[Fwd: economy based cluster scheduling: a new way to allocating resources in clusters...]
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Rajkumar Buyya raj at cs.mu.oz.auWed Aug 7 02:35:06 PDT 2002
- Previous message: Windows HPC
- Next message: Beep a node
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Dear All, I am forwarding this message as I believe that it will be interest to some of you. Particularly: Libra: Economic-based Scheduler for Service-Oriented Cluster Computing You can find a link to it from: http://gridbus.org/ It is currently implemented as a plug-in scheduler for PBS. A short message on this work that I posted to TFCC discussion list is given below. Of course if you like to extend/collaborate on future work of this project, you are welcome: ----------- I would like share information about our recent work on cluster scheduling and seek your comments on the same [This is also an attempt to apply our previous work (for the past 3+ years) on Economic paradigms for Grid Computing, more about which can be found at: http://buyya.com/ecogrid/]: In an attempt to apply Economy and user QoS requirements driven mechanism for Cluster resource management and scheduling, we created a new scheduler (called Libra, a component of Gridbus). It essentially allows users to submit jobs to cluster in normal way, but users have an opportunity to add two more parameters: "deadline" and "budget" to their "job script" before submission. We found that this approach is able to complete more jobs within deadline than without it. Additionally this approach provides incentive for users for submitting a job to cluster with relaxed deadline. This initial concept is being demonstrated by implementing a pluggable Libra scheduler for PBS (and currently porting it to SGE). A technical writeup on this can be accessed from: http://gridbus.org/ [click on Libra or it can be directly accessed from: http://buyya.com/papers/libra.pdf] I personally believe that a service like this at cluster level will help in providing an enforceable QoS services and supporting so called "service-level agreements", a term quite often used within Web services community I guess. Comments (even collaborations) are welcome and appreciated as this is an ongoing effort. cheers Raj ------------------------------------------------------------------------ Rajkumar Buyya Grid Computing and Distributed Systems (GRIDS) Lab: http://gridbus.org Dept. of Computer Science and Software Engineering The University of Melbourne SEECS Building, Room. 3.38 221 Bouverie St., Carlton Melbourne, Australia Phone: +61-3-8344 9317 (office); +61-3-9571 3629 (home) Fax: +61-3-9348 1184; eFax: +1-801-720-9272 Email: rajkumar at buyya.com | raj at cs.mu.oz.au URL: http://www.buyya.com | http://www.csse.melbourne.edu/~raj ------------------------------------------------------------------------
- Previous message: Windows HPC
- Next message: Beep a node
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
