K. Connelly and A. A. Chien. FM-QoS: Real-time Communication using Self-synchronizing Schedules. In Proc. of SC97: High Performance Networking and Computing, November 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of QoS in clusters. Low cost packet scheduling and queuing algorithms for high performance networks are described in [17] In [18] Gerla et al. describe three schemes for providing QoS in wormhole routed networks. However, these require modifications to switches or extra NICs on hosts. FM QoS [19] provides predictable performance by devising conflict free communication schedules. In [20] and [21] the authors propose new hardware or hardware software infrastructures to support QoS. The Globus project s GARA framework [22,23] provides end to end QoS for different types of resources. 8. ....

K. Connelly, A. Chien. FM-QoS: Real-time communication using self-synchronization schedules. In Proc. Of Supercomputing conference, San Jose, CA, November 1997.

....communication schedules to avoid resource con icts for network links and outputs. Elimination of such resource con icts leads to predictable communication performance. FM QoS uses a Petri net model to characterize the structure of the self organizing schedules and to tolerate the clock drifts [26]. 3 Problem Description All the existing variations of the parametric time based scheduling method are based on a single node model. They mainly focus on non preemptive periodic aperiodic hard real time tasks with inter task relative timing constraints. Our basic objective is to extend the ....

Kay Connelly and Andrew A. Chien. FM-QoS: Real-time communication using self-synchronizing schedules. In High Performance Networking and Computing, San Jose, California, pages 15-21, November 1997. ACM SIGARCH and IEEE-CS-TCCA.

....The primary challenge associated with using Myrinet for real time communication is to ensure bounded fall through times across the worm hole switches. Myrinet worm hole switches have predictably low latencies, that are less than a micro second only in the absence of switch contention. FM QoS [17] is a messaging layer that overcomes switch contention by building a global view of time among LANai interfaces [8] in the Myrinet network. FM QoS uses feedback from the actual latency incurred by specially introduced traffic, to correct for drifts among Real Time Clocks (RTCs) on the individual ....

....ensure conflict free link schedules without requiring a second level of link scheduling at the MCP. In other words, contention of link bandwidth and Myrinet worm hole switches is eliminated without explicit synchronization among the MCPs. In this sense, our approach is orthogonal to that of FM QoS [17] which aims to achieve LANai level synchronization. 7 4.4 Priority Inversion Priority inversion is minimized by migrating most of the protocol processing activity into the tasks context. Sending tasks marshal their messages to LANai memory, and receiving tasks transfer messages from LANai ....

Connelly, K. and A. Chien. FM-QoS: Real-time communication using self-synchronizing schedules. In Proceedings of Supercomputing 1998.

....Myrinet switches. So far, Myrinet has been primarily used for highperformance communication. Popular messaging layers such as GM [12] FM [13] and BIP [14] are designed for low latency and high bandwidth, with no particular support for Quality of Service or real time communication. FM QoS [6] is the only known messaging layer providing predictably low latencies between network interfaces across a Myrinet network of which we are aware. 3 Myrinet worm hole switches have predictably low fall through latencies of the order of 500 ns [4] However, the presence of contending traffic at a ....

....low latencies between network interfaces across a Myrinet network of which we are aware. 3 Myrinet worm hole switches have predictably low fall through latencies of the order of 500 ns [4] However, the presence of contending traffic at a switch can cause unacceptably high jitters in latency [6]. Thus the key to enabling predictable communication across a worm hole routed network such as Myrinet lies in generating conflict free schedules. This directly necessitates building a global view of time across the network. This global view of time has to be sufficiently fine grained and accurate ....

K.Connelly,and A. Chien.. "FM-QoS: Real-time communication using self-synchronizing schedules" in Proceedings of Supercomputing, 1998.

.... software layering, which resulted in 90 of the FM performance being delivered to the applications [27] FM has been used as a testbed for research on novel mechanisms for coscheduling on network of workstations [36] and on the introduction of QoS guarantees in a wormhole routing interconnect [11]. HPVM is available for download on our web site at http: www csag.ucsd.edu. 3 The Machine: HPVM III Cluster Software The software is based on the Illinois HPVM technology, which includes Illinois Fast Messages 3 [31] to deliver high bandwidth (92MB s) with low overhead (a few seconds) and ....

Kay Connelly and Andrew A. Chien. FM-QoS: Real-time communication using self-synchronizing schedules. In Proceedings of SC97, November 1997. Available from http://www.supercomp.org/sc97/program/TECH/ CONNELLY/INDEX.HTM.

....the entire system dramatically easier to install, manage, and use. Fast Messages and HPVM have also been used as a testbed for research on novel mechanisms for co scheduling on a network of workstations [19] and for research on the introduction of QoS guarantees in a wormhole routing interconnect [6]. HPVM is available for download on our Web site at http: www csag.ucsd.edu .There mainder of this paper summarizes the improvements to the HPVM system in HPVM 1.9. 3 HPVM Low Level Design Review Overall, the structure of FM is straightforward. Packets must traverse a pipeline across ....

K. Connelly and A. A. Chien. FM-QoS: Real-time communication using selfsynchronizing schedules. In SC97 Proceedings, November 1997. 4

....at the granularity of two kilobyte packets and can preserve this synchronization at a traffic overhead of less than 1 . More specifically, delivery can be guaranteed with a latency of approximately 20 microseconds, application to application. For more information on this approach see [2]. 5.3 Perspective Compared to both virtual circuit and physical circuit approaches, global scheduling reduces the hardware requirements for network switches. Simplicity may enable faster or cheaper switches, and further requiring no additional hardware 5 The prototype for FM QoS is being built ....

Kay Connelly and Andrew Chien. FM-QoS: Real-time communication using selfsynchronizing schedules. In Proceedings of Supercomputing Conference, San Jose, CA, November 1997.

No context found.

K. Connelly and A. A. Chien. FM-QoS: Real-time Communication using Self-synchronizing Schedules. In Proc. of SC97: High Performance Networking and Computing, November 1997.

No context found.

K. Connelly and A. A. Chien, "FM-QoS: Real-Time Communication Using Self-Synchronizing Schedules," in Proceedings of Supercomputing Conference, November 1997.

No context found.

K. Connelly and A. A. Chien. FM-QoS: Real-Time Communication Using Self-Synchronizing Schedules. In Proc. of Supercomputing Conf. (SC'97), November 1997.

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