Cirne, W., and Marzullo, K. The computational Co-op: Gathering clusters into a metacomputer. In PPS/SPDP'99 Symposium (1999).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Therefore, we cannot use preemption, what makes it harder to implement a predictable schedule. However, the previous work we have done in the Computational Co op suggests that, even when preemption is not an option, it is possible to be enough predictable to support application scheduling [Cirne 99a] The Bushel of AppLeS problem seems to be more critical when parallel ap plications are involved because they are not work conservative across partition sizes. That is, for most parallel applications, the amount of computational work they require varies depending on the size of the ....

....their users. Predictability comes from the fine grained implementation of the policy that determines how resources are to be shared. In a smaller scale, some researchers have started exploring the predictability of resource scheduling as a way to enable multiple schedulers to coexist [Chapin 95] Cirne 99a] Harty 96] These results are naturally more metacomputing oriented. The focus here is on where to draw the line dividing the responsibility of resource and application schedulers, and what interface should one export to the other. Predictability appears as a requisite for application ....

Walfredo Cirne and Keith Marzullo. The Computational Co-op: Gathering Clusters into a Metacomputer. In Proceeding of IPPS/SPDP'99, April 1999.

No context found.

Cirne, W., and Marzullo, K. The computational Co-op: Gathering clusters into a metacomputer. In PPS/SPDP'99 Symposium (1999).

No context found.

Cirne, W., and Marzullo, K. The computational Co-op: Gathering clusters into a metacomputer. In PPS/SPDP'99 Symposium (1999).

....users, but also that its users will access the other institution s resources, maybe in equal proportions. Existing solutions in grid computing allow these two institutions to de ne some policies in their resource sharing, creating static constraints and guarantees to the users of the grid [15, 28, 12]. However, if a third institution C joins the grid, new agreements must be negotiated between the institutions and con gured on each of them. We can easily see that these mechanisms are neither scalable nor exible enough to the large scale grids scenarios. 2.1 Related work Although grid ....

....needed to use grids matured. Past e orts have been spent in de ning mechanisms that support static access policies and constraints to allow the building of metacomputing infrastructures across di erent administrative domains like in the Condor system [28] and in the Computational Co op [12]. Since 1984 the Condor system has used di erent mechanisms for allowing a Condor user to access resources across institutional boundaries. After trying to use institutional level agreements [16] Condor was changed to a user to institution level [28] to provide exibility, as requested by its ....

Cirne, W., and Marzullo, K. The computational Co-op: Gathering clusters into a metacomputer. In PPS/SPDP'99 Symposium (1999).

....a fine level of control on how the resources are shared among their users. Predictability comes as a result of this fine level of control. To a lesser degree, some researchers have started exploring the predictability of resource scheduling as a way to enable multiple schedulers to coexist [20] [23] [57] These results are naturally more oriented to grid computing. The focus here is on where to draw the line dividing the responsibility of resource and application schedulers, and what interface should one export to the other. Predictability appears as a requisite for good application ....

Walfredo Cirne and Keith Marzullo. The Computational Co-op: Gathering Clusters into a Metacomputer. In Proceeding of IPPS/SPDP'99, April 1999.

....hand, has either ignored the differences between Sarah and Bob s requirements or has gravitated towards Bob s since they are more technically challenging. Moreover, by committing to a particular Grid Computing infrastructure (e.g. Condor pools [11] Globus resources [7] Computational Co ops [4], or Nile farms [1] both Sarah and Bob must restrict their application to utilize only processors available through such an infrastructure. Bob may be happy to do this, but Sarah might have access to many other processors elsewhere. And, since her application is coarse grained, using whatever ....

....applications (as Condor [11] and high energy physics applications (as Nile [1] for example. Yet other efforts have addressed specific aspects of the Grid Computing infrastructure, such as supporting the federation of independent sites into large scale grids, as the Computational Co op [4]. All these projects view Grid Computing infrastructure being deployed as universally available system services. 3c) 3b) task done (4) remote exec (3) playpen, file xfer, and remote exec (3a) 2) add task (1) Home Machine Grid Machine Task Manager User Agent Server home ....

W. Cirne and K. Marzullo. The Computational Co-op: Gathering Clusters into a Metacomputer. In Proceeding of IPPS/SPDP'99, April 1999.

....other hand, has either ignored the differences between Sarah and Bob s requirements or has gravitated towards Bob s since they are more technically challenging. Moreover, by committing to a particular Grid Computing infrastructure (e.g. Condor pools [10] Globus resources [7] Computational Co ops [4], or Nile farms [1] both Sarah and Bob must restrict their application to utilize 2 only processors available through such an infrastructure. Bob may be happy to do this, but Sarah might have access to many other processors elsewhere. And, since her application is coarse grained, using whatever ....

....applications (as Condor [10] and high energy physics applications (as Nile [1] for example. Yet other efforts have addressed specific aspects of the Grid Computing infrastructure, such as supporting the federation of independent sites into large scale grids, as the Computational Co op [4]. All these projects view Grid Computing infrastructure being deployed as universally available system services. Open Grid, in opposition, adopts a user centric approach for providing Grid Computing services. Usercentric approaches are recognized as the best strategy for scheduling in grids, in ....

W. Cirne and K. Marzullo. The Computational Co-op: Gathering Clusters into a Metacomputer. In Proceeding of IPPS/SPDP'99, April 1999.

....system load. That is, in order to make reasonable decisions, the application scheduler needs information on how the resource schedulers are going to deal with its requests. Although some have proposed mechanisms to promote the effective communication among the different schedulers in the system [10,7], the resource schedulers currently in use have not been designed with this need in mind. Therefore, researchers in metacomputing have developed tools that monitor and forecast how long a request is going to take to run over a particular set of resources (e.g. 50] However, today there is no ....

W. Cirne and K. Marzullo, "The computational co-op: gathering clusters into a metacomputer". In Second Merged Symposium IPPS/SPDP 1999, 13th International Parallel Processing Symposium & 10th Symposium on Parallel and Distributed Processing, pp. 160--166, April 1999.

....the current system load. That is, in order to make reasonable decisions, the meta scheduler needs information on how the machines schedulers are going to deal with its requests. Although some have proposed mechanisms to promote effective communication among the different schedulers in the system [11,8], the machine schedulers currently in use have not been designed with this need in mind. Therefore, researchers in metacomputing have developed tools that monitor and forecast how long a request is going to take to run over a particular set of resources (e.g. 61] Today there is no such tool ....

W. Cirne and K. Marzullo, "The computational co-op: gathering clusters into a metacomputer". In Second Merged Symposium IPPS/SPDP 1999, 13th International Parallel Processing Symposium & 10th Symposium on Parallel and Distributed Processing, pp. 160--166, April 1999.

....the scheduler cannot wait for running processes to finish (which is when the amount of resources consumed is known) before performing a new allocation. In fact, non preemptive proportional share scheduling of multiple resources is NP hard, even when the execution times of the processes are known [4]. 4 The ratio r t is called pass in [19] Hence, we provide an algorithm that asymptotically guarantees fairness and bounds how unfair the allocation can be at any point in time. Since how much resource the running processes are going to consume is unknown, we make allocations decisions ....

Walfredo Cirne and Keith Marzullo. The Computational Coop: Gathering Clusters into a Metacomputer. UCSD Technical Report CS99-611, January 1999. www-cse.ucsd.edu/users/walfredo/resume.html#publications

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC