System for power savings in server farms

• US 7,756,972 B2
• Filed: 12/06/2005
• Issued: 07/13/2010
• Est. Priority Date: 12/06/2005
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Ma, Kevin
• Primary Examiner(s): Follansbee; John
• Assistant Examiner(s): Chou; Alan S
• Application Number:
  US11/295,413
• Publication Number:
  US 20070130341A1
• Time in Patent Office: 1,680 Days
• Field of Search: 709/224, 709/226, 709/229
• US Class Current: 709/224
• CPC Class Codes: G06F 1/26 Power supply means, e.g. regulation thereof for memories G11C
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 2209/5011 Pool
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources
• CPC Class Codes: G06F 9/5094 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 Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/1008 based on parameters of servers, e.g. available memory or workload monitoring of computer activity G06F11/30
• CPC Class Codes: H04L 67/1012 based on compliance of requirements or conditions with available server resources
• CPC Class Codes: H04L 67/1017 based on a round robin mechanism
• CPC Class Codes: H04L 67/1023 based on a hash applied to IP addresses or costs
• 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
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A threshold-based approach is used to assign tasks to servers in a server farm. A “number of connections” count or “connection rate” is maintained for an active server and when the count achieves a first value a wake up signal is sent to a next (inactive) server to alert the inactive server to become active and periodic keep alive signals are initiated. When the number of connections next reaches a second, higher value, subsequent requests for connections are sent to the newly active server. Variations of this approach allow for servers and their associated processes and devices to be handled in clusters for power management scaling efficiency. Servers can be placed in the inactive mode when the number of connections reaches a lower threshold. Other criteria can be used to determine activation of servers, processors or other devices or processes. Other types of resource use monitoring can also be used with this invention.

Claims

1. A method for controlling power consumption of a server, the method comprising:

setting a limit of resource use for a first active server;

measuring the resource use of the first active server;

receiving a request for additional resource use;

when the measured resource use of the first active server plus the request for additional resource are in a predetermined relationship with a pre-active threshold of a first connection rate, then

initiating powering up of a inactive server into an active state in response to the measured resource use of the first active server achieving the pre-active threshold, the first connection rate of the pre-active threshold being less than a second connection rate of an active threshold , wherein said initiating powering up of the inactive server includes;

sending a power up signal to the inactive server to initiate powering up of the inactive server before the second connection rate of the active threshold is reached; and

upon reaching the pre-active threshold, initiating periodic keep alives to the inactive server to keep the inactive server powered up after the pre-active threshold is met and before the inactive server is activated to be assigned requests for additional resource use, wherein the periodic keep alives do not place the inactive server into an active state;

placing the inactive server into the active state as a second active server when the active threshold is met; and

assigning requests for additional resource use to the second active server.

2. The method of claim 1, wherein a resource use includes a network connection request.

3. The method of claim 2, wherein the network connection request includes a request for web page content.

4. The method of claim 2, wherein the network connection request includes a request for data from a database.

5. The method of claim 1, wherein the resource use includes a connection rate, wherein measuring the resource use includes measuring a current connection rate, the method further comprising:

determining connection rate of the pre-active threshold for the first server; and

detecting when incoming requests for additional connections would exceed the pre-active threshold.

6. The method of claim 1, further comprising:

setting an inactive threshold that is of a third connection rate lower than the pre-active threshold, wherein the second active server is transitioned to standby upon reaching the inactive threshold;

measuring a resource use of the second active server; and

placing the second active server into the standby state when the measured resource use of the second active server achieves a predetermined relationship to the inactive threshold.

7. The method of claim 6, the method further comprising:

measuring a rate of current connections of the second active server; and

placing the second active server into an inactive state when the rate of current connections of the second active server achieves the inactive threshold.

8. The method of claim 6, wherein the keep-alives are sent after the active threshold is met, wherein when the standby threshold is met, the keep-alives are not sent.

9. The method of claim 1, wherein the first and second active servers each include a cluster of servers.

10. The method of claim 1, wherein each server has associated devices, the method further comprising:

placing devices associated with a particular computer into an active state when the particular server is placed into an active state, and placing the devices into an inactive state when the particular server is placed into an inactive state.

11. The method of claim 1, wherein a system includes multiple levels, wherein multiple servers are included in each level, the method further comprising:

assigning different weights to two or more of the levels; and

using the different weights to determine the predetermined relationship with the set limit, wherein a decision to place any servers in a particular level into an active state includes use of a weight associated with the particular level.

12. The method of claim 1 additionally comprising measuring a time for the second inactive server to transition into the active state.

13. The method of claim 12 additionally comprising moving the pre-active threshold or a standby threshold based on the time for transition of the second inactive server into the active state.

14. The method of claim 1 additionally comprising:

measuring a frequency of server-standby timeouts for the second active server; and

moving the pre-active threshold or a standby threshold based on frequency of server-standby timeouts.

15. A method for managing power among servers in a server farm, the method comprising:

predetermining a threshold limit on the number of connections for a first server in the server farm;

monitoring the first server to determine when a number of connections being handled by the first server is near the threshold limit;

powering up a second server from a low-power state to a active state when the first server'"'"'s number of connections is near the threshold limit, wherein a particular server in the low-power state consumes energy at a lower rate than the particular server in the active state;

initiating periodic keep alives to the second server being powered up to keep the second server powered up before the second server is activated to be assigned connections,

wherein the periodic keep alives do not place the inactive server into an active state;

activating the second server upon reaching the threshold limit;

assigning connections to the second server;

monitoring the second server to determine when its number of connections falls below a lower limit; and

in response to the second server'"'"'s connections falling below the lower limit, placing the second server into the low-power state.

16. The method of claim 15 additionally comprising:

measuring the time for the second server to transition from the active state into the low-power state; and

moving when the second server is powered up the threshold limit or a standby threshold based on the time for transition of the second inactive server into the active state.

17. An apparatus for controlling power to servers, the apparatus comprising:

a computer processor coupled to a plurality of servers;

a machine-readable storage medium coupled to the computer processor, wherein the machine-readable storage medium includes one or more instructions executable by the computer processor for;

setting a limit of resource use for a first active server;

measuring the resource use of the first active server;

receiving a request for additional resource use;

when the measured resource use of the first active server plus the request for additional resource are in a predetermined relationship with a pre-active threshold of a first connection rate, then

initiating powering up of a inactive server into an active state in response to the measured resource use of the first active server achieving the pre-active threshold, the first connection rate of the pre-active threshold being less than a second connection rate of an active threshold, wherein said initiating powering up of the inactive server includes;

sending a power up signal to the inactive server to initiate powering up of the inactive server before the second connection rate of the active threshold is reached; and

upon reaching the pre-active threshold, initiating periodic keep alives to the inactive server to keep the inactive server powered up after the pre-active threshold is met and before the inactive server is activated to be assigned requests for additional resource use, wherein the periodic keep alives do not place the inactive server into an active state;

placing the inactive server into an active state as a second active server when the active threshold is met; and

assigning requests for additional resource use to the second active server.

18. A machine-readable storage medium including instructions executable by a computer processor, the machine-readable storage medium comprising:

one or more instructions for setting a limit of resource use for a first active server;

one or more instructions for measuring the resource use of the first active server;

one or more instructions for receiving a request for additional resource use;

one or more instructions for when the measured resource use of the first active server plus the request for additional resource are in a predetermined relationship with a pre-active threshold of a first connection rate, then

one or more instructions for initiating powering up of a inactive server into an active state in response to the measured resource use of the first active server achieving the pre-active threshold, the first connection rate of the pre-active threshold being less than an active threshold of a second connection rate, wherein said one or more instructions for initiating powering up of the inactive server includes;

one or more instructions for sending a power up signal to the inactive server to initiate powering up of the inactive server before the second connection rate of active threshold is reached; and

upon reaching the pre-active threshold, one or more instructions for initiating periodic keep alives to the inactive server being powered up to keep the inactive server powered up after the pre-active threshold is met and before the inactive server is activated to be assigned requests for additional resource use, wherein the periodic keep alives do not place the inactive server into an active state; and

one or more instructions for placing the inactive server into an active state as a second active server when the active threshold is met; and

one or more instructions for assigning requests for additional resource use to the second active server.

19. A method for controlling power consumption of servers in a group of servers, the method comprising:

assigning servers to an active pool or an inactive pool;

setting a limit of resource use for the active pool;

measuring the resource use of the active pool;

receiving a request for additional resource use;

when the measured resource use of the active pool plus the request for additional resource are in a predetermined relationship with a pre-active threshold of a first connection rate, then

initiating powering up of at least one server into an active pool in response to the measured resource use of the active pool achieving the pre-active threshold, the first connection rate of the pre-active threshold being less than a second connection rate of an active threshold;

subsequent to the step of initiating powering up of and before placing at least one server from the inactive pool into the active pool, initiating periodic keep alives to the at least one server to keep the at least one server powered up after activating the at least one server and before the at least one server is placed in the active pool to be assigned connections, wherein the periodic keep alives do not place the inactive server into an active state;

placing the at least one server from the inactive pool into the active pool upon reaching the active threshold; and

upon reaching the active threshold, assigning requests for additional resource use to the active pool, the active pool including the at least one server.

20. The method of claim 19, wherein measuring the resource use includes measuring a current connection rate, the method further comprising:

determining a threshold connection rate for the active pool; and

detecting when incoming requests for additional connections would exceed the pre-active threshold.

21. The method of claim 19, wherein said resource use includes a network connection request which includes a request for web page content or a request for data from a database.