home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!decwrl!spool.mu.edu!agate!darkstar.UCSC.EDU!osr
- From: philk@fox.cis.ohio-state.edu (Phillip E Krueger)
- Newsgroups: comp.os.research
- Subject: Re: Difference between DYNAMIC and ADAPTIVE load balancing???
- Date: 5 Nov 1992 05:12:08 GMT
- Organization: The Ohio State University Dept. of Computer and Info. Science
- Lines: 43
- Approved: comp-os-research@ftp.cse.ucsc.edu
- Message-ID: <1daaf8INNeb4@darkstar.UCSC.EDU>
- References: <1d9guvINN6rp@darkstar.UCSC.EDU>
- NNTP-Posting-Host: ftp.cse.ucsc.edu
- Originator: osr@ftp
-
- In article <1d9guvINN6rp@darkstar.UCSC.EDU> Amin writes:
- >
- >Hi there. Subject says it all. I have seen many papers about dynamic and
- >adaptive load balancing, but I couldn't understand the difference.
- >Are these really two different categories in load balancing, or are these
- >two words used interchangeably in the literature. Please post your
- >response, so that other people can post their views about the responses.
- >
- >Thanks.
- >Amin
-
- The difference between static and dynamic (I'll get to adaptive in a second)
- load distributing (or global scheduling) algorithms is the use of system
- state information. While static algorithms make no use of such information,
- dynamic algorithms use such information to improve individual job transfer
- decisions.
-
- Adaptive algorithms constitute a subset of dynamic algorithms. Adaptive
- algorithms go further in their use of system state information: Such
- information may be used to modify the parameters of the algorithm, or even to
- choose which load distributing strategy is used. For example, you might find
- that load distributing algorithm A is best under system conditions X, while
- algorithm B is better under conditions Y. An adaptive algorithm might choose
- which strategy to pursue based on whether the conditions it observes are closer
- to X or Y. As a more detailed example, sender-initiated algorithms tend to
- perform better than receiver-initiated algorithms when the overall system load
- is low, though receiver-initiated algorithm performs better when the system
- load is high (sender-initiated algorithms are disastrous at high loads). An
- adaptive algorithm might choose to follow a sender-initiated strategy when the
- load is low, but switch to a receiver-initiated strategy if the load becomes
- high.
-
- Examples of adaptive algorithms can be found in:
- N. G. Shivaratri and P. Krueger, "Two Adaptive Location Policies for Global
- Scheduling Algorithms," Proc. 10th International Conference on Distributed
- Computing Systems, pp. 502-509 (May 1990).
-
- -- Phil Krueger
- --
- Assistant Professor
- Ohio State University philk@cis.ohio-state.edu
- Dept. of Computer and Info. Science ..!osu-cis!cis.ohio-state.edu!philk
- 2036 Neil Ave. Columbus, OH 43210-1277 (614) 292-2565
-