Machine learning prediction of virtual computing instance transfer performance

Patent metadata

Field	Value
Publication number	US-10853116-B2
Application number	US-201816040272-A
Country	US
Kind code	B2
Filing date	Jul 19, 2018
Priority date	Jul 19, 2018
Publication date	Dec 1, 2020
Grant date	Dec 1, 2020

Abstract

The disclosure provides an approach for preventing the failure of virtual computing instance transfers across data centers. In one embodiment, a flow control module collects performance information primarily from components in a local site, as opposed to components in a remote site, during the transfer of a virtual machine (VM) from the local site to the remote site. The performance information that is collected may include various performance metrics, each of which is considered a feature. The flow control module performs feature preparation by normalizing feature data and imputing missing feature data, if any. The flow control module then inputs the prepared feature data into machine learning model(s) which have been trained to predict whether a VM transfer will succeed or fail, given the input feature data. If the prediction is that the VM transfer will fail, then remediation actions may be taken, such as slowing down the VM transfer.

First claim

We claim:

1. A computer-implemented method, comprising:
receiving first performance information associated with a first data transfer path and a plurality of other data transfer paths;
predicting, using a first machine learning model and the first performance information, a respective success rate of transferring a virtual computing instance from a first data center to a second data center over each of the first data transfer path and the other data transfer paths;
transferring the virtual computing instance from the first data center to the second data center over the first data transfer path responsive to determining that the respective success rate of transferring the virtual computing instance over the first data transfer path is higher than the success rate of transferring the virtual computing instance over the other data transfer paths;
receiving second performance information associated with the transfer of the virtual computing instance from the first data center to the second data center over the first data transfer path;
predicting, using a second machine learning model and the received second performance information, whether the transfer of the virtual computing instance over the first data transfer path will succeed; and
responsive to predicting that the transfer of the virtual computing instance will not succeed, reducing a rate at which the virtual computing instance is transferred.

2. The computer-implemented method of claim 1 , further comprising, selecting the first second machine learning model from a plurality of trained machine learning models based, at least in part, on the received second performance information.

3. The computer-implemented method of claim 1 , wherein the second machine learning model is trained using training data that includes performance and success information associated with at least one of a high throughput, high latency transfer of a virtual computing instance from a third data center to a fourth data center or a low latency, low throughput transfer of a virtual computing instance from the third data center to the fourth data center.

4. The computer-implemented method of claim 1 , further comprising, re-training the second machine learning model using at least the received second performance information.

5. The computer-implemented method of claim 1 , wherein the received second performance information includes at least one of a disk input/output read rate, a network latency, a network throughput, a network packet loss, a compression ratio, a compression throughput, a data insertion rate, or a data delivery rate.

6. The computer-implemented method of claim 1 , wherein the second machine learning model is one of a support vector machine, a logistic regression model, a neural network, or a naive Bayesian network.

7. The computer-implemented method of claim 1 , wherein the second machine learning model outputs one of a label indicating whether the transfer of the virtual computing instance will succeed or fail or a predicted probability that the transfer of the virtual computing instance will succeed.

8. The computer-implemented method of claim 1 , further comprising, responsive to predicting that the transfer of the virtual computing instance will succeed, maintaining the rate at which the virtual computing instance is transferred.

9. The computer-implemented method of claim 1 , wherein the virtual computing instance is a virtual machine.

10. The method of claim 1 , wherein the first data transfer path comprises a first gateway in the first data center and a second gateway in the second data center, and wherein the second performance information comprises data received from the first gateway.

11. A non-transitory computer-readable storage medium comprising instructions, which when executed by a computing system, causes the computing system to carry out operations comprising:
receiving first performance information associated with a first data transfer path and a plurality of other data transfer paths;
predicting, using a first machine learning model and the first performance information, a respective success rate of transferring a virtual computing instance from a first data center to a second data center over each of the first data transfer path and the other data transfer paths;
transferring the virtual computing instance from the first data center to the second data center over the first data transfer path responsive to determining that the respective success rate of transferring the virtual computing instance over the first data transfer path is higher than the success rate of transferring the virtual computing instance over the other data transfer paths;
receiving second performance information associated with the transfer of the virtual computing instance from the first data center to the second data center over the first data transfer path;
predicting, using a second machine learning model and the received second performance information, whether the transfer of the virtual computing instance over the first data transfer path will succeed; and
responsive to predicting that the transfer of the virtual computing instance will not succeed, reducing a rate at which the virtual computing instance is transferred.

12. The computer-readable storage medium of claim 11 , the operations further comprising, selecting the second machine learning model from a plurality of trained machine learning models based, at least in part, on the received second performance information.

13. The computer-readable storage medium of claim 11 , wherein the second machine learning model is trained using training data that includes performance and success information associated with at least one of a high throughput, high latency transfer of a virtual computing instance from a third data center to a fourth data center or a low latency, low throughput transfer of a virtual computing instance from the third data center to the fourth data center.

14. The computer-readable storage medium of claim 11 , the operations further comprising, re-training the second machine learning model using at least the received second performance information.

15. The computer-readable storage medium of claim 11 , wherein the received second performance information includes at least one of a disk input/output read rate, a network latency, a network throughput, a network packet loss, a compression ratio, a compression throughput, a data insertion rate, or a data delivery rate.

16. The computer-readable storage medium of claim 11 , wherein the second machine learning model is one of a support vector machine, a logistic regression model, a neural network, or a naive Bayesian network.

17. The computer-readable storage medium of claim 11 , wherein the second machine learning model outputs one of a label indicating whether the transfer of the virtual computing instance will succeed or fail or a predicted probability that the transfer of the virtual computing instance will succeed.

18. The computer-readable storage medium of claim 11 , the operations further comprising, responsive to predicting that the transfer of the virtual computing instance will succeed, maintaining the rate at which the virtual computing instance is transferred.

19. A system, comprising:
a memory; and
a processor storing one or more applications, which, when executed on the processor, perform operations comprising:
receiving first performance information associated with a first data transfer path and a plurality of other data transfer paths;
predicting, using a first machine

Assignees

• Vmware Inc

Inventors

• Cui Liang
• Ekbote Siddharth
• Wu Weiqing
• Sabin Todd

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F18/2411

  based on the proximity to a decision surface, e.g. support vector machines · CPC title

• G06F11/3419

  by assessing time · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F11/3452

  Performance evaluation by statistical analysis · CPC title