SERVER MANAGEMENT APPARATUS AND SERVER MANAGEMENT METHOD

• US 20110099403A1
• Filed: 02/18/2010
• Published: 04/28/2011
• Est. Priority Date: 10/26/2009
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Ohta, Tomoya, Miyata, Yasushi, Nishiyama, Hiroyasu
• Granted Patent: US 8,286,165 B2
• Time in Patent Office: Days
• US Class Current: 713/323
• CPC Class Codes: G06F 1/32 Means for saving power
• CPC Class Codes: G06F 2009/45562 Creating, deleting, cloning virtual machine instances
• CPC Class Codes: G06F 2009/4557 Distribution of virtual machine instances; Migration and load balancing
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/5072 Grid computing
• CPC Class Codes: G06F 9/5088 involving task migration
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A server management apparatus for lowering migration costs during scale-in/scale-out and workload consolidation of a cluster system(s) to thereby reduce power consumption is disclosed. The apparatus manages a physical server group which renders operative a virtual server group thereon and, when putting into practice a cluster system including a plurality of virtual servers placed in the physical server group, manages the layout state of virtual servers pursuant to the load state of the virtual server group. When executing scale-in, a virtual server operating on a physical server with the minimum number of operative virtual servers is specified as a shutdown target. When executing scale-out, a workload variation is predicted to control a scale-out destination of cluster system so that load variation-resembled cluster systems gather on the same physical server. The scale-in execution timing is delayed if the predicted load variation tends to rise and accelerated if it falls.

Claims

1. A server management apparatus for managing a physical server group including physical servers rendering operative thereon a virtual server group having a plurality of virtual servers and for managing, when putting into practice a cluster system containing therein said plurality of virtual servers placed in said physical server group, a placement state of said virtual servers depending on a load state of said virtual server group, wherein said server management apparatus comprises:

a load information collection unit which collects load information of the virtual server group constituting said cluster system;

a configuration information collection unit which collects configuration information indicating that said virtual server group is presently assigned to which one of said physical servers;

a configuration change judgment unit which determines from the collected load information whether the virtual servers constituting said cluster system are reducible in number;

a configuration change target selection unit which selects from the collected configuration information a physical server which is least in operative virtual server number; and

a configuration change execution unit which deactivates a virtual server operating on the selected physical server.

2. The server management apparatus according to claim 1, wherein said server management apparatus further executes, prior to the deactivation of the virtual server specified as a deactivation target, scale-in processing for stopping request allocation to said virtual server of the deactivation target and for deleting it from the cluster system.

3. The server management apparatus according to claim 1, wherein said server management apparatus is further arranged so that;

said configuration change judgment unit determines from said collected load information whether the number of virtual servers constituting the cluster system is excessive or deficient;

when it is judged that the number of virtual servers constituting said cluster system is deficient, said configuration change target selection unit selects a physical server which is largest in operative virtual server number from the configuration information collected by said configuration information collection unit; and

said configuration change execution unit activates a virtual server on the selected physical server.

4. The server management apparatus according to claim 3, wherein said server management apparatus further executes scale-out for adding the activated virtual server to said cluster system and for starting request allocation.

5. The server management apparatus according to claim 1, wherein said server management apparatus further comprises a load variation similarity calculation unit which derives from the load information which has been collected by said load information collection unit a degree of similarity of load variation between a plurality of cluster systems, and wherein

said configuration change judgment unit judges from said collected load information whether scale-out for virtual server addition is necessary or not,

when a result of the judgment indicates that the scale-out is necessary, said configuration change target selection unit selects a physical server containing a larger number of virtual servers constituting a cluster system which is high in the derived degree of load variation similarity;

and

said configuration change execution unit determines the selected physical server to be a scale-out destination and then executes the scale-out.

6. The server management apparatus according to claim 1, wherein said server management apparatus further comprises:

a physical server power supply control unit which determines whether a virtual server is absent with respect to said physical server as a result of configuration change performed by said configuration change execution unit and for executing, when it is judged that such virtual server is absent, power-off of the physical server,

7. The server management apparatus according to claim 6, wherein after said configuration change execution unit has executed the scale-in,

said configuration change judgment unit judges whether it is possible to power off the physical server by operative virtual server migration,

said configuration change target selection unit migrates one or more than one virtual server between physical servers when it is judged that it is possible to power off the physical server,

said configuration change execution unit judges whether there is a virtual server operating on a physical server of migration source, and

said physical server power supply control unit performs power-off of the physical server if the virtual server presence judgment indicates that no operative virtual servers exist therein.

8. The server management apparatus according to claim 2, wherein said server management apparatus further comprises a load variation prediction unit which determines from the load information collected from said load information collection unit whether a load at a present time point is on an upward trend or on a downward trend with respect to each cluster system, and wherein

said load variation prediction unit modifies a threshold for use in scale-in execution possible/impossible judgment so that it becomes higher if the load is on the downward trend, and, if the load is on the upward trend, modifies the threshold for use in scale-in execution possible/impossible judgment so that it becomes lower in value.

9. The server management apparatus according to claim 5, wherein said configuration information collection unit further collects as said configuration information an amount of unused computing resources held by said physical server,

said configuration change judgment unit judges based on the unused computing resource amount thus collected whether scale-up for increasing the computing resource amount to be assigned to virtual server is executable or not,

said configuration change target selection unit applies scale-up to this virtual server when the judgment indicates that the scale-up is executable, and

said configuration change execution unit manages a load dispersion device which allocates a request from a client to a virtual server group constituting the cluster system and controls said load dispersion device to thereby increase an amount of request allocation to the virtual server which has executed the scale-up to an extent greater than that of other virtual servers within the cluster system.

10. The server management apparatus according to claim 2, wherein said server management apparatus stores in a storage device an operational policy including a minimal physical server number during the scale-in,

said configuration change target selection unit selects a virtual server to be deactivated in such a manner that its physical server number is not less than the minimal physical server number upon execution of the scale-in, and

the scale-in of the virtual server to be deactivated is executed.

11. The server management apparatus according to claim 2, wherein said load information collection unit further collects a power consumption amount of the operative physical servers and stores it as said configuration information, and

said configuration change target selection unit acquires from said configuration information a power consumption amount of a physical server on which the virtual server to be deactivated when the scale-in is performed is rendered operative, and determines a virtual server operating on a physical server which is greater in the acquired power consumption amount to be the virtual server that is deactivated when the scale-in is done.

12. The server management apparatus according to claim 5, wherein said load information collection unit collects, when said physical server is registered to said server management apparatus, the configuration information containing therein the computing resource amount and registers it to a storage device,

said configuration change judgment unit acquires, at a time of scale-out target judgment, the computing resource amount of a physical server under suspension and determines a scale-out destination while letting the physical server under suspension be included therein,

said configuration change target selection unit is such that said server management apparatus activates the physical server under suspension when the server under suspension is specified as the scale-out destination by the judgment, and

said configuration change execution unit activates a virtual server on the activated physical server to thereby perform the scale-out,

13. The server management apparatus according to claim 7, wherein when migrating said one or more than one virtual server between physical servers, in activation/deactivation operations of the virtual server operating on a physical server of migration source which constitutes said cluster system and a virtual server which is in a halt state in a physical server of migration destination,

said server management apparatus requests the migration destination physical server to activate a virtual server of migration destination being set in the halt state in the migration destination physical server, requests a load dispersion device which allocates a request from a client to start request allocation to the migration destination virtual server, requests said load dispersion device to stop the request allocation to a virtual server of migration source, and requests the migration source physical server to shut down the migration source virtual server operating in said migration source physical serve r.

14. The server management apparatus according to claim 5, wherein said load variation similarity calculation unit acquires the load information collected from said load information collection unit in units of cluster systems and determines those cluster systems which are identical to each other for a longer time in cluster system load increasing time period and in cluster system load decreasing time period to be cluster systems high in degree of similarity.

15. The server management apparatus according to claim 14, wherein said load variation similarity calculation unit acquires, in units of cluster systems, past load variation data based on the load information collected from said load information collection unit and an operative virtual server number at that time, and determines cluster systems which are identical to each other for a longer time in the operative virtual server number of cluster system to be cluster systems high in similarity degree,

16. The server management apparatus according to claim 15, wherein said load variation similarity calculation unit determines cluster systems which are identical to each other in time period which is less in number of operative virtual servers of cluster system rather than in time period which is larger in number of operative virtual servers of cluster system to be higher in similarity degree.

17. The server management apparatus according to claim 15, wherein said load variation similarity calculation unit determines cluster systems which are identical to each other in the operative virtual server number in a time zone which is nearer to a judgment execution time point after this judgment execution time point.

18. The server management apparatus according to claim 15, wherein said load variation similarity calculation unit extracts the similarity degree of a load variation from the load information at a timing being the same as a timing for execution of an extraction work.

19. The server management apparatus according to claim 15, wherein said load variation similarity calculation unit applies weighting to a number of virtual servers which have been rendered operative in the past of a prespecified time in such a way as to determine that the similarity degree of a load variation of nearer past is higher.

20. The server management apparatus according to claim 8, wherein said load variation prediction unit further divides at predetermined time intervals data obtained by dividing per prespecified time period the load variation data based on said load information collected from said load information collection unit, performs calculation to determine whether the load variation is on an upward trend or on a downward trend within such time period, and judges a present state to be an unstable state if calculated numbers of the upward trend and the downward trend are both beyond a specified value.

21. The server management apparatus according to claim 8, wherein said load variation prediction unit applies weighted averaging to the load variation data based on the load information collected from said load information collection unit in such a way as to enlarge a load variation of nearer past and simultaneously lessen a load variation of farther past.

22. The server management apparatus according to claim 8, wherein said load variation prediction unit predicts a load variation tendency from load information at a timing which is the same as a timing for execution of the prediction of load variation tendency.

23. A server management method for use in a server management apparatus operative to manage a physical server group including physical servers for rendering operative thereon a virtual server group having a plurality of virtual servers, said method being for managing a placement state of the virtual servers depending on a load state of said virtual server group when handling for practical use a cluster system containing therein said plurality of virtual servers placed in said physical server group, said method comprising the steps of:

collecting load information of the virtual server group constituting said cluster system,

collecting configuration information indicating that said virtual server group is currently assigned to which one of said physical servers,

determining from the collected load information whether scale-in processing for deactivating a virtual server is executable or not,

selecting from the collected configuration information a physical server which is least in number of operative virtual servers, and

selecting a virtual server operating on the selected physical server as a target to be deactivated after execution of the scale-in and then executes the scale-in.

24. The server management method according to claim 23, wherein said server management apparatus further performs operations including;

determining from said collected load information whether scale-out processing for adding a virtual server is needed or not,

when a result of the determination indicates that the scale-in is needed, selecting from said collected configuration information a physical server which is largest in number of operative virtual servers, and

activating a virtual server on the selected physical server to thereby execute the scale-out,

25. The server management method according to claim 23, wherein said server management apparatus further performs operations including:

deriving from the load information collected in past times a degree of similarity of a load variation between a plurality of cluster systems,

judging from the collected load information whether the scale-out for virtual server addition is needed or not,

when the judgment indicates that the scale-out is needed, selecting a physical server on which a larger number of virtual servers constituting a cluster system being high in the derived load variation similarity exist thereon, and

determining the selected physical server to be a scale-out destination to thereby execute the scale-out.