B. Verghese, A. Gupta, and M. Rosenblum. Performance isolation: sharing and isolation in shared-memory multiprocessors. In Proc. 8th International Conference on Architectural Support for Programming Languages and Operating Systems, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....bandwidth that a single http client or virtual host server gets, both techniques can be used to achieve similar ends. Also, many have studied general quality of service (QoS) support for server applications. The reservation domains of Eclipse [5] the Software Performance Units of Verghese et al. [20], and Resource Containers [4] can group a set of processes or threads as a unit for resource scheduling. This is similar to our job classes. The Nemesis kernel [15] also provides QoS with rate based real time scheduling for I O as well as CPU. However, those systems are integrated deep into the ....

B. Verghese, A. Gupta, and M. Rosenblum, "Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors," ASPLOS. Oct. 1998, San Jose, CA, pp. 181-192.

....is needed because multiple I O channels may belong to a single application and it would be more logical to charge the resource usage to the application as a whole. There have been attempts to address the issue of resource accounting in a shared memory multi processing environment. Verghese et al. [8] propose abstracting system resources into a logical entity known as a system performance unit (SPU) SPUs associate a resource to a group of processes that provide a single service. For each service, the CPU usage time is measured by counting the number of time slots the service requires in a ....

Ben Verghese, Anoop Gupta, and Mendel Rosenblum. Performance isolation: Sharing and isolation in shared-memory multiprocessors. In Proceedings of 8th International Conference on Architectural Support for Programming Languages and Operating Systems (APLOS-VIII), pages 181-182, San Jose, California, October 1998.

....Layer 5 6 7 switches such as those from Alteon [2] Cisco [10] Foundry [15] at most support only content aware and load balancing request dispatching to a Web server cluster, but not Web server QoS. Several research e orts focused on QoS guarantee for speci c type of resource, such as CPU [25, 6, 16, 29, 8, 19, 17], disk [20, 9, 28] and on shared network link [21, 24, 14] Almost all of these systems can be abstracted into the same implementation framework as described in this paper. In particular, one can show the uid fair queuing model [27] provides the theoretical basis for all these resource ....

B. Verghese, A. Gupta, and M. Rosenblum. \Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors." In Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

....not the framework, should determine how tickets are adjusted among entities. These issues have yet to be considered, although other improvements, such as the removal of the upper and lower limits imposed by currencies [Sullivan and Seltzer, 2000] have been made by subsequent research. 3. 7 IRIX [Verghese et al. 1998] presents a performance isolation model for managing resources in the IRIX operating system. It partitions resources into isolated units called Software Performance Units (SPUs) Processes in the same SPU contend for resources allocated to the unit but do not experience any performance degradation ....

Verghese, B., Gupta, A., and Rosenblum, M. (1998). Performance isolation: Sharing and isolation in shared-memory multiprocessors. In Proceedings of the 8th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) , pages 181-192, San Jose, Ca, USA.

.... attack on the other users on the system, intentionally or unintentionally, using a fork bomb, for example, where as many processes as possible are started up as fast as possible (essentially, each process continuously starts up copies of itself) One method for isolating entities is proposed in [VGR98] This method uses Software Performance Units (SPU) as the unit of isolation. A thread that belongs to one SPU will not be e ected by the resource usage of a thread in another SPU (although thread is no protection between threads in the same SPU) In this system a SPU is allocated an integer ....

Ben Verghese, Anoop Gupta, and Mendel Rosenblum. Performance isolation: sharing and isolation in shared-memory multiprocessors. In asplos, pages 181-192. acm, 1998.

....switches such as those from Alteon [2] Cisco [10] Foundry [15] at most support only content aware and loadbalancing request dispatching to a Web server cluster, but not Web server QoS. Several research efforts focused on QoS guarantee for specific type of resource, such as CPU [25] 6] 16] [29], 8] 19] 17] disk [20] 9] 28] and on shared network link [21] 24] 14] Almost all of these systems can be abstracted into the same implementation framework as described in this paper. In particular, one can show the fluid fair queuing model [27] provides the theoretical basis for ....

B. Verghese, A. Gupta, and M. Rosenblum. "Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors." In Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

....the flamework, should determine how tickets are adjusted among entities. These issues have yet to be considered, although other improvements, such as the removal of the upper and lower limits imposed by cur rencies [Sullivan and Seltzer, 2000] have been made by subsequent research. 3. 7 IRIX [Verghese et al. 1998] presents a performance isola tion model for managing resources in the IRIX operating system. It partitions resources into isolated units called Software Performance Units (SPUs) Pro cesses in the same SPU contend for resources allo cated to the unit but do not experience any performance ....

Verghese, B., Gupta, A., and Rosenblum, M. (1998). Performance isolation: Sharing and isolation in shaxed-memory multiprocessors. In Proceedings of the 8th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) , pages 181-192, San Jose, Ca, USA. ACM.

....topologically close to the processes they belong to. The DSM property calls for memory allocation schemes that exploit Memory Locality Domains, i.e. clusters of neighboring memory and processor modules, where computations execute in isolation from other computations running on different modules [19]. Memory isolation schemes are not always sufficient for parallel programs that utilize large numbers of possibly distanced processor and memory modules, programs with dynamic and unpredictable memory access patterns, or even programs that migrate between processors due to the operating system ....

B. Verghese, A. Gupta and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared Memory Multiprocessors. Proc. of the 8th International Conference on Architectural Support for Programming Languages and Operating Systems, pp. 181--192, San Jose (USA), 1998.

....Work The growing popularity of the multimedia applications has resulted in several research efforts that have focused on the design of predictable resource allocation mechanisms. Consequently, in the recent past, several techniques have been proposed for the predictable allocation of processor [8, 10, 12, 17, 19, 20, 25, 26, 28, 29, 31], network interface [2, 4, 6, 9, 11, 23] and disk [1, 14, 30] bandwidth. While each effort differs in the exact mechanism employed to provide predictable performance (e.g. admission control, rate based allocation, fair queuing) the broad goals are similar add quality of service support to an ....

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in SharedMemory Multiprocessors. In Proceedings of ASPLOS-VIII, San Jose, CA, pages 181-192, October 1998.

....processors. Frequent repartitioning can be expensive; doing so infrequently can result in imbalances (and unfairness) across partitions. While fairness may be elusive (or expensive) in such an approach, it has, nevertheless, been successfully employed to isolate applications from one another [1, 10, 29]. In summary, GPS based fair scheduling algorithms or simple modifications thereof are unsuitable for fair allocation of resources in multiprocessor environments. To overcome this limitation, we propose a CPU scheduling algorithm for multiprocessors that: i) explicitly distinguishes between ....

....Resource containers [2] define a new abstraction to account for resource usage; our effort is complementary since we focus on resource allocation, while they focus on accounting for resource usage. Finally, application isolation and resource allocation in clustered environments has been studied in [1, 10, 29]. 7 Concluding Remarks In this paper, we argued that the infeasibility of certain weight assignments causes unfairness or starvation in many existing proportional share schedulers when employed for multiprocessor servers. We presented a novel weight readjustment algorithm to translate infeasible ....

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors. In Proceedings of ASPLOS-VIII, San Jose, CA, pages 181--192, October 1998.

.... In order to cope with groups of resources that must be allocated in a global and interrelationship aware way, several research activities currently focus on global resource allocation schemes and exploit new abstractions for resource principals such as process groups [LMB 96, BGzS98, VGR98] or resource containers [BMD99] Knowledge Sharing The actual resource requiremenst of multimedia applications are hard to predict. The most important factor of in uence is the bitrate produced by codecs, which, using variable bitrate video codecs rises dramatically when scenes change or have ....

B. Vergehese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared Memory Multiprocessors. In Proceedings of the 8th International Conference on Architectural Support for Programming Languages and Operating Systems, October 1998.

....it can enable some seek reduction, by slightly relaxing the timescale on whichit operates. This allows the scheduler to service # requests for the rst process for every 2# requests for the second, where each set mightcontain sequential requests. One variation on this theme is suggested in [29], where the scheduler picks from processes with virtual clocks between ######## and ######## # (where # is a relaxation threshold, and could be 1 second) Among these, it chooses the request with the smallest positioning time. Weproposeacombination heuristic for this scheduler, thus hinting at ....

....of webservers [3] It is important for these disk schedulers to adhere to their contract; anticipatory scheduling facilitates this for applications issuing synchronous I O. In practice, proportional share disk schedulers will almost always be deployed in combination with a seek reducing scheduler [29]. Our experiments have demonstrated how the combination heuristic brings about simultaneous improvement of both contract adherence and performance. Real time disk schedulers (either pure or in combination with seek reducing schedulers) are commonly used to serve and view multimedia content [9, ....

B. Verghese, A. Gupta, and M. Rosenblum. Performance isolation: Sharing and isolation in shared memory multiprocessors. In ASPLOS, Oct. 1998.

....break down due to aggressive clients. It is not a complete QoS solution bacouse it only takes into account bandwidth resource which does not necessiraly reflect the real system resource consumption. Several research efforts focused on QoS guarantee for specific type of resource, such as CPU[25, 6, 16, 29, 8, 19, 17], disk[20, 9, 28] and on shared network link[21, 24, 14] Almost all of these systems can be abstracted into the same implementation frmework as described in this thesis. In particular, one can show the fluid fair queuing model[27] provides the theoretical bases for all these resource schedulers. ....

{SPU} B. Verghese, A. Gupta, and M. Rosenblum. "Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors." In Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

....it can enable some seek reduction, by slightly relaxing the timescale on which it operates. This allows the scheduler to service n requests for the rst process for every 2n requests for the second, where each set might contain sequential requests. One variation on this theme is suggested in [29], where the scheduler picks from processes with virtual clocks between minclock and minclock (where is a relaxation threshold, and could be 1 second) Among these, it chooses the request with the smallest positioning time. We propose a combination heuristic for this scheduler, thus hinting ....

....of webservers [3] It is important for these disk schedulers to adhere to their contract; anticipatory scheduling facilitates this for applications issuing synchronous I O. In practice, proportional share disk schedulers will almost always be deployed in combination with a seek reducing scheduler [29]. Our experiments have demonstrated how the combination heuristic brings about simultaneous improvement of both contract adherence and performance. Real time disk schedulers (either pure or in combination with seek reducing schedulers) are commonly used to serve and view multimedia content [9, ....

B. Verghese, A. Gupta, and M. Rosenblum. Performance isolation: Sharing and isolation in shared memory multiprocessors. In ASPLOS, Oct. 1998.

....to a local replica of stored content, as described in 5.1. This is the only support for controlling I O resources in our current prototype. Our approach could extend to manage multiple resources given support in the server node OS for enforcing assignments of disk, memory, and network bandwidth [44, 40, 30]. Economic problems involving multiple complementary goods are often intractable, but the problem is simplified in Muse because customer utility functions specify value indirectly in terms of delivered performance; the resource allotment to achieve that performance need not be visible to ....

....provisioning choices securely and with out burdening server applications. Decoupling resource principals from the kernel s process and thread abstractions improves generality: examples in the literature include resource groups in Opal [14] activities in Rialto [24] Software Performance Units [44], scheduling domains in Nemesis [30] and the resource containers [10] implemented for FreeBSD by Mohit Aron [8] and used in our prototype. Aron has shown how to extend resource principals to cluster reserves spanning nodes [9, 8] Scheduling for performance isolation. To ensure performance ....

[Article contains additional citation context not shown here]

Ben Verghese, Anoop Gupta, and Mendel Rosenblum. Performance Isolation: Sharing and Isolation in Shared Memory Multiprocessors. In Proceedings of the 8th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), October 1998.

....MOSIX [4] provides load balancing and preemptive migration for traditional UNIX processes. Chowdhury et al. 5] characterized when to reconfigure sequential workloads. DEMOS MP [14] Accent [23] Locus [20] and V [19] all provided manual or semi automated migration of processes. Verghese et. al [21] proposed a way to isolate the performance of applications running on an SMP system. While their approach requires changes to similar parts of the operating system, their primary goal was to increase fairness to all applications, while our goal is to create an inherently unfair priority level for ....

....the same test but replacing the Ethernet with Myrinet for the CVM application) As a result, we plan to implement this feature in the near future. In particular, we are likely to employ the network priority system proposed by Druschel and Peterson [6] and the I O priority policy proposed in [21]. It is possible to further enhance the virtual memory system to increase the speed at which pages are reclaimed from the guest processes by the host processes. In particular, dirty guest pages require writing back to the swap device before they can be allocated to the host process. One extension ....

B. Verghese, A. Gupta, and M. Rosenblum, "Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors, " ASPLOS. Oct. 1998, San Jose, CA, pp. 181-192.

....Work The growing popularity of the multimedia applications has resulted in several research efforts that have focused on the design of predictable resource allocation mechanisms. Consequently, in the recent past, several techniques have been proposed for the predictable allocation of processor [7, 9, 14, 21, 22], network interface [3, 5, 8, 18] and disk [1, 11, 23] bandwidth. While each effort differs in the exact mechanism employed to provide predictable performance (e.g. admission control, rate based allocation, fair queuing) the broad goals are similar add quality of service support to an ....

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors. In Proceedings of ASPLOS-VIII, San Jose, CA, pages 181--192, October 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance isolation: sharing and isolation in shared-memory multiprocessors. In Proc. 8th International Conference on Architectural Support for Programming Languages and Operating Systems, 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance isolation: sharing and isolation in shared-memory multiprocessors. In Proceedings of 8th International Conference on Architectural Support for Programming Languages and Operating Systems, 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in SharedMemory Multiprocessors. In Proc. of the ACM 8th Intl. Conf. on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors. In Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors. In Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, San Jose, CA, Oct. 1998.

No context found.

Ben Verghese, Anoop Gupta, and Mendel Rosenblum. Performance isolation: Sharing and isolation in shared-memory multiprocessors. In Proceedings of 8th International Conference on Architectural Support for Programming Languages and Operating Systems (APLOS-VIII), pages 181--182, San Jose, California, October 1998.

No context found.

Ben Verghese, Anoop Gupta, and Mendel Rosenblum. Performance isolation: Sharing and isolation in shared-memory multiprocessors. In 8th International Conference on Architectural Support for Programming Languages and Operating Systems, pages 181--192, San Jose, California, October 1998.

No context found.

B. Verghese, A. Gupta, and M. Rosenblum. Performance Isolation: Sharing and Isolation in Shared-Memory Multiprocessors. In Proc. 8th Intl. Conf. on Architectural Support for Programming Languages and Operating Systems, Oct. 1998.

First 50 documents

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