Apparatus and method that empirically measures capacity of multiple servers and forwards relative weights to load balancer

• US 6,571,288 B1
• Filed: 04/26/1999
• Issued: 05/27/2003
• Est. Priority Date: 04/26/1999
• Status: Expired due to Term

• Current Assignee: QNAP Systems Incorporated
• Original Assignee: Hewlett-Packard Company (HP Inc.)
• Inventors: Sarukkai, Sekhar R.
• Primary Examiner(s): Jaroenchonwanit, Bunjob
• Application Number:
  US09/299,323
• Time in Patent Office: 1,492 Days
• Field of Search: 709/224, 709/223, 709/226, 370/252, 714/46
• US Class Current: 709/226
• CPC Class Codes: G06F 2209/508 Monitor
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/10015 Access to distributed or replicated servers, e.g. using brokers
• CPC Class Codes: H04L 67/1029 using data related to the state of servers by a load balancer
• CPC Class Codes: H04L 67/1031 Controlling of the operation of servers by a load balancer, e.g. adding or removing servers that serve requests

Abstract

A system for and a method of empirically measuring the capacity of multiple servers in a cluster including at least one capacity prober for measuring the load capacity of each server and for forwarding the relative weights to a load balancer for the cluster. During off peak operating hours, one at a time each server in the cluster is taken off line and stress tested to measure the capacity of the server. The remaining servers in the cluster remain on line to service customers. The relative weights for each of the servers are collected and updated in the load balancer. In this way, the operation of the cluster is better optimized.

Claims

1. An apparatus for measuring the capacity of at least one server comprising:

a plurality of servers in a cluster;

a network;

a load balancer; and

a plurality of capacity probers;

wherein each of the plurality of capacity probers being associated with one of the plurality of servers in the cluster, and being configured to perform a capacity test on its associated server, such that during the capacity test, the associated server of the plurality of servers is rendered off-line;

wherein the capacity test involves running a predetermined amount of workload sets on the selected server;

wherein each of the plurality of the capacity probers communicates a predetermined parameter of the selected server to a load balancer, such that the predetermined parameter relates to the capacity test;

wherein the predetermined parameter is set to zero, and returns the predetermined parameter back to an original value after the test is performed, thereby returning the selected server to on-line status; and

wherein the capacity prober changes the value of the predetermined parameters for all the servers in the cluster to a new set of predetermined parameters in the load balancer, after the capacity test has been preformed on the last server in the cluster, such that the new set of predetermined parameters indicate the results of the capacity test.

2. The apparatus according to claim 1, wherein the predetermined parameters are the relative weight of the selected server.

3. The apparatus according to claim 1, wherein the workload sets are derived from actual requests over a predetermined time period.

4. A method of empirically measuring the capacity of a plurality of servers in a cluster, the method comprising the steps of:

selecting one of the plurality of servers;

rendering the selected server off-line;

setting a variable parameter for the selected server to zero in a load balancer;

stressing the selected server with a predetermined amount of workload sets;

measuring the performance of the selected server;

resetting the predetermined parameter for the selected server to a previous value in the load balancer; and

rendering the selected server in service.

5. The method according to claim 4, wherein the step of stressing the selected server comprises the steps of:

deriving the predetermined amount of workload sets from a log of actual requests; and

running the workload sets on the selected server.

6. The method according to claim 5, wherein the step of stressing the selected server further comprises the step of:

measuring the predetermined parameter for the selected server based the running of the workload set.

7. The method according to claim 4, wherein the step of rendering the selected server in service further comprises the step of:

setting the measured predetermined parameter value for the selected server to a new value in the load balancer.

8. A system for empirically measuring the capacity of multiple servers in a cluster having a plurality of servers, a network and a load balancer, the method comprising the steps of:

means for selecting one of the plurality of servers;

means for rendering the selected server off-line;

means for setting a variable parameter for the selected server to zero in a load balancer;

means for stressing the selected server with a predetermined amount of work load sets;

means for measuring the performance of the selected server;

means for resetting the predetermined parameter for the selected server to a previous value in the load balancer; and

means for rendering the selected server in service.

9. The method according to claim 8, wherein the means for rendering the selected server in service further comprises the step of:

means for setting the measured predetermined parameter value for the selected server to a new value in the load balancer.

10. The method according to claim 8, wherein the means for stressing the selected server comprises the step of:

means for deriving a workload set from a log of actual requests; and

means for running the workload set on the selected server.

11. The method according to claim 10, wherein the means for stressing the selected server further comprises the step of:

means for measuring the predetermined parameter for the selected server based the running of the workload set.