Server farm and method for operating the same

Patent metadata

Field	Value
Publication number	US-9959149-B1
Application number	US-201615290106-A
Country	US
Kind code	B1
Filing date	Oct 11, 2016
Priority date	Oct 11, 2016
Publication date	May 1, 2018
Grant date	May 1, 2018

Abstract

A method for operating a server farm with a plurality of servers operably connected with each other includes: receiving a job request of a computational task to be handled by the server farm; determining, from the plurality of servers, one or more servers operable to accept the job request; determining a respective effective energy efficiency value associated with at least the one or more servers; and assigning the computational task to a server with the highest effective energy efficiency value. The effective energy efficiency value is defined by a service rate of the respective server divided by a difference between an energy consumption rate value when the respective server is busy and an energy consumption rate value when the respective server is idle. The present invention also relates to a server farm operated by the method.

First claim

The invention claimed is:

1. A method for operating a server farm with a plurality of servers operably connected with each other, the method comprising the steps of:
receiving a job request of a computational task to be handled by the server farm;
determining, from the plurality of servers, one or more servers operable to accept the job request;
determining a respective effective energy efficiency value associated with at least the one or more servers, the respective effective energy efficiency value of the servers being defined by: a service rate of the respective server divided by a difference between an energy consumption rate value when the respective server is busy and an energy consumption rate value when the respective server is idle; and
assigning the computational task to a server with the highest effective energy efficiency value;

for maximizing energy efficiency of the server farm, defined by a ratio of a long-run expected throughput divided by an expected energy consumption rate value.

2. The method in accordance with claim 1 , further comprising the step of:
sorting the one or more servers according to the respective determined effective energy efficiency values.

3. The method in accordance with claim 1 , wherein the step of determining from the plurality of servers one or more servers operable to accept the job request comprises determining, from the plurality of servers, all servers operable to accept the job request.

4. The method in accordance with claim 1 , wherein the plurality of servers cannot be powered off during operation of the server farm.

5. The method in accordance with claim 1 , wherein assignment of computation tasks in the server farm is substantially independent of an arrival rate of computation tasks at the server farm.

6. The method in accordance with claim 1 , wherein assignment of computation tasks in the server farm is substantially independent of a respective size of the computation tasks received at the server farm.

7. The method in accordance with claim 1 , wherein the plurality of servers each includes a finite buffer for queuing job requests.

8. The method in accordance with claim 7 , wherein the one or more servers operable to accept the job request each has at least one vacancy in their respective buffer.

9. The method in accordance with claim 1 , wherein the server farm is heterogeneous in that the plurality of servers can have different server speeds, energy consumption rates, and/or buffer sizes.

10. The method in accordance with claim 1 , wherein the server farm is a non-jockeying server farm in which computational task being handled by one of the plurality of servers cannot be reassigned to other servers.

11. A server farm comprising:
a plurality of servers operably connected with each other;
one or more processor operably connected with the plurality of server, the one or more processor being arranged to
receive a job request of a computational task to be handled by the server farm;
determine, from the plurality of servers, one or more servers operable to accept the job request;
determine a respective effective energy efficiency value associated with at least the one or more servers, the respective effective energy efficiency value being defined by: a service rate of the respective server divided by a difference between an energy consumption rate value when the respective server is busy and an energy consumption rate value when the respective server is idle; and
assign the computational task to a server with the highest effective energy efficiency value;

for maximizing energy efficiency of the server farm, defined by a ratio of a long-run expected throughput divided by an expected energy consumption rate value.

12. The server farm in accordance with claim 11 , wherein the one or more processor is further operable to:
sort the one or more servers according to the respective determined effective energy efficiency values.

13. The server farm in accordance with claim 11 , wherein the one or more processor is further operable to: determine, from the plurality of servers, all servers operable to accept the job request.

14. The server farm in accordance with claim 11 , wherein the plurality of servers cannot be powered off during operation of the server farm.

15. The server farm in accordance with claim 11 , wherein the one or more processor is arranged such that assignment of computation tasks in the server farm is substantially independent of an arrival rate of computation tasks at the server farm.

16. The server farm in accordance with claim 11 , wherein the one or more processor is arranged such that assignment of computation tasks in the server farm is substantially independent of a respective size of the computation tasks received at the server farm.

17. The server farm in accordance with claim 11 , wherein the plurality of servers each includes a finite buffer for queuing job requests; and wherein the one or more servers operable to accept the job request each has at least one vacancy in their respective buffer.

18. The server farm in accordance with claim 11 , wherein the server farm is heterogeneous in that the plurality of servers can have different server speeds, energy consumption rates, and/or buffer sizes.

19. The server farm in accordance with claim 11 , wherein the server farm is a non-jockeying server farm in which computational task being handled by one of the plurality of servers cannot be reassigned to other servers.

20. The server farm in accordance with claim 11 , wherein the one or more processors are incorporated in at least one of the plurality of servers.

Assignees

• Univ City Hong Kong

Inventors

• Fu Jing
• Morgan William
• Guo Jun
• Zukerman Moshe
• Wong Wing Ming Eric

Classifications

• G06F9/5027

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title

• G06F9/5094 Primary

  where the allocation takes into account power or heat criteria (power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206) · CPC title

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title