On demand node and server instance allocation and de-allocation

• US 20050038789A1
• Filed: 08/12/2004
• Published: 02/17/2005
• Est. Priority Date: 08/14/2003
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Stamos, James W., Lu, Edwina, Chidambaran, Lakshminarayanan, Pingte, Rajendra
• Granted Patent: US 7,441,033 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources
• CPC Class Codes: H04L 41/0668 by dynamic selection of recovery network elements, e.g. replacement by the most appropriate element after failure
• CPC Class Codes: H04L 41/5009 Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
• CPC Class Codes: H04L 41/5025 by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
• 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: H04L 67/1034 Reaction to server failures by a load balancer

Abstract

An approach efficiently and dynamically allocates and de-allocates database instances and nodes between databases. The approach performs this function in a way that accounts for the users of and services hosted by multiple databases.

Claims

1. A method for managing resources in a multiple node system, the method comprising the steps of:

said multiple node system detecting a violation of a service-level agreement on a multi-node server, wherein said multi-node server includes a set of server instances that run a first set of nodes;

in response to detecting said violation, said multiple node system determining whether to resolve said violation by using a node that is not a member of said first set of nodes to host another server instance of said multi-node server; and

in response to determining to resolve said violation by using a node not a member of said first set of nodes, attempting to resolve said violation by performing steps that include;

said multiple node system finding another node to use to host a second server instance of said multi-node server, and

said multiple node system starting said second server instance of said multi-node server on said other node.

2. The method of claim 1, wherein:

after finding said other node and before starting said second server instance, said node hosts a third server instance; and

the step of attempting to resolve said violation further includes shutting down said third server instance before starting said server instance.

3. The method of claim 1, wherein the step of finding another node further includes:

generating a set of candidates that correspond to another multi-node server hosted by another set of nodes different than said first set of nodes; and

selecting a victim from said set of candidates from which to deallocate a node.

4. The method of claim 3, wherein:

the steps further include transmitting a request to a multi-node server corresponding to a candidate of said set of candidates to volunteer to quiesce a server instance on a node; and

selecting a victim includes selecting a victim from a candidate of a plurality of multi-node servers that transmit, in response to said request to quiesce, a message to volunteer to quiesce.

5. The method of claim 4, wherein the step of transmitting a request to a multi-node server includes broadcasting the request to the set of candidates.

6. The method of claim 4, wherein the step of selecting a victim from a candidate of a plurality of multi-node servers includes selecting a victim based on one or more factors that include at least one of:

priority data that is supplied by a user and that indicates a priority of a candidate of said plurality of multi-node servers relative to other candidates of said plurality of multi-node servers; and

a quiescing cost associated with quiescing a server instance on a node.

7. The method of claim 2, wherein the step of attempting to resolve said violation includes:

transmitting a request to a second multi-node server to prepare a server instance to shut down; and

receiving a message from the second multi-node server indicating that a server instance is prepared to shut down.

8. The method of claim 7, wherein:

the message indicating a server instance is prepared further includes data identifying the second server instance and node.

9. A method used for managing resources in a multiple node system, the method comprising the steps of:

a first multi-node server running on a plurality of nodes of said multiple node system, said first multi-node server comprising a plurality of server instances running on said plurality of nodes;

said first multi-node server receiving a request to acknowledge whether at least one server instance may be relinquished;

in response to receiving said request, said first multi-node server determining whether using at least one less server instance causes a service-level violation; and

if said first multi-node server determines that at least one server instance does not cause a service-level violation, then transmitting a message indicating that at least one server instance of a plurality of server instances may be relinquished.

10. The method of claim 9, wherein determining whether using at least one less server instance causes a service-level violation includes determining whether a service-level agreement for a type of work performed by said first multi-node server would be violated.

11. The method of claim 10, wherein the service-level agreement for a type of work performed is based on a cardinality of server instances allocated to that type of work.

12. The method of claim 9, wherein determining whether using at least one less server instance causes a service-level violation includes determining whether there is sufficient available unused CPU utilization to absorb workload of the plurality of nodes with at least one less node.

13. The method of claim 9, wherein:

said multiple node system includes a plurality of other multi-node servers including a second multi-node server;

a priority is associated with said plurality of other multi-node servers;

a request for another server-instance for said second multi-node server causes transmitting the request to acknowledge whether at least one server instance may be relinquished;

the steps further include, in response to receiving said message indicating that at least one server instance may be relinquished, allocating said at least one server instance to a multi-node server associated with a higher level of said priority; and

said multi-node server associated with a higher level of said priority is either said second multi-node server or a different multi-node server of said plurality of other multi-node servers.

14. A method used for managing resources in a multiple node system, the method comprising the steps of:

a multi-node server running on a plurality of nodes of said multiple node system, said multi-node server comprising a plurality of server instances running on the plurality of nodes;

said multi-node server receiving a request to acknowledge whether at least one server instance may be relinquished;

in response to receiving said request, selecting, based on a measure of performance, a particular server instance of said plurality of server instances on which to cease performing work for clients of said multi-node server; and

causing said particular server instance to cease performing work for said clients.

15. The method of claim 14, wherein said measure of performance is based on CPU utilization of a node on which a server instance resides.

16. The method of claim 14, wherein:

the particular server instance has a session established for a client of said multi-node server;

the steps further include determining whether said session may be migrated from said particular server instance to another of said plurality of server instances; and

the step of causing said particular server instance to cease performing work for said clients is performed only if determining that said session may be migrated from said particular server instance.