 of correct processing of the second workload 104 up to a defined volume, such as 1,000 requests per second. The server set 106 may be arranged to fulfill the performance guarantees 114 of the workloads 104, e.g., by supplementing, adapting, and/or optimizing the configuration of servers 108 comprising the data set.
When a server set 106 is arranged such that a particular server 108 processes different workloads 104, particularly for different clients 102, problems may arise due to the concurrent and/or consecutive sharing of the server 108. As a first example, the server 108 may have to secure and isolate the workloads 104 of the first client 102 from the workloads 104 of the second client 102, e.g., to prevent the second load 104 from accessing proprietary information of the first client 102 and tampering with the operation of the first workload 104. The significance of isolation may grow with the scalability of the server set 106; e.g., a particular server 108 may process hundreds or even thousands of workloads 104 of various clients 102, and safeguarding the information of each client 102 may be a high priority. As a second example, the server 108 may comprise a limited set of computational resources, such as processor capacity, storage capacity, and network bandwidth. An overconsumption of the computational resources of the query processing server 108 by the first workload 104 may create a shortage of computational resources of the server 108 for the second workload 104, such as limited processing capacity; an exhaustion of available storage; and/or constrained network bandwidth. Such overconsumption by the first workload 104 may lead to delays or even a failure in the processing of the second workload 104, including a failure of the performance guarantee 114 of the second workload 104, such as an inability of the query processing server 108 to scale up in order to handle peak volume of the second client 102. In view of such concerns, techniques may be utilized to allocate and compartmentalize the computational resources of the server 108 for each workload 104, such as processor time-slicing and per-workload quotas or caps on storage and network capacity. Using such techniques, a server 108 may limit the adverse effects of an overconsumption of resources to the workload 104 responsible for the overconsumption; e.g., increased processing demand by the first workload 104 may result in delayed completion of the first workload 104 without impacting the processing of the second workload 104. However, resource limitations may also consume computational resources (e.g., processor time-slicing among a set of workloads 104 may present overhead due to context-switching). Such inefficiency may scale with the scalability of the server set 106, such as using a particular server 108 to process hundreds or even thousands of workloads 104, so the isolation and allocation techniques may have to be implemented with careful attention to efficiency.
Other developments may also present a potential source of failure of a performance guarantee 114 of a particular workload 104. For example, computational resources may be limited by systemic factors, such as a shortage of storage capacity due to a failure of a hard disk drive in a storage array, or a surge in a computational process, such as a background maintenance process. An introduction of line noise into a network connection, such as due to electromagnetic interference or a faulty cable, may lead to diminished throughput and increased latency. The server set 106 may experience a failure of a particular server 108 and may have to re-route the workload 104 of the failed server to other servers 108 of the server set 106, thereby increasing the computational load of the individual servers 108. Any such change in the performance of the server set 106 may interfere with the process path 112 of a workload 104, which may risk failing the performance guarantee 114.
In view of such risks, load-balancing techniques are often utilized to detect and mitigate computational overload of a particular server 108. For example, the respective servers 108 of the server set 106 may include a monitoring process of a performance capability, such as available processor capacity, storage, network throughput, and latency. The performance capabilities may also include considerations such as resiliency to data loss, e.g., the volume of data stored by the server 108 that has not yet been transmitted to a replica, as a measure of the risk of data loss in the event of a failure of the server 108. The server 108 may track the performance capabilities and, if detecting a potential shortage that risks failing a performance guarantee 114, may invoke a variety of “self-help” measurements to alleviate the shortage. For example, in the event of a shortage of processing capacity, the server 108 may place the processor into a “boost” mode; awaken and utilize dormant processing capacity, such as additional processing cores; and/or reduce or suspend some deferrable processing, such as maintenance tasks. In the event of a shortage of storage capacity, the server 108 may delete or compress data that is not currently in use, including significant data that may later be restored from a replica. In the event of a shortage of network capacity, the server 108 may suspend processes that are consuming network capacity, or shift network bandwidth allocation from processes that are tolerant of reduced network bandwidth and latency to processes that are sensitive to constrained network bandwidth or latency.
Alternatively or additionally to such “self-help” techniques, the server 108 may report the performance capability shortage to a network administrator or network monitoring process, which may intercede to reconfigure the server set 106. For example, the computational load of a storage server 108 may be alleviated by provisioning a new server 108, replicating the data set onto the new server 108, and altering process paths 112 to utilize the new server 108. However, reconfiguration of the architecture of the server set 106 may be a comparatively expensive step, and/or may involve a delay to implement, during which time the performance guarantee 114 of a workload 104 may fail.
However, these and other techniques may be inadequate to address a particular source of interference with the processing of the server set 106 that may jeopardize performance guarantees 114.
As further illustrated in the example scenario 100 of
The query processing server 108 may respond to the processing jam 118 in numerous ways. For example, the query processing server 108 may retry the delivery of the workload 104 for a period of time, in case the processing jam 118 is ephemeral and is momentarily alleviated. The query processing server 108 may utilize an outbound queue for the workload 104 that the storage server 108 may be able to work through and empty when the processing capacity shortage 116 is alleviated, or that may be transferred to a replica of the storage server 108 following a reconfiguration of the server set 106. However, these techniques may also fail if the processing jam 118 is prolonged and a substitute for the storage server 108 is unavailable. The outbound queue of the query processing server 108 may also overflow, or the workloads 104 allocated to the query processing server 108 may begin to starve, inducing a failure of the performance guarantee 114. In some cases, the source of the fault may be misattributed to the query processing server 108, since the performance guarantees 114 failed while the query processing server 108 retained the workloads 104 for a prolonged period. For example, an automated diagnostic process may identify the query processing server 108 as a processing bottleneck, and may initiate a failover of the query processing server 108 that fails to resolve the actual limitation of the performance of the server set 106.
As further illustrated in the example scenario 100 of
In view of the problems depicted in the example scenario 100 of
Additionally, because such incidents may occur suddenly, may quickly present risks to the failure of a performance guarantee 114, and may often be only transient, it may be advantageous to utilize techniques that may be applied rapidly and without involving a significant and potentially expensive allocation of resources, such as inducing failover of the afflicted server 108 to a substitute server 108. It may also be advantageous to utilize techniques that may be applied automatically in the locality of the afflicted server 108, without necessarily resorting to a centralized manager that holistically evaluates the server set 106 to identify potential solutions, and/or without involving a human administrator who may not be able to respond to the processing capacity shortage 116 in due time.
Additionally, because the effects of a processing jam 118 may spill over onto other servers 108 in a process path 112, it may be advantageous to provide techniques that may be easily propagated to a broader neighborhood of the afflicted server 108, and therefore expand to incorporate the other servers 108 in the resolution of the processing capacity shortage 116.
At a first time 210, the servers 108 of the server set may apply the tasks 110 to the workload 104, where each server 108 completes the task 110 on a portion of the workload 104 and delivers the partially completed workload 104 to the next downstream server 108 of the process path 112. Additionally, the servers 108 may individually monitor the performance capabilities 202, and compare the performance capabilities 202 with the performance guarantee 114. For example, if the performance guarantee 114 comprises a maximum latency, such as 10 milliseconds, the respective servers 108 may monitor the duration of completing the task 110 over a selected portion of the workload 104 to ensure that the task 110 is completed within 2.5 milliseconds on each server 108. If the performance guarantee 114 comprises a maximum volume of unreplicated data that is subject to data loss, the server 108 may monitor and manage a queue of unreplicated data that is awaiting synchronization with a replica. In this manner, the respective servers 108 may ensure that the performance capabilities 202 of the individual servers 108 are sufficient to satisfy the performance guarantee 114; such that maintaining adequate individual performance capabilities 202 of all servers 108 in the server path 112 results in a satisfaction of the performance guarantee.
However, at a second time 212, the third server 108 in the process path 112 may detect a diminished performance capability 202, such as limited processing capacity, storage capacity, or network bandwidth. Comparison of the diminished performance capability 202 with the performance guarantee 114 may reveal a processing capacity shortage 116 that introduces a risk 204 of failing the performance guarantee 114 for the workload 114.
In some circumstances, the third server 108 may be capable of utilizing “self-help” measures to restore the performance capability 202. In other circumstances, the processing capacity shortage 116 may rapidly be identified as severe and unresolvable, such as a complete failure of a storage device that necessitates substitution of the third server 108. However, in some circumstances, the diminished performance capability 202 may be resolved by temporarily reducing the workload 104 handled by the third server 108. Such reduction of the workload 104 may be achieved by reducing the delivery of the workload 104 to the third server 108 by the upstream servers 108 of the process path 112. Such reduction may provide a window of opportunity in which the third server 108 may apply the available performance capabilities 202 to a workload 104 of reduced volume, which may enable the third server 108 to catch up with the volume of the workload 104. For instance, the third server 108 may utilize an input buffer of workloads 104 delivered by the upstream server 108. If the rate at which the workload 104 is delivered into the input buffer exceeds the rate at which the third server 108 removes and completes the workload 104 from the input buffer, the input buffer may steadily grow to reflect a deepening processing queue with a growing latency. Reducing the input rate of delivery of the workload 104 into the input buffer below the rate at which the third server 108 takes the workload 104 out of the input buffer may shrink the input buffer and enable the third server 108. When the input buffer is depleted or at least reduced to an acceptable latency, and/or the cause of the diminished performance capability 202 and processing capacity shortage is resolved, the input rate to the input buffer may be restored.
As further shown in the example scenario 200 of
The second server 108 may continue to apply the rate limit 208 to task 110 for the duration of the processing capacity shortage 116 of the third server 108. For example, the third server 108 may eventually report an abatement of the processing capacity shortage 116, or the second server 108 may detect such abatement, e.g., by detecting an emptying of the outbound queue to the third server 108, at which point the second server 108 may remove the rate limit 208 and resume unrate limited processing of the task 110 for the workload 104. Alternatively, if the processing capacity shortage 116 is prolonged or indefinite, or if the second server 108 identifies that applying the rate limit 208 to the task 110 may impose a new risk 204 of failing the performance guarantee 114, the second server 108 may propagate the performance capability alert 206 to the next upstream server 108 of the process path 112, i.e., the first server 108. The first server 108 may similarly respond to the performance capability alert 206 by applying a rate limit 208 to the task 110 of the first server 108 over the workload 104. The first server 108, as the intake point of the workload 104, may reduce the commitment of the entire process path 112 to a smaller workload volume over which the performance capability 202 may be guaranteed even while afflicted with the processing capacity shortage 116. In this manner, the backward propagation of performance capability alerts 206 and the application of a rate limit 208 the task 110 of the second server 108 operate as a form of “backpressure” on the upstream servers 108 of the process path 112, which reduces the computational overload of the third server 108 and promotes the satisfaction of the performance guarantee 114 of the server set 106 over the workload 104 in accordance with the techniques presented herein.
A first technical effect that may arise from the techniques presented herein involves the resolution of the processing capacity shortage 116 of a downstream server 108 and the risk 204 of failing the performance guarantee 114 of the workload 104 through the application of backpressure on upstream servers 108 of the process path 112.
The use of a rate limit 208 by the second server 108 in accordance with the techniques presented herein may effectively address the process capacity shortage 116 and the risk 204 of failure of the performance guarantee 114 of the workload 104 in numerous ways. As a first such example, it may be feasible for the second server 108 to apply the rate limit 208 to the task 110 may be feasible by the second server 108 without the introduction of the rate limit 208 exacerbating the risk 204 of failing the performance guarantee 114. For example, the second server 108 may have a surplus performance capability 202, and may be capable of working through the workload 104 significantly faster than required by the performance guarantee 114 (e.g., the second server 108 may have a maximum allocation of 2.5 milliseconds to perform the task 110 over the workload 104 within the performance guarantee 114, but may be capable of completing the task 110 in only 0.7 milliseconds). That is, the application of the rate limit 208 to the task 110 may offload some of the delay caused by the processing capacity shortage 116 from the third server 108 to the second server 108, thus enabling the third server 108 to work through a backlog of the workload 104 and restore the performance capability 202.
As a second such example, the second server 108 and third server 108 may share two workloads 104, wherein the processing capacity shortage 116 may introduce a risk 204 of failing the performance guarantee 114 of the first workload 104, but may pose no risk 204 of failing the performance guarantee 114 of the second workload 104 (e.g., the first workload 104 may be sensitive to latency, while the second workload 104 may be relatively tolerant of latency). The application of the rate limit 208 to the task 110 of the second server 108 may reduce the rate of delivery to the third server 108 of both the first workload 104 and the second workload 104. The reduced volume of the second workload 104 may enable the third server 108 to apply the performance capability 202 to work through a backlog of the first workload 104 and therefore alleviate the processing capacity shortage 116, without introducing a risk 204 of failing a performance guarantee 114 for the second workload 104 that is not significantly affected by increased latency.
As a third such example, the use of a performance capability alert 206 and rate limit 208 may be applied to further upstream servers 108. For example, in the example scenario 200 of
A second technical effect that may arise from the techniques presented herein involves the capability of the server set 106 to respond to performance capacity shortages in an efficient, rapid, and automated manner. As a first such example, the techniques presented herein may be applied without conducting a holistic, extensive analysis of the capacity of the server set 106, such as may be performed by a network monitoring or network administrator, to determine the root cause of the processing capacity shortage 116 and to assess the available options. Rather, the server 108 afflicted by diminished performance capability 202 may simply detect the processing capacity shortage 116 and transmit the performance capability alert 206 to the upstream server 108. As a second such example, the techniques presented herein do not involve a significant and potentially expensive reconfiguration of the server set 106 or a commitment of resources, such as provisioning a substitute server for the afflicted server 108, which may involve remapping associations within the server set 106 and/or introduce a delay in the recovery process. In some cases, the delay involved in applying the recovery may outlast the duration of the processing capacity shortage 116. In other cases, the performance guarantee 114 for the workload 104 may fail during the delay involved in applying such heavy recovery techniques. In some circumstances, such recovery may impose additional computational load on the afflicted server 108, thus hastening the failure of the performance guarantee 114. By contrast, the comparatively simple techniques presented herein are applicable merely by transmitting the performance capability alert 206 to the upstream server 108 and causing the second server 108 to apply the rate limit 208 to the task 110 may be applied rapidly and with a negligible expenditure of resources, and may therefore be effective at resolving some processing capacity shortages 116, particularly serious but ephemeral shortages, that other techniques may not adequately address. Moreover, the transmission of the performance capability alert 206 and the application of the rate limit 208 to the task 110 utilize currently existing and available resources and capabilities of the downstream and upstream servers (e.g., processor clock rate adjustment, adjustment of thread and process priorities, and/or the use of queues), and do not depend upon the introduction of complex new process management machinery or protocols.
A third technical effect that may arise from the techniques presented herein involves the extension of the process to reduce or avoid the risk 204 of failing the performance guarantee 114 altogether. In the example scenario 200 of
The example system 308 comprises a task processor 310, which performs a task 110 of the workload according to a performance guarantee, wherein the workload 104 is processed through the server set 106 according to a process path 112 that includes an upstream server 108 and a downstream server 108 relative to the example server 302. The example system 308 also includes a task rate limit 314, which receives a performance capability alert 206 from a downstream server 108 of the process path 112, e.g., in response to a comparison of the performance capability 202 of the downstream server 108 with the performance guarantee 114 of the workload 104, which indicates a processing capacity shortage 116 and a risk 204 of failing the performance guarantee 114 of the workload 104. Responsive to the performance capability alert 206, the task rate limit 314 applies a rate limit 208 to the task 110 performed on the workload 104 to reduce the computational load of the downstream server 108. The example system 308 also includes a workload streamer 312, which, after completion of the task 110 on the workload 104, delivers the workload 104 to the downstream server 108 of the process path 112. In this manner, the example system 308 enables the example server 302 to apply the task 110 to the workload 104 as part of the process path 112 in accordance with the techniques presented herein.
The example method 400 begins at 402 and involves executing 404, by the server, instructions that cause the server to perform in the following manner. The execution of the instructions causes the server 108 to receive 406 a workload 104 from an upstream server 108 of a process path 112, wherein the workload 104 is associated with a performance guarantee 114. The execution of the instructions also causes the server 108 to perform 408 a task 110 on the workload 104. The execution of the instructions also causes the server 108 to identify 410 a performance capability 202 of the server 108. The execution of the instructions also causes the server 108 to compare 412 the performance capability 202 with the performance guarantee 114 of the workload 104. The execution of the instructions also causes the server 108 to respond to determining that the performance capability 202 risks failing the performance guarantee 114 by transmit 414 a performance capability alert 206 to the upstream server 108. The execution of the instructions also causes the server 108 to respond to a performance capability alert 206 received from a downstream server 108 of the process path 112 by rate-limiting 416 the task 110 performed on the workload 104 to reduce the computational load of the downstream server 108. In this manner, the example method 400 may enable the server 108 to process the workload 104 as part of the process path 112 in accordance with the techniques presented herein, and so ends at 418.
The first example method 500 begins at 502 and involves configuring a server 108 of the server set 106 that is within the process path 112 to process the workload 104 in the following manner. The server 108 performs 506 a task 110 on the workload 104 according to the performance guarantee 114. The server 108 further receives 508 a performance capability alert 206 from the downstream server 108, wherein the performance capability alert 206 indicates that a computational load of the downstream server 108 risks failing the performance guarantee 114 for the workload 104. Responsive to the performance capability alert 206, the server 108 further rate-limits 510 the task 110 of the server 108 to reduce the workload delivered to the downstream server 108. After performing the task 110 on the workload 104, the server 108 further delivers 512 the workload 104 to a downstream server 108 of the process path 112. In this manner, the example method 500 may enable the server 108 to operate as part of a server set 106 to participate in the processing of the workload 104 in accordance with the techniques presented herein, and so ends at 514.
Still another embodiment involves a computer-readable medium comprising processor-executable instructions configured to apply the techniques presented herein. Such computer-readable media may include various types of communications media, such as a signal that may be propagated through various physical phenomena (e.g., an electromagnetic signal, a sound wave signal, or an optical signal) and in various wired scenarios (e.g., via an Ethernet or fiber optic cable) and/or wireless scenarios (e.g., a wireless local area network (WLAN) such as WiFi, a personal area network (PAN) such as Bluetooth, or a cellular or radio network), and which encodes a set of computer-readable instructions that, when executed by a processor of a device, cause the device to implement the techniques presented herein. Such computer-readable media may also include (as a class of technologies that excludes communications media) computer-computer-readable memory devices, such as a memory semiconductor (e.g., a semiconductor utilizing static random access memory (SRAM), dynamic random access memory (DRAM), and/or synchronous dynamic random access memory (SDRAM) technologies), a platter of a hard disk drive, a flash memory device, or a magnetic or optical disc (such as a CD-R, DVD-R, or floppy disc), encoding a set of computer-readable instructions that, when executed by a processor of a device, cause the device to implement the techniques presented herein.
An example computer-readable medium that may be devised in these ways is illustrated in
The techniques discussed herein may be devised with variations in many aspects, and some variations may present additional advantages and/or reduce disadvantages with respect to other variations of these and other techniques. Moreover, some variations may be implemented in combination, and some combinations may feature additional advantages and/or reduced disadvantages through synergistic cooperation. The variations may be incorporated in various embodiments (e.g., the first example server 302 and/or the example system 308 of
E1. Scenarios
A first aspect that may vary among implementations of these techniques relates to scenarios in which the presented techniques may be utilized.
As a first variation of this first aspect, the presented techniques may be utilized with a variety of servers 108 and server sets 106, such as workstations, laptops, consoles, tablets, phones, portable media and/or game players, embedded systems, appliances, vehicles, and wearable devices. The server may also comprise a collection of server units, such as a collection of server processes executing on a device; a personal group of interoperating devices of a user; a local collection of server units comprising a computing cluster; and/or a geographically distributed collection of server units that span a region, including a global-scale distributed database. Such servers 108 may be interconnected in a variety of ways, such as locally wired connections (e.g., a bus architecture such as Universal Serial Bus (USB) or a locally wired network such as Ethernet); locally wireless connections (e.g., Bluetooth connections or a WiFi network); remote wired connections (e.g., long-distance fiber optic connections comprising Internet); and/or remote wireless connections (e.g., cellular communication). Additionally, such servers 108 may serve a variety of clients 102, such as a client process on one or more of the servers 108; other servers 108 within a different server set 106; and/or various client devices that utilize the server 108 and/or server group on behalf of one or more clients 102 and/or other devices.
As a second variation of this first aspect, the server set 106 may present a variety of services that involve applying tasks 110 to workloads 108. As a first such example, the service may comprise a distributed database or data storage system, involving tasks 110 such as receiving the data; storing the data; replicating and/or auditing the data; evaluating queries over the data; and/or running reports or user-defined functions over the data. As a second such example, the service may comprise a content presentation system, such as a news service, a social network service, or social media service, which may involve tasks 110 such as retrieving and storing content items; generating new content items; aggregating content items into a digest or collage; and transmitting or communicating the content items to clients 102. As a third such example, the service may comprise a media presentation system, which may involve tasks 110 such as acquiring, storing, cataloging, and archiving the media objects; rendering and presenting media objects to clients 102; and/or tracking engagement of the clients 102 with the media objects. As a fourth such example, the service may comprise a software repository, which may involve tasks 110 such as storing and cataloging software; deploying software to various clients 102; and receiving and applying updates such as patches and upgrades to the software deployed of the clients 102. As a fifth such example, the service may comprise a gaming system, which may involve tasks 110 such as initiating game sessions; running game sessions; and compiling the results of game sessions among various clients 102. As a sixth such example, the service may comprise an enterprise operational service that provides operational computing for an enterprise, which may involve tasks 110 such as providing a directory of entities such as individuals and operating units; exchanging communication among the entities; controlling and managing various processes; monitoring and logging various processes, such as machine sensors; and generating alerts. Those of ordinary skill in the art may devise a range of scenarios in which a server set 106 configured in accordance with the techniques presented herein may be utilized.
E2. Performance Capabilities and Performance Guarantees
A second aspect that may vary among embodiments of the techniques presented herein involves the performance capabilities 202 monitored by the servers 108 and the comparison with performance guarantees 114 over the workload 104 to identify a processing capacity shortage 116 and a risk 204 of failing the performance guarantee 114.
As a first variation of this second aspect, the performance capabilities 202 may include, e.g., processor capacity; storage capacity; network bandwidth; availability of the server set 106; scalability to handle fluctuations in the volume of a workload 104; resiliency to address faults such as the failure of a server 108; latency of processing the workload 104 through the server set 106; and/or adaptability to handle new types of workloads 104.
As a second variation of this second aspect, the performance guarantees 114 of the workloads 104 may involve, e.g., a processing latency, such as a maximum end-to-end processing duration for processing the workload 104 to completion; a processing throughput of the workload 104, such as a sustainable rate of completed items; a processing consistency of the workload 104, such as a guarantee of consistency among portions of the workload 104 processed at different times and/or by different servers 108; scalability to handle a peak volume of the workload 104 to a defined level; a processing replication of the workload 104, such as a maximum volume of unreplicated data that may be subject to data loss; and/or a minimum availability of the server set 106, such as a “sigma” level.
As a third variation of this second aspect, a server 108 may identify the performance capabilities 202 in various ways. As a first such example, a server 108 may predict the performance capability of the server 202 over the workload 104, such as an estimate of the amount of time involved in applying the task 110 to the workload 104 or a realistically achievable throughput of the server 108. Such predictions may be based, e.g., upon an analysis of the workload 104, a set of typical performance characteristics or heuristics of the server 108, or previous assessments of processing the task 110 over similar workloads 104. Alternatively or additionally, the server 108 may measure the performance capability 202 of the server while performing the workload 104. Such measurement may occur with various granularity and/or periodicity, and may involve techniques such as low-level hardware monitors (e.g., hardware timers or rate meters) and/or high-level software monitors (e.g., a timer placed upon a thread executing the task 110). As a third such example, a server 108 may not actively monitor the performance capability 202 but may receive an alert if an apparent processing capacity shortage 116 arises (e.g., a message from a downstream server 108 of a reduced delivery of the completed workload 104).
As a fourth variation of this second aspect, a server 108 may and compare such performance capabilities 202 with the performance guarantees 114 of the workload 104 in various ways. As a first such example, the server 108 may compare an instantaneous measurement of the performance capability 202 with an instantaneous performance guarantee 114, such as a current data transfer rate compared with a minimum acceptable data transfer rate, and/or periodic measurements, such as a number of completed tasks 110 over a workload 104 in a given period vs. a quota of completed tasks 110. As a second such example, the server 108 may compare a trend in the performance capability 202, e.g., detecting a gradual reduction of processing capacity over time that, while currently satisfying the performance guarantee 114, may indicate an imminent or eventual risk 204 of failing the performance guarantee 114, such as a gradually diminishing rate of completed tasks 110. As a third such example, a workload 104 may be associated with a set of at least two performance guarantees 114 for at least two performance capabilities 202 and a priority order of the performance guarantees 114 (e.g., a first priority of a maximum latency of processing individual tasks 110 over the workload 104 at a typical rate of 10 milliseconds, but in the event of an ephemeral failure of the first performance guarantee 114, a second priority of a maximum throughput of processing tasks 110 of the workload 104 within a given period, such as at least 100 tasks completed per second). Such prioritization may enable the performance guarantees 114 to be specified in a layered or more nuanced manner. The server 108 may compare the respective performance capabilities 202 of the server 202 according to the priority order to evaluate the risk 204 of failing the collection of performance guarantees 114 for the workload 114.
As a fifth variation of this third aspect, a performance capability alert 206 may be relayed from a downstream server 108 to an upstream server 108 in a variety of ways. As a first such example, the performance capability alert 206 may comprise a message initiated by the downstream server 208 and transmitted to the upstream server 108 in response to the identification of a risk 204 of failing the performance guarantee 114. The message may be delivered in-band (e.g., as part of an ordinary communication stream) or out-of-band (e.g., using a separate and dedicated communication channel). As a second such example, the performance capability alert 206 may comprise a performance metric that is continuously and/or periodically reported by the downstream server 108 to the upstream server 108 (e.g., an instantaneous measurement of processing capacity), where the upstream server 108 may construe a fluctuation of the metric as a performance capability alert 206 (e.g., the downstream server 108 may periodically report its latency in completing the task 110 over the workload 104, and the metric may reveal an excessive latency that is approaching a maximum latency specified by the performance guarantee 114). As a third such example, the performance capability alert 206 may comprise part of a data structure shared by the downstream server 108 and the upstream server 108, such as a flag of a status field or a queue count of a workload queue provided at the interface between the downstream server 108 and the upstream server 108. As a fourth such example, the performance capability alert 206 may comprise a function of the upstream server 108 that is invoked by the downstream server 108, such as an API call, a remote procedure call, a delegate function, or an interrupt that the downstream server 108 initiates on the upstream server 108. Many such techniques may be utilized to compare the performance capability 202 to the performance guarantee 14 to identify a risk 204 of failing the performance guarantee 114 of the workload 104 in accordance with the techniques presented herein.
E3. Task Rate-Limiting
A third aspect that may vary among embodiments of the presented techniques involves the manner of applying a rate limit 208 to a task 110 over the workload 104 in accordance with the techniques presented herein.
As a first variation of this third aspect, illustrated in a first example scenario 710, a server 302 may rate limit a task 110, responsive to receiving a performance capability alert 206 from a downstream server 108, by reducing the performance capabilities 202 of the server 108. As a first example, the server 208 may reduce a processor speed 702 of a processor 304, such as reducing the clock speed or core count that is applied to perform the task 110 over the workload 104. As a second example, the server 302 may reduce a thread priority of the task 110, such that a multiprocessing processor 304 performs increments of the task 110 less frequently, or even suspends the task 110 temporarily if other tasks 110 are of higher priority. Other types of performance capabilities 202 that may be reduced for the workload 104 include volatile or nonvolatile memory allocation; network bandwidth; and/or access to a peripheral device such as a rendering pipeline. In some scenarios, the server 302 may rate limit the task 110, relative to a severity of the performance capability alert 206, such as the degree of constraint on the network capacity of the downstream server 108 or the pending volume of unprocessed work that the downstream server 108 has to work through to alleviate the performance capability alert 206.
As a second variation of this third aspect, illustrated in a second example scenario 712, a server 302 may rate limit a task 110, responsive to receiving a performance capability alert 206 from a downstream server 108, by temporarily refusing to accept the workload 104 from an upstream server 108, e.g., by initiating a processing jam 118. The processing jam 118 may be initiated in increments, such that the upstream server 108 is only capable of sending batches of the workload 104 to the server 302 in intervals that are interspersed by a cessation of the workload 104 arriving at the server 302. Alternatively or additionally, the server 302 may reduce an acceptance rate of the workload 104 from the upstream server 108; e.g., the upstream server 108 may utilize an output queue of workload 104 to deliver to the server 302, and the server 302 may only check the output queue at an interval, or at a reduced interval, thereby slowing the rate at which the server 302 accepts workload 104 from the upstream server 108 and delivers the workload 104 to the downstream server 108.
As a third variation of this third aspect, illustrated in a third example scenario 714, a server 108 may rate limit a task 110, responsive to receiving a performance capability alert 206 from a downstream server 108, by utilizing one or more queues that slow the intake and/or delivery of the workload 104 to the downstream server 108. As a first such example, the server 302 may implement an input queue 704 that enqueues the task 110 for the workload 104 for a delay period, and withdraw the task 110 from the input queue to perform the task 110 on the workload 104 only after the delay period. As a second such example, the server 302 may implement an output queue 706 with a delivery delay that slows the rate at which processed work is delivered to the downstream server 108.
As a fourth variation of this third aspect, a server 302 may rate limit the task 110 over the workload 104 only within the performance guarantee 114 of the workload 104. For example, the performance guarantee 114 may comprise a maximum 10-millisecond latency of processing the workload 104 through the process path 112, and a particular server 302 may be permitted to expend up to 2.5 milliseconds per task 110 while the task 110 remains in conformity with the performance guarantee 114. If the server 302 typically performs the task 110 in 0.7 milliseconds, the server 302 may rate limit the task 110 for up to or close to an additional 1.8 milliseconds to reduce the rate at which the workload 104 is delivered to the downstream server 108. If further rate-limiting is to be applied, instead of introducing a new risk 204 of failing the performance guarantee 114, the server 302 may refrain from further rate-limiting the task 110. Instead, as shown in the fourth example scenario 716 of
E4. Process Path Adaptation
A fourth aspect that may vary among embodiments of the techniques presented herein involves adjustment of the process path 112 to adapt to a processing capacity shortage 116. In some scenarios, rate-limiting the tasks 110 of upstream servers 108 may be adequate to resolve a processing capacity shortage 116. However, in other scenarios, the processing capacity shortage 116 may be severe, prolonged, and/or of indefinite duration, such that in addition to rate-limiting a task 110 of an upstream server 108, the server set 106 may implement more significant steps to maintain the satisfaction of the performance guarantee 114.
As a first variation of this fourth aspect, illustrated in a first example scenario 808, a server 302 may respond to a performance capability alert 206 of a downstream server 108 by redirecting the process path 112 through the server set 102 to provide a substitute server 802 in lieu of the server 108 exhibiting the performance capacity shortage 116. The substitute server 802 may be previously allocated and allocated and ready for designation as a substitute server 802, or may be newly provisioned for 802 and inserted into the process path 112. Alternatively, the substitute server 802 may already exist in the process path 112 of the workload 104 or in another process path 112 of the server set 104, and the task 110 performed by the server 108 may be transferred to the substitute server 802 along with the workload 104.
As a second variation of this fourth aspect, a server 302 may respond to a performance capability alert 206 by expanding a computational resource set of the server 108 exhibiting the processing capacity shortage 116. As a first such example, the server 108 may comprise a virtual machine, and the processing resources allocated to the virtual machine may be increased (e.g., raising a thread priority and/or processing core usage of the virtual machine). As a second such example, illustrated in a second example scenario 810, the server 108 exhibiting the processing capacity shortage 116 may be supplemented by the addition of an auxiliary server 804 that expands the processing capacity of the server 108. For example, the workload 104 may be shared between the server 108 and the auxiliary server 804 until the processing capacity shortage 116 of the server 108 is alleviated.
As a third variation of this fourth aspect, a server 108 exhibiting a processing capacity shortage 116 may experience a processing capacity shortage 116 that risks failing a performance guarantee 114 of a first workload 104, but that presents lower or no risk of failing performance guarantees 114 of other workloads 104 of the server 108. The server 108 may therefor prioritize the processing of the first workload 104 over the other workloads 104 alleviate the processing capacity shortage 106. As a first such example, illustrated in a third example scenario 812, the server 108 may adjust by reducing a process priority 806 of another workload 104 that the server 108 processes, e.g., a workload 104 that involves no performance guarantee 114, or may involve a second performance guarantee 114 that is amply satisfied (e.g., a dependency upon a different type of performance capability of the server 108, such as a CPU-bound workload as compared with a network-bound workload). The relative adjustment of the process priorities 806 may enable the server 108 to work through a backlog and resolve the processing capacity shortage 116. As a second such example, where the server 108 processes a third task 110 for a third workload 104 according to a third process path 112, the server 108 may redirect the second process path for the third workload 110 through a substitute server 802. The server 108 may therefore reserve a greater proportion of computational resources to address the processing capacity shortage 116.
As a fourth variation of this fourth aspect, a server 108 that implements rate-limiting of a task 110 in order to alleviate a processing capacity shortage 116 of a downstream server 108 may curtail or end the rate-limiting of the task 110 based upon an alleviation of the performance capability shortage 116 of the downstream sever 108. As a first such example, a downstream server 108 initiating the performance capability alert 206 may send a notification to an upstream server 302 applying the rate-limiting to indicate an abatement of the processing capacity shortage 116. As a second such example, an upstream server 302 applying rate-limiting to a task 110 may detect an abatement of the processing capacity shortage 116, e.g., as a depletion of an output queue of workloads 104 to deliver to the downstream server 108. As a third such example, the upstream server 302 may apply the rate-limiting only for a set interval, such as one second, and may then remove the rate-limiting, such that a persistence of the processing capacity shortage 116 at the downstream server 108 may result in a second performance capability alert 206 and a reapplication of the rate limit to the process 110. In some scenarios, the reapplication may occur at an increasing interval (e.g., first one second, then two seconds, etc.) to reduce an inefficiency of the transmission and receipt of multiple performance capability alerts 206, which may reduce the ability of the downstream server 108 to alleviate the processing capacity shortage 116.
As a fifth variation of this fourth aspect, the adjustments of the process paths 112 may be requested and/or implemented by the server 108 experiencing the processing capacity shortage 116. As another example, the adjustments of the processing paths 112 may be requested and/or implemented by the upstream server 302, e.g., upon determining that the rate-limiting of the task 110 by the upstream server 302 is insufficient to resolve a processing capacity shortage 116 that is prolonged, indefinite, overly frequent, and/or unresolvable by rate-limiting. As yet another example, the adjustments of the processing paths 112 may be implemented at the request of an automated network monitor or network administrator. Many such techniques may be utilized to provide further adaptations of the server set 106, in conjunction with the rate-limiting of the task 110 by the upstream server 302, in accordance with the techniques presented herein.
Although not required, embodiments are described in the general context of “computer readable instructions” being executed by one or more computing devices. Computer readable instructions may be distributed via computer readable media (discussed below). Computer readable instructions may be implemented as program modules, such as functions, objects, Application Programming Interfaces (APIs), data structures, and the like, that perform particular tasks or implement particular abstract data types. Typically, the functionality of the computer readable instructions may be combined or distributed as desired in various environments.
In other embodiments, device 902 may include additional features and/or functionality. For example, device 902 may also include additional storage (e.g., removable and/or non-removable) including, but not limited to, magnetic storage, optical storage, and the like. Such additional storage is illustrated in
The term “computer readable media” as used herein includes computer storage media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions or other data. Memory 908 and storage 910 are examples of computer storage media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVDs) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by device 902. Any such computer storage media may be part of device 902.
Device 902 may also include communication connection(s) 916 that allows device 902 to communicate with other devices. Communication connection(s) 916 may include, but is not limited to, a modem, a Network Interface Card (NIC), an integrated network interface, a radio frequency transmitter/receiver, an infrared port, a USB connection, or other interfaces for connecting computing device 902 to other computing devices. Communication connection(s) 916 may include a wired connection or a wireless connection. Communication connection(s) 916 may transmit and/or receive communication media.
The term “computer readable media” may include communication media. Communication media typically embodies computer readable instructions or other data in a “modulated data signal” such as a carrier wave or other transport mechanism and includes any information delivery media. The term “modulated data signal” may include a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal.
Device 902 may include input device(s) 914 such as keyboard, mouse, pen, voice input device, touch input device, infrared cameras, video input devices, and/or any other input device. Output device(s) 912 such as one or more displays, speakers, printers, and/or any other output device may also be included in device 902. Input device(s) 914 and output device(s) 912 may be connected to device 902 via a wired connection, wireless connection, or any combination thereof. In one embodiment, an input device or an output device from another computing device may be used as input device(s) 914 or output device(s) 912 for computing device 902.
Components of computing device 902 may be connected by various interconnects, such as a bus. Such interconnects may include a Peripheral Component Interconnect (PCI), such as PCI Express, a Universal Serial Bus (USB), Firewire (IEEE 1394), an optical bus structure, and the like. In another embodiment, components of computing device 902 may be interconnected by a network. For example, memory 908 may be comprised of multiple physical memory units located in different physical locations interconnected by a network.
Those skilled in the art will realize that storage devices utilized to store computer readable instructions may be distributed across a network. For example, a computing device 920 accessible via network 918 may store computer readable instructions to implement one or more embodiments provided herein. Computing device 902 may access computing device 920 and download a part or all of the computer readable instructions for execution. Alternatively, computing device 902 may download pieces of the computer readable instructions, as needed, or some instructions may be executed at computing device 902 and some at computing device 920.
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
As used in this application, the terms “component,” “module,” “system”, “interface”, and the like are generally intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. One or more components may be localized on one computer and/or distributed between two or more computers.
Furthermore, the claimed subject matter may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware, or any combination thereof to control a computer to implement the disclosed subject matter. The term “article of manufacture” as used herein is intended to encompass a computer program accessible from any computer-readable device, carrier, or media. Of course, those skilled in the art will recognize many modifications may be made to this configuration without departing from the scope or spirit of the claimed subject matter.
Various operations of embodiments are provided herein. In one embodiment, one or more of the operations described may constitute computer readable instructions stored on one or more computer readable media, which if executed by a computing device, will cause the computing device to perform the operations described. The order in which some or all of the operations are described should not be construed as to imply that these operations are necessarily order dependent. Alternative ordering will be appreciated by one skilled in the art having the benefit of this description. Further, it will be understood that not all operations are necessarily present in each embodiment provided herein.
Any aspect or design described herein as an “example” is not necessarily to be construed as advantageous over other aspects or designs. Rather, use of the word “example” is intended to present one possible aspect and/or implementation that may pertain to the techniques presented herein. Such examples are not necessary for such techniques or intended to be limiting. Various embodiments of such techniques may include such an example, alone or in combination with other features, and/or may vary and/or omit the illustrated example.
As used in this application, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless specified otherwise, or clear from context, “X employs A or B” is intended to mean any of the natural inclusive permutations. That is, if X employs A; X employs B; or X employs both A and B, then “X employs A or B” is satisfied under any of the foregoing instances. In addition, the articles “a” and “an” as used in this application and the appended claims may generally be construed to mean “one or more” unless specified otherwise or clear from context to be directed to a singular form.
Also, although the disclosure has been shown and described with respect to one or more implementations, equivalent alterations and modifications will occur to others skilled in the art based upon a reading and understanding of this specification and the annexed drawings. The disclosure includes all such modifications and alterations and is limited only by the scope of the following claims. In particular regard to the various functions performed by the above described components (e.g., elements, resources, etc.), the terms used to describe such components are intended to correspond, unless otherwise indicated, to any component which performs the specified function of the described component (e.g., that is functionally equivalent), even though not structurally equivalent to the disclosed structure which performs the function in the herein illustrated example implementations of the disclosure. In addition, while a particular feature of the disclosure may have been disclosed with respect to only one of several implementations, such feature may be combined with one or more other features of the other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms “includes”, “having”, “has”, “with”, or variants thereof are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term “comprising.”
This application is a continuation of, and claims priority under 35 U.S.C. §§ 119-120 to, U.S. Patent Application No. 62/668,226, entitled “DISTRIBUTED DATABASES,” filed on May 7, 2018, the entirety of which is hereby incorporated by reference as if fully rewritten herein.
| Number | Date | Country | |
|---|---|---|---|
| 62668226 | May 2018 | US |
