James D. Salehi, James F. Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols. Technical Report UM-CS  Home/Search   Document Details and Download   Summary   Related Articles   Check

....conflicts [19, 50] University of Arizona has given special attention to caching for communication, both on the data side [59] and on the instruction side [57] For the instruction side, they investigate in how to use compiler techniques to increase cache performance. Salehi, Kurose and Towsley [66] have studied cache behavior in the context of parallelized communication protocols. Braun and Diot [21] report on, and compare cache hit rates of, an ILP implementation and a corresponding non ILP implementation. The original ILP concept was proposed by Clark and Tennenhouse [27] They and ....

James D. Salehi, James F. Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols. Technical Report UM-CS

....has been done on cache behavior. University of Arizona has given special attention to caching for communication, both on the data side [11] and on the instruction side [10] For the instruction side, they investigate in how to use compiler techniques to increase cache performance. Salehi et al. [14] have studied cache behavior in the context of parallelized communication protocols. The original ILP concept was proposed by Clark and Tennenhouse [5] It has been successfully used to fold together checksum computation with the kernel to user copy in UNIX by, e.g. Partridge and Pink [12] ....

James D. Salehi, James F. Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols. Technical Report UM-CS-1994-075, University of Massachusetts, Amherst, MA 01003, USA, October 1994.

....affinity based scheduling techniques (which have been explored in the context of general parallel programs (i.e. non network application processing) 15, 28, 46, 68, 76] present a promising research opportunity in parallel networking. We have recently completed an extensive study in this area [61, 63, 62]; the results are summarized in Section 3 and several extensions proposed. We also seek to explore the design of parallel networking for communication intensive continuous media ap 1 We use the terms message and packet generally interchangeably. 2 Affinity based scheduling techniques manage ....

.... scheduling of parallel network processing We have completed an initial study of the benefits of affinity based scheduling of parallel network processing, in a setting in which protocol processing executes on the multiprocessor host concurrently with a general workload of nonprotocol activity [61, 63, 62]. The results presented in [63, 62] suggest the tremendous importance of accounting for affinity in multiprocessor protocol scheduling. For the UDP IP FDDI protocol stack in our experiments, the maximum reduction in per message latency ranges from 30 to 50 , providing significant reduction in ....

[Article contains additional citation context not shown here]

James Salehi, James Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols. Technical Report UM-CS-1994-075, University of Massachusetts, Amherst, October 1994. Available via FTP from gaia.cs.umass.edu in pub/Sale94:Scheduling.Z.

....applications. We show instead how to parameterize the analytic model with experimental timing measurements. Finally, in the design of the experiments formulated to yield these measurements, we illustrate an experimental method for isolating the indi 2 These observations lead us to propose in [17] a hybrid approach for a specific class of streams, which offers the best overall performance yielding high message throughput, high intra stream scalability, and robustness in the presence of bursty arrivals. Main Memory system bus FDDI interface network packet P P P P 1 2 3 4 1 T T 2 idle ....

....of affinity based overhead. This paper is organized as follows. We begin by presenting the problem formulation in Section 2. In Section 3 we summarize the salient aspects of the multiprocessor simulation model and the analytic model of message execution time; extensive details are provided in [17]. Section 4 discusses the implementation based experiments performed to measure the parameters needed by the analytic model. Performance results are presented in Section 5. We discuss related work in Section 6, and summarize our work in Section 7. 2 Problem formulation 2.1 Affinity scheduling of ....

[Article contains additional citation context not shown here]

J. Salehi, J. Kurose and D. Towsley. "Scheduling for Cache Affinity in Parallelized Communication Protocols". TRUMCS -1994-075, U. Massachusetts, Oct. 1994. (available via ftp from gaia.cs.umass.edu in pub/Sale94:Scheduling.ps.Z)

....parallelism is generally necessary on multiprocessor machines, since the alternative would restrict aggregate network access to the bandwidth capacity of a single processor. While previous studies have explored affinity based scheduling of non network related application processing, our work [30, 31] is the first to apply the technique to operating system network processing. To illustrate the motivation for considering affinity based scheduling here, consider the following. In our experimental environment (consisting of a parallelized x kernel [12, 25] running on a Silicon Graphics ....

....Given that processor speeds have tended to increase much faster than memory access times (a trend that is likely to continue [9, 21, 37, 39] this empirical observation suggests that affinity based scheduling represents a promising research opportunity in parallel networking. In our earlier work [30, 31], we evaluated the benefits of affinity based scheduling of receive side UDP IP FDDI processing, for a shared memory multiprocessor concurrently executing a background workload of non protocol activity. We found that the scheduling technique can significantly reduce packet latency, and in some ....

James Salehi, James Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols (extended abstract). In ACM International Conference on Measurement and Modeling of Computer Systems (ACM SIGMETRICS), pages 311-- 312, Ottawa, Canada, May 1995.

....24, 1995 Abstract In this paper, we present further results in processor cache affinity scheduling of parallel network protocol processing, in a setting in which protocol processing executes on the multiprocessor host concurrently with a general workload of non protocol activity. In earlier work [31, 32] we evaluated affinity based scheduling of receive side protocol processing under two parallelization approaches: Locking and Independent Protocol Stacks (IPS) In this work, we i) evaluate affinity based scheduling of send side UDP IP FDDI processing, ii) examine the performance of ....

....operating system both within and among connections can both increase the bandwidth and decrease the latency of multiprocessor communication. While previous studies have explored the benefits of affinity based scheduling of non network related application processing [7, 10, 22, 40] our work [31, 32, 33] is the first to apply the technique to operating system network processing. In our experimental environment (consisting of a parallelized x kernel [13, 26] running in user space on an 8 processor MIPS R4400 based SGI Challenge XL) packet execution times can vary by as much as a factor of four, ....

[Article contains additional citation context not shown here]

James Salehi, James Kurose, and Don Towsley. Scheduling for cache affinity in parallelized communication protocols. Technical Report UM-CS-1994-075, University of Massachusetts, Amherst, October 1994. Available via FTP from gaia.cs.umass.edu in pub/Sale94:Scheduling.Z.

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