Allocating operators of a streaming application to virtual machines based on monitored performance

• US 9,794,139 B2
• Filed: 03/07/2014
• Issued: 10/17/2017
• Est. Priority Date: 03/07/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bragstad, Lance, Branson, Michael J., Cao, Bin, Carey, James E., Odden, Mathew R.
• Primary Examiner(s): Chou, Alan S
• Application Number:
  US14/200,118
• Publication Number:
  US 20150256482A1
• Time in Patent Office: 1,320 Days
• Field of Search: 709219
• CPC Class Codes: G06F 11/32 with visual or acoustical indication of the functioning of the machine
• CPC Class Codes: G06F 11/3404 for parallel or distributed programming
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3442 for planning or managing the needed capacity
• CPC Class Codes: G06F 2009/4557 Distribution of virtual machine instances; Migration and load balancing
• CPC Class Codes: G06F 2201/81 Threshold
• CPC Class Codes: G06F 2201/815 Virtual
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 2209/501 Performance criteria
• CPC Class Codes: G06F 2209/5017 Task decomposition
• CPC Class Codes: G06F 2209/506 Constraint
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/50 Allocation of resources, e.g. of the central processing unit [CPU]
• CPC Class Codes: G06F 9/5038 considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration scheduling strategies G06F9/4881 and subgroups
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06F 9/5066 Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs mappping at compile time, see G06F8/451
• CPC Class Codes: H04L 41/0896 Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities flow or congestion control using dynamic resource allocation, e.g. in-call renegotiation, H04L47/76
• CPC Class Codes: H04L 41/5009 Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
• CPC Class Codes: H04L 41/5025 by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
• CPC Class Codes: H04L 41/5096 wherein the managed service relates to distributed or central networked applications
• CPC Class Codes: H04L 43/04 : Processing captured monitoring data, e.g. for logfile generation
• CPC Class Codes: H04L 47/783 : Distributed allocation of resources, e.g. bandwidth brokers
• CPC Class Codes: H04L 65/60 : Network streaming of media packets

Abstract

Performance thresholds are defined for operators in a flow graph for a streaming application. A streams manager deploys the flow graph to one or more virtual machines (VMs). The performance of each portion of the flow graph on each VM is monitored. A VM is selected. When the performance of the portion of the flow graph in the selected VM does not satisfy the defined performance threshold(s), a determination is made regarding whether the portion of the flow graph is underperforming or overperforming. When the portion of the flow graph is underperforming, the portion of the flow graph is split into multiple portions that are implemented on multiple VMs. When the portion of the flow graph is overperforming, a determination is made of whether a neighbor VM is also overperforming. When a neighbor VM is also overperforming, the two VMs may be coalesced into a single VM.

Claims

1. A computer-implemented method executed by at least one processor for running a streaming application, the method comprising:

deploying a streaming application that comprises a flow graph that includes a plurality of operators that process a plurality of data tuples on at least one virtual machine (VM);

defining a plurality of performance thresholds for the plurality of operators in the flow graph;

monitoring performance of the streaming application;

when performance of a first portion of the streaming application in a selected VM does not satisfy at least one of the plurality of performance thresholds, determining when the first portion of the streaming application in the selected VM is underperforming, and when the first portion of the streaming application in the selected VM is underperforming, splitting the first portion of the streaming application in the selected VM into second and third portions of the streaming application that are deployed to a plurality of VMs; and

continuously monitoring performance of all portions of the flow graph and dynamically changing allocation of portions of the flow graph that are underperforming by splitting at least one portion of the flow graph to different VMs and changing allocation of portions of the flow graph that are overperforming by coalescing at least one portion of the flow graph to a single VM.

2. The method of claim 1 wherein when performance of the first portion of the streaming application in the selected VM does not satisfy the at least one of the plurality of performance thresholds, determining when the first portion of the streaming application in the selected VM is overperforming, and when the first portion of the streaming application in the selected VM is overperforming and when a second VM that has an operator in a fourth portion of the flow graph that communicates with an operator in the first portion of the streaming application in the selected VM is also overperforming, coalescing the first portion and the fourth portion to run on a single VM.

3. The method of claim 1 wherein deploying the streaming application on the at least one VM comprises initially deploying the flow graph on a single VM.

4. The method of claim 1 wherein deploying the streaming application on the at least one VM comprises requesting a cloud manager provision the plurality of VMs, and after the cloud manager provisions the plurality of VMs, the streams manager deploys the flow graph to the plurality of VMs.

5. The method of claim 1 wherein the plurality of performance thresholds includes at least one threshold that indicates when at least one operator in the flow graph is underperforming.

6. The method of claim 1 wherein the plurality of performance thresholds includes at least one threshold that indicates when at least one operator in the flow graph is overperforming.

7. A computer-implemented method executed by at least one processor for running a streaming application, the method comprising:

deploying a streaming application that comprises a flow graph that includes a plurality of operators that process a plurality of data tuples on a single virtual machine (VM) by requesting a cloud manager provision the single VM, and after the cloud manager provisions the single VM, the streams manager deploys the flow graph to the single VM;

defining a plurality of performance thresholds for the plurality of operators in the flow graph, wherein the plurality of performance thresholds includes at least one threshold that indicates when at least one operator in the flow graph is underperforming and at least one threshold that indicates when at least one operator in the flow graph is overperforming;

monitoring performance of the streaming application;

when performance of the streaming application in the single VM does not satisfy at least one of the plurality of performance thresholds, determining when the streaming application in the single VM is underperforming, and when the streaming application in the single VM is underperforming, splitting the streaming application in the single VM into first and second portions of the streaming application that are deployed to first and second VMs by requesting the cloud manager provision the first and second VMs;

when performance of the first portion of the streaming application in the first VM does not satisfy at least one of the plurality of performance thresholds, determining when the first portion of the streaming application in the first VM is underperforming, and when the first portion of the streaming application in the first VM is underperforming, splitting the first portion of the streaming application in the first VM into third and fourth portions of the streaming application that are deployed to respective third and fourth VMs by requesting the cloud manager provision the third and fourth VMs;

when performance of the second portion of the streaming application in the second VM does not satisfy at least one of the plurality of performance thresholds, determining when the second portion of the streaming application in the second VM is underperforming, and when the second portion of the streaming application in the second VM is underperforming, splitting the second portion of the streaming application in the second VM into fifth and sixth portions of the streaming application that are deployed to respective fifth and sixth VMs;

when performance of the first portion of the streaming application in the first VM does not satisfy the at least one of the plurality of performance thresholds, determining when the first portion of the streaming application in the first VM is overperforming, and when the first portion of the streaming application in the first VM is overperforming and when a different VM that has an operator in a seventh portion of the flow graph that communicates with an operator in the first portion of the streaming application in the selected VM is also overperforming, coalescing the first portion and the seventh portion to run on a single VM; and

continuously monitoring performance of all portions of the flow graph and dynamically changing allocation of portions of the flow graph that are underperforming by splitting at least one portion of the flow graph to different VMs and changing allocation of portions of the flow graph that are overperforming by coalescing at least one portion of the flow graph to a single VM.