Data processing system and method for iteratively re-distributing objects across all or a minimum number of processing units

• US 8,224,938 B2
• Filed: 07/08/2005
• Issued: 07/17/2012
• Est. Priority Date: 08/13/2004
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Sauermann, Volker
• Primary Examiner(s): Dollinger, Tonia L. M.
• Assistant Examiner(s): Murray, Daniel C
• Application Number:
  US11/176,268
• Publication Number:
  US 20060020767A1
• Time in Patent Office: 2,566 Days
• Field of Search: 709/226, 709/223, 718/102, 718/105
• US Class Current: 709/223
• CPC Class Codes: G06F 9/5066 Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs mappping at compile time, see G06F8/451

Abstract

A data processing system has a program module for determining a re-distribution of objects, such as database tables, over processing units of a cluster of processing units, such as in a blade computing environment. An estimate of the required duration for applying the re-distribution is calculated in order to facilitate an administrator'"'"'s decision when to initiate the re-distribution procedure.

Claims

1. A data processing system for assigning a set of objects to processing units of a cluster of processing units, the data processing system comprising:

first determining means that determines a re-distribution of the objects over the processing units, the re-distribution specifying which objects to assign to a set of required processors among the processing units, wherein the objects include database tables, wherein each processing unit includes a memory unit;

calculating means that calculates an expected duration required for applying the re-distribution, wherein the calculation is based on at least a time parameter and the aggregated size of the objects to be redistributed, the time parameter providing an estimate of an amount of down time for the processing units;

a user interface that shows the expected duration to a user; and

a processor for performing operations including;

prompting the user, by the user interface, to select whether to distribute the objects according to one of a first and second modes;

wherein

in the first mode, the set of required processors includes all the processing units in the cluster; and

in the second mode, the set of required processors consists of a minimum number of processing units determined using a two-dimensional minimization procedure comprising calculating a sorting index for each of the objects based on the normalised object size and load of each of the objects;

creating an object sequence for re-distribution by sorting the objects by size;

iteratively redistributing the objects in the object sequence across the required processing units by, for each required processing unit;

determining a gap in the associated memory unit, wherein the gap represents free space on the memory unit;

identifying the largest remaining object in the object sequence that will fit into the gap;

adding the identified object to the memory unit; and

removing the identified object from the object sequence; and

after the object sequence has no more objects to redistribute, outputting the assignment of objects.

2. The data processing system of claim 1, wherein the calculating means that calculates the expected duration is further operable to:

determine a sub-set of the processing units affected by the re-distribution, and

determine a sub-set of the set of objects, the sub-set of the set of objects consisting of objects that are assigned to the sub-set of the processing units in the re-distribution.

3. The data processing system of claim 1, further comprising input means for a user'"'"'s acceptance of a suggested re-distribution by a single input action.

4. The data processing system of claim 3, further comprising second determining means that determines a start time for applying the re-distribution using a load profile.

5. The data processing system of claim 1, further comprising storage means for storing a current distribution and the re-distribution.

6. The data processing system of claim 5, wherein the system comprises a blade server, and the blade server is one of the effected processing units.

7. The data processing system of claim 1, wherein the system provides a management process of a distributed search engine.

8. A method of assigning a set of objects to processing units of a cluster of processing units, the method comprising:

determining, via a processor, a re-distribution of the objects over the processing units, the re-distribution specifying which objects will be assigned to a set of required processors among the processing units, wherein the objects include database tables, wherein each processing unit includes a memory unit;

calculating, via the processor, an expected duration required for applying the re-distribution of the objects, wherein the calculating is based on at least a time parameter and the aggregated size of the objects to be redistributed, the time parameter providing an estimate of an amount of down time for the processing units;

showing, via a display device, the expected duration to a user;

prompting the user, on the display device, to select whether to distribute the objects according to one of a first and second modes;

wherein

in the first mode, the set of required processors includes all the processing units in the cluster; and

in the second mode, the set of required processors consists of a minimum number of processing units determined using a two-dimensional minimization procedure comprising calculating a sorting index for each of the objects based on the normalised object size and load of each of the objects;

creating, via the processor, an object sequence for re-distribution by sorting the objects by size;

iteratively redistributing, via the processor, the objects in the object sequence across the required processing units by, for each processing unit;

determining a gap in the associated memory unit, wherein the gap represents free space on the memory unit;

identifying the largest remaining object in the object sequence that will fit into the gap;

adding the identified object to the memory unit; and

removing the identified object from the object sequence; and

after the object sequence has no more objects to redistribute, outputting, via the display device, the assignment of objects.

9. The method of claim 8, further comprising:

determining a sub-set of the processing units affected by the re-distribution; and

determining a sub-set of the set of objects, the sub-set of the set of objects consisting of objects that are assigned to processing units of the sub-set of the processing units.

10. The method of claim 9, further comprising entering a user'"'"'s acceptance of the re-distribution by means of a single input action.

11. The method of claim 10, further comprising determining the start time for applying the re-distribution using a load profile.

12. A non-transitory computer-readable storage medium comprising computer executable instructions, which instructions when executed perform stages, comprising:

determining a re-distribution of objects over a cluster of processing units, the re-distribution specifying which objects to assign a set of required processors among the processing units;

calculating an expected duration required for applying the re-distribution of the objects, wherein the calculating is based on at least a time parameter and the aggregated size of the objects to be redistributed, the time parameter providing an estimate of an amount of down time for the processing units;

showing the expected duration to a user;

prompting the user to select whether to distribute the objects according to one of a first and second modes;

wherein

in the first mode, the set of required processors includes all the processing units in the cluster; and

in the second mode, the set of required processors consists of a minimum number of processing units determined using a two-dimensional minimization procedure comprising calculating a sorting index for each of the objects based on the normalised object size and load of each of the objects;

creating an object sequence for re-distribution by sorting the objects by size;

iteratively redistributing the objects in the object sequence across the effected processing units by, for each processing unit;

determining a gap in the associated memory unit, wherein the gap represents free space on the memory unit;

identifying the largest remaining object in the object sequence that will fit into the gap;

adding the identified object to the memory unit; and

removing the identified object from the object sequence; and

after the object sequence has no more objects to redistribute, outputting the assignment of objects