Dynamically identifying target capacity when scaling cloud resources

Patent metadata

Field	Value
Publication number	US-9722945-B2
Application number	US-201414307745-A
Country	US
Kind code	B2
Filing date	Jun 18, 2014
Priority date	Mar 31, 2014
Publication date	Aug 1, 2017
Grant date	Aug 1, 2017

Abstract

Embodiments are directed to preventing flapping when auto-scaling cloud resources. In one scenario, a computer system accesses information specifying a target operational metric that is to be maintained on a plurality of cloud resources. The computer system determines a current measured value for the target operational metric for at least some of the cloud resources. The computer system further calculates a scaling factor based on the target operational metric and the current measured value, where the scaling factor represents an amount of variance between the target operational metric and the current measured value. The computer system also calculates a delta value representing a modified quantity of cloud resources modified by the calculated scaling factor and determines whether a scaling action is to occur based on the calculated delta value.

First claim

We claim:

1. A computer-implemented method for auto-scaling cloud resources in order to increase or decrease the number of virtual machine (VM) instances used to meet a current computing load being handled by a service, the computer-implemented method being performed by one or more processors executing computer executable instructions for the computer-implemented method, and the computer-implemented the method comprising acts of:
periodically accessing information at given intervals, wherein the accessed information comprises:
a determination of how many VM instances (n) are currently being used to meet a computing load of a service; and
measured performance metrics that quantify the computing load the service is currently handling;

for at least one given measured performance metric, obtaining a per-capita load (PCL) by dividing the given measured performance metric by the number of VM instances being used to meet the computing load of the service;
accessing a user-defined per-capita target threshold (PCTT) representing an amount of the computing load each VM instance is to handle for the given performance metric;
determining a scaling action to be taken, but without flapping the VM instances by causing them to enter an undesirable cycle of alternately removing and then adding the same number of VM instances until the current computing load changes, by performing the following:
calculating a scaling factor based on dividing the PCL by the PCTT for the given performance metric, wherein the scaling factor represents an amount of variance between the at least one given measured performance metric and the user-defined per-capita target threshold (PCTT) representing an amount of the computing load each VM instance is to handle for the given performance metric;
calculating the number of VM instances required to scale to the scaling factor by determining a delta value based on the difference between i) the number of VM instances (n) currently being used to meet the computing load of the service and ii) n times the scaling factor;
determining whether a scaling action is to occur based on the calculated delta value; and

when determined, performing the scaling action as indicated by the calculated delta value.

2. The computer-implemented method of claim 1 , wherein the scaling action comprises adding virtual machines when the delta value is positive or removing virtual machines when the delta value is negative.

3. The computer-implemented method of claim 1 , wherein the scaling action comprises increasing or decreasing the size of at least one virtual machine.

4. The computer-implemented method of claim 1 , wherein a at least one of the performance metrics is prioritized higher than at least one other performance metric.

5. The computer-implemented method of claim 4 , wherein health of the virtual machines is prioritized over cost savings.

6. The computer-implemented method of claim 4 , wherein the scaling factor is calculated for each of the measured performance metrics and wherein a specified scaling factor is selected based on the prioritization.

7. The computer-implemented method of claim 6 , wherein the specified scaling factor comprises a maximum scaling factor that represents the highest amount of variance between the at least one given measured performance metric and the user-defined per-capita target threshold (PCTT).

8. The computer-implemented method of claim 1 , further comprising:
determining that at least one performance metric, if implemented, would cause at least one service level agreement (SLA) to no longer be enforceable; and
generating a notification that the selected operational metric would result in the SLA no longer being enforceable.

9. The computer-implemented method of claim 1 , wherein upon determining that a scaling action is to occur based on the calculated delta value, the computer-implemented method further comprises calculating a projected impact the scaling action would have on the total capacity of the VM instances currently being used to meet the computing load of the service.

10. A computer-implemented method for auto-scaling cloud resources in order to increase or decrease the number of virtual machine (VM) instances used to meet a current computing load being handled by a service, the computer-implemented method being performed by one or more processors executing computer executable instructions for the computer-implemented method, and the computer-implemented the method comprising acts of:
periodically accessing information at given intervals, wherein the accessed information comprises:
a determination of how many VM instances (n) are currently being used to meet a computing load of a service; and
measured performance metrics that quantify the computing load the service is currently handling;

for at least one given measured performance metric, obtaining a per-capita load (PCL) by dividing the given measured performance metric by the number of VM instances being used to meet the computing load of the service;
accessing a user-defined per-capita target threshold (PCTT) representing an amount of the computing load each VM instance is to handle for the given performance metric;
determining a scaling action to be taken, but without flapping the VM instances by causing them to enter an undesirable cycle of alternately removing and then adding the same number of VM instances until the current computing load changes, by performing the following:
calculating a scaling factor based on dividing the PCL by the PCTT for the given performance metric, wherein the scaling factor represents an amount of variance between the at least one given measured performance metric and the user-defined per-capita target threshold (PCTT) representing an amount of the computing load each VM instance is to handle for the given performance metric;
calculating the number of VM instances required to scale to the scaling factor by determining a delta value based on the difference between i) the number of VM instances (n) currently being used to meet the computing load of the service and ii) n times the scaling factor; and
determining whether a scaling action is to occur based on the calculated delta value, and when a scaling action is determined, performing the following:
projecting the impact the scaling action will have on total capacity of the service to meet the current computing load; and
if the projected impact indicates that the scaling action will result in a new total capacity of the service that is unable to handle the current computing load, reducing the magnitude of the scaling action until the new total capacity will be able to serve the current computing load.

11. The computer-implemented method of claim 10 , wherein the scaling action comprises adding virtual machines when the delta value is positive or removing virtual machines when the delta value is negative.

12. The computer-implemented method of claim 10 , wherein the scaling action comprises increasing or decreasing the size of at least one virtual machine.

13. The computer-implemented method of claim 10 , wherein at least one of the performance metrics is prioritized higher than at least one other performance metric.

14. The computer-implemented method of claim 13 , wherein health of the virtual machines is prioritized over cost savings.

15. The computer-implemented method of claim 13 , wherein the scaling factor is calculated for each of the measured performance metrics and wherein a specified scaling factor is selected based on the prioritization.

16. The computer-implemented method of claim 15 , wherein t

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Siciliano Stephen
• Lamanna Charles
• Grebnov Ilya

Classifications

• G06F9/5072 Primary

  Grid computing · CPC title

• H04L43/0817

  by checking functioning · CPC title

• H04L41/0836

  to enhance reliability, e.g. reduce downtime · CPC title

• G06F9/5083

  Techniques for rebalancing the load in a distributed system · CPC title

• H04L47/781

  Centralised allocation of resources · CPC title