Load management to reduce communication signaling latency in a virtual machine environment

• US 7,730,269 B2
• Filed: 08/29/2006
• Issued: 06/01/2010
• Est. Priority Date: 08/29/2006
• Status: Active Grant

• Current Assignee: X Corp. (f/k/a Twitter, Inc.) (X Holdings Corp.)
• Original Assignee: International Business Machines Corporation
• Inventors: Martin, Brian Keith, Hrischuk, Curtis E., Burckart, Erik John
• Primary Examiner(s): Elmore; Stephen C
• Assistant Examiner(s): Giardino, Jr.; Mark A
• Application Number:
  US11/468,141
• Publication Number:
  US 20080059747A1
• Time in Patent Office: 1,372 Days
• Field of Search: 707/206, 711/170, 711/154
• US Class Current: 711/154
• CPC Class Codes: G06F 12/0253 Garbage collection, i.e. reclamation of unreferenced memory

Abstract

A computer implemented method for reducing communication signaling protocol latency. An acceptable level of latency is specified. Automatic memory management activities are monitored based on specified parameters to calculate a level of activity that determines whether a reduction of activity is required.

Claims

1. A computer implemented method of optimizing an internet signaling protocol latency of an application server, the computer implemented method comprising:

receiving a plurality of application memory requests at a load balancer associated with the application server;

calculating, at the load balancer, an automatic memory management latency based on an automatic memory management activity in response to receiving each of the application memory requests;

evaluating, at the load balancer, based on the calculated automatic memory management latency, whether the automatic memory management latency is within a specified threshold that is acceptable for the internet signaling protocol latency of the application server, wherein the internet signaling protocol latency is based on the automatic memory management latency; and

throttling an application memory request acceptance rate of the application memory requests at the load balancer based on a result of the evaluation which indicates the automatic memory management latency exceeds the specified threshold.

2. The computer implemented method of claim 1, wherein the calculated latency is based on a heap that is scanned upon an occurrence of a garbage collection event.

3. The computer implemented method of claim 1, wherein the calculated latency is based on a rate of object reclamation by a garbage collection event.

4. The computer implemented method of claim 1, wherein the internet signaling protocol comprises a session initiation protocol (SIP).

5. The computer implemented method of claim 1, wherein calculating the automatic memory management latency comprises:

developing a first function that inputs a rate of consumption of a heap by input requests, a rate at which the heap is being reclaimed by the automatic memory management activity, a hardware platform characteristic, and a performance measurement of a hardware server on the platform; and

developing a second function that inputs a rate at which objects are created by input requests, a rate at which objects are being reclaimed by a garbage collection thread, the hardware platform characteristic, and the performance measurement of the hardware server on the platform.

6. The computer implemented method of claim 5, wherein the evaluation comprises determining whether a result of the second function is less than an acceptable internet signaling protocol latency based on a slowness of object reclamation.

7. The computer implemented method of claim 6, wherein the evaluation comprises determining whether a result of the first function is less than an acceptable internet signaling protocol latency based on a slowness of heap reclamation.

8. The computer implemented method of claim 1, wherein throttling an application memory request acceptance rate comprises rejecting memory request.

9. The computer implemented method of claim 1, further comprising predicting a future automatic memory management activity based on the evaluation.

10. The computer implemented method of claim 6, wherein receiving the plurality of application memory requests comprises receiving at one application server, and responsive to determining that the second function is less than an acceptable internet signaling protocol latency based on a slowness of object reclamation and the first function is less than an acceptable internet signaling protocol latency based on a slowness of heap reclamation, sending the memory request to another application server.

11. The computer implemented method of claim 1, wherein the automatic memory management activity comprises a garbage collection activity.

12. The computer implemented method of claim 11, wherein throttling the application memory request acceptance rate further comprises determining if an average garbage collection activity time is above a specified threshold.

13. A method for accepting application memory requests for an application server at a load balancer associated with the application server, the application memory requests consuming heap memory associated with the application server, the method comprising:

receiving an application memory request at the load balancer;

determining a first function related to an expected garbage collection latency based on an amount of heap that needs to be scanned during a garbage collection event;

determining a second function related to an expected garbage collection latency based on a number of objects in the heap that are not collected during the garbage collection event; and

wherein determining the first and second functions are performed at the load balancer;

rejecting the application memory request at the load balancer when the first and second functions exceed respective first and second thresholds; and

accepting the application memory request at the load balancer when the first and second functions are under the first and second thresholds, respectively.

14. The method of claim 13, wherein:

determining the first function inputs a rate of consumption of the heap by input requests, a rate at which the heap is being reclaimed by the automatic memory management activity, a hardware platform characteristic, and a performance measurement of a hardware server on the platform; and

determining the second function inputs a rate at which objects are created by input requests, a rate at which objects are being reclaimed by a garbage collection thread, the hardware platform characteristic, and the performance measurement of the hardware server on the platform.

15. The method of claim 14 wherein determining the second function comprises determining whether a result of the second function is less than an acceptable internet signaling protocol latency based on a slowness of object reclamation.

16. The method of claim 15, wherein determining the first function comprises determining whether a result of the first function is less than an acceptable internet signaling protocol latency based on a slowness of heap reclamation.