DISTRIBUTED DATA PROCESSING SYSTEM

• US 20110167112A1
• Filed: 08/24/2009
• Published: 07/07/2011
• Est. Priority Date: 09/08/2008
• Status: Active Grant

• Current Assignee: British Telecommunications PLC (BT Group PLC)
• Original Assignee: British Telecommunications PLC (BT Group PLC)
• Inventors: Fisher, Michael, McKee, Paul, Mazzucco, Michele
• Granted Patent: US 9,015,227 B2
• Time in Patent Office: Days
• US Class Current: 709/203
• CPC Class Codes: G06F 2209/5019 Workload prediction
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/1023 based on a hash applied to IP addresses or costs

Abstract

A distributed data processing system is disclosed. A problem with distributed data processing systems, especially data centres, is the low level of utilisation of those data centres. This is, in part, owing to data centre providers being concerned not to attract financial penalties levied by their customers in response to the quality of service provided to those customers by the data centre falling below service level agreed in service level agreements. Embodiments are disclosed which tackle this problem by having the data centre control the load placed upon it by carrying out an assessment of the likely impact of performing a requested bundle of jobs on jobs which are already running in the data centre. The decision whether to provide the requested bundle of jobs is then made by weighing-up the expected increase in penalties payable to customers against the revenue expected in return for running the requested bundle of jobs.

Claims

1. A computer network comprising a client computer, a service provisioning system comprising a job admission controller and one or more server computers, and

communication links connecting said client computer and service provisioning system;

said client computer being arranged in operation to;

i) generate a job bundle request message representing a request for said one or more server computers to carry out a plurality of instances of a data processing job;

ii) send said job bundle request message to said one or more server computers;

said job admission controller being arranged in operation to;

i) receive said job bundle request message;

ii) estimate whether admitting said bundle of jobs would overload said service provisioning system;

iii) determine whether said service provisioning system is to provide said bundle of jobs in accordance with said estimation;

iv) indicate said determination to said client computer;

said client computer further being arranged in operation to;

v) on finding said indication to indicate that said bundle is admitted, send job requests belonging to said bundle to said service provisioning system;

said one or more server computers being arranged in operation to receive job requests in said bundle and to perform the jobs requested therein.

2. A computer network according to claim 1 wherein said job bundle request message includes a measure indicative of the predicted average arrival rate of jobs in said bundle.

3. A computer network according to claim 1 wherein said job admission controller is further arranged in operation to monitor a measure indicative of the current average arrival rate of jobs of a given type and the current average service time of jobs of a given type and to derive therefrom an estimate of the current load offered on the service provisioning system by jobs of said given type.

4. A computer network according to claim 3 wherein said estimation involves said job admission controller calculating the probability that jobs of said type will be performed with a quality worse than a threshold performance level in determining whether to perform said requested bundle of jobs.

5. A computer network according to claim 4 wherein said one or more server computers are further arranged in operation to monitor the variation in inter-arrival time of jobs of a given type and the variation in the service time of jobs of a given type, said variations also being taken into account in calculating the probability that said requested bundle of jobs will be performed with a quality worse than a threshold performance level.

6. A computer network according to claim 1 wherein said one or more server computers perform a plurality of different job types, and said determination involves calculating an allocation of server computers to job types assuming that the requested bundle is admitted, and thereafter calculating the probability that the requested bundle of jobs will be performed with a quality worse than a threshold performance level.

7. A computer network according to claim 4, wherein said job admission controller is in communication with a store storing a financial penalty associated with performing a requested bundle of jobs with a quality worse than a threshold performance level, and a financial reward associated with performing a requested bundle of jobs with a quality better than said threshold performance level, said job admission controller determining whether to perform said bundle of jobs in dependence on the expected revenue from said bundle given said stored penalty and reward and the probability of said quality exceeding said threshold and/or the probability of said quality falling below said threshold.

8. A computer network according to claim 6 wherein said job admission controller is in communication with a store storing, for each type of job, a threshold performance level, a financial penalty associated with performing a requested bundle of jobs with a quality worse than said threshold performance level, and a financial reward associated with performing a requested bundle of jobs with a quality better than said threshold performance level, said job admission controller determining whether to perform said bundle of jobs in dependence on the expected revenue from said bundle given said stored penalty and reward and the probability of said quality exceeding said threshold and/or the probability of said quality falling below said threshold, and on the expected revenue from bundles of jobs which said one or more server computers have already started performing.

9. A computer network according to claim 1 wherein said service provisioning system is arranged in operation to respond to said job request by finding whether said job request belongs to an admitted job bundle, and to meet said job request on finding that said service request does belong to an admitted bundle.

10. A computer network according to claim 1 wherein said job admission controller indicates the result of said determination by:

a) generating a bundle admission indication message indicating whether said service provisioning system is to provide said bundle of jobs; and

iv) sending said job bundle admission message to the sender of said service bundle request message;

said client computer learning said determination by reading said bundle admission message.

11. A computer network according to claim 10 wherein said bundle admission indication message includes a job bundle identifier.

12. A data processing system arranged in operation to:

i) receive a job bundle request message representing a request for said data processing system to carry out a plurality of instances of a data processing service;

ii) determine whether said data processing system is to provide said bundle of jobs; and

iii) indicate the result of said determination to the sender of said job bundle request message.

13. A method of operating a data processing system comprising:

i) receiving a job bundle request message representing a request for said data processing system to carry out a plurality of instances of a data processing service;

ii) determining whether said data processing system is to provide said bundle of jobs; and

iii) indicating the result of said determination to the sender of said job bundle request message.

14. A computer program executable to perform the method of claim 13.

15. A computer-readable medium with the computer program of claim 14 stored thereon.