Performance diagnostic for virtual machines

• US 10,339,683 B2
• Filed: 09/23/2015
• Issued: 07/02/2019
• Est. Priority Date: 09/23/2015
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Guo, Shugang, Lu, Zhongping, Feng, Jin, Zhang, Yixiang
• Primary Examiner(s): Tran, Dzung
• Application Number:
  US14/862,163
• Publication Number:
  US 20170084062A1
• Time in Patent Office: 1,378 Days
• Field of Search: 702179, 705 35, 705 39, 705 38
• CPC Class Codes: G06F 11/3452 Performance evaluation by statistical analysis
• CPC Class Codes: G06F 11/3476 Data logging G06F11/14, G06F11/2205 take precedence
• CPC Class Codes: G06F 17/40 Data acquisition and logging for input to computer G06F3/00
• CPC Class Codes: G06F 2201/815 Virtual
• CPC Class Codes: G06F 40/174 Form filling; Merging
• CPC Class Codes: G06F 7/08 Sorting, i.e. grouping record carriers in numerical or other ordered sequence according to the classification of at least some of the information they carry by merging two or more sets of carriers in ordered sequence G06F7/16
• CPC Class Codes: G06T 11/206 Drawing of charts or graphs

Abstract

A method for a processor to provide a chart of a performance metric in a collection interval includes creating regions by dividing the collection interval into regions of increasingly smaller time intervals and determining a mean and a variance for each region based on data points in that region, sorting the regions by their variances and means, and processing the sorted regions. Processing the sorted regions includes removing any child region when its parent region has a variance that substantially represents the child region, and replacing any two neighboring or intersecting regions with a merged region comprising the two neighboring or intersecting regions when the merged region has a variance that substantially represents the two neighboring or intersecting regions. The method further includes generating the chart by visually indicating highest ranking regions by variance in the chart and displaying the chart or transmitting the chart over a computer network.

Claims

1. A virtual machine system with improved performance diagnostic for virtual machines, comprising:

networked host computers running the virtual machines, the virtual machines running applications;

a virtual machine manager to manage the virtual machines, the virtual machine manager comprising;

a performance analyzer to provide a plurality of highest ranking regions in a regions list comprising data points of a performance metric for a virtual machine, the performance analyzer comprising;

a region abstractor to create regions of various time interval in the regions list, each region being a parent, a child, or a neighbor to a number of other regions in the region list;

a region sorter to sort the regions in the regions list by variance and mean;

a region pruner, comprising;

a child-parent region pruner to remove any child region from the regions list when its parent region has a variance that substantially represents the child region; and

a neighbor region merger to process the regions list after the child-parent region pruner, wherein the neighbor region merger is to replace any two neighboring regions in the regions list with a merged region comprising the two neighboring regions when the merged region has a variance that substantially represents the two neighboring regions;

a statistics subsystem to;

collect the data points from the host computers and provide them to the performance analyzer;

generate a chart of the data points;

receive the plurality of highest ranking regions in the regions list from the performance analyzer and visually indicate them in the chart;

display the chart or transmit it over a computer network; and

display recorded events and recorded alarms corresponding to the plurality of highest ranking regions in the regions list or transmit them over the computer network; and

an events and alarms subsystem to record events and alarms for the virtual machine and provide the recorded events and the recorded alarms corresponding to the plurality of highest ranking regions in the regions list to the statistics subsystem.

2. The system of claim 1, wherein the region abstractor creates the regions by:

creating a first region with its mean and variance based on a data list and adding the first region to the regions list;

splitting the data list into left and right data lists;

when the left data list is not less than a first minimum time interval, creating a second region with its mean and variance based on the left data list and adding the second region to the regions list;

when the left data list is less than the first minimum time interval and the right data list is not less than a second minimum time interval, creating a third region with its mean and variance based on the right data list and adding the third region to the regions list; and

when the left data list is less than the first minimum time interval, the right data list is less than the second minimum time interval, but another right data list at a higher level is not less than the second minimum time interval, creating a fourth region with its mean and variance based on the other right data list and adding the fourth region to the regions list.

3. The system of claim 1, wherein the region sorter sorts the regions in the regions list first by their variance and then any regions with a same variance the region sorter sorts them by their means.

4. The system of claim 1, wherein the child-parent pruner is to:

when a variance of the child region is smaller than the parent region, remove the child region; and

when the variance of the parent region is smaller than the child region, remove the child region when a ratio between the variance of the parent region and the variance of the child region is greater than a similarity factor, or otherwise removing the parent region.

5. The system of claim 1, wherein the neighbor region merger is to:

determine a ratio of the variance of the merged region to a higher of two variances of the two neighboring regions;

when the ratio is greater than a merge factor;

set the variance of the merged region equal to the higher variance;

remove one of the two neighboring regions having the higher variance; and

replace the other of the two neighboring regions with the merged region.

6. The system of claim 1, wherein the region pruner further comprises an intersection region merger to process the regions list after the neighbor region merger, the intersection region merger is to replace any two intersecting regions with a merged region comprising the two intersecting regions when the merged region has a variance that substantially represents the two intersecting regions.

7. The system of claim 6, wherein the intersection region merger is to:

determine a ratio of the variance of the merged region to a higher of two variances of the two intersecting regions;

when the ratio is greater than a merge factor;

set the variance of the merged region equal to the higher variance;

remove one of the two intersecting regions having the higher variance; and

replace the other of the two intersecting regions with the merged region; and

when the ratio is not greater than a threshold, removing the other of the two interesting regions.

8. The system of claim 6, wherein the region pruner further comprises an intersection region pruner to process the regions list after the intersection region merger, the intersection region pruner is to remove any region that has a variance less than another region when the region is a child or parent of the other region or the region intersects the other region.

9. The system of claim 8, wherein the region pruner further comprises a region shrinker to process the regions list after the intersection region pruner, the region shrinker is to remove data points at ends of any region when a new slope fitted to remaining data points in the region without the removed data points is greater than or equal to an old slope fitted to original data points in the region including the removed points.

10. The system of claim 9, wherein the region pruner further comprises a region feeder to process the regions list after the region shrinker, the region feeder is to add one or more points to start or end of any region when a ratio of a new variance of the region with the added data points to an old variance of the region without the added points is greater than a merge factor.