B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurable protocols: Experience with a prototype and an architectural framework. In International Conference on Network Protocols, pages 234--242, San Francisco, CA, October 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Plan 9 approach, threads execute all the way up down the protocol stack in the common case thus, neither functional nor pipelined parallelism are realized. Several researchers have built protocol development frameworks which explicitly support experimentation with parallelization alternatives [25, 44, 65]. In the context of the ASX development environment, Schmidt and Suda [65] explore alternative protocol process architectures on a typical shared memory platform. In their study, the performance of task based and message based architectures are compared with respect to maximum supportable ....

....presentation layer processing. Scalability limitations are not observed for connection level message parallelism in the absence of XDR processing, although in its presence a limit is observed (and conjectured to reflect bus contention) A second protocol framework is presented by Lindgren et al. [44]. In that work, the authors describe a number of experiments with a prototype custom protocol stack on a 32 node BBN Butterfly. The stack, which includes DES encryption, exhibits 8 way functional parallelism (including separate send receive pipelines) Experimental results show that DES can be ....

B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurable protocols: Experience with a prototype and an architectural framework. In International Conference on Network Protocols, pages 234--242, San Francisco, CA, October 1993.

....algorithms implemented as separate modules. Interactions between the modules in this model are predefined and fixed, in contrast with the more flexible possibilities of our model. Several other efforts have concentrated on supportingparallel execution of modular protocols, including [GNI92, LAKS93] While similar to our work in the sense of decomposing protocols along semantic lines, these efforts differ in their emphasis on using parallel execution to improve throughput and latency for high performance scientific applications. They also retain a single level composition model, which we ....

B. Lindgren, M. Ammar, B. Krupczak, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. In Proceedings of the International Conference on Network Protocols, March 1993.

....take advantage of the machines full capabilities. One way to improve performance in the network protocol subsystem to exploit the availability of multiple processors in the host. The use of parallelism in network protocol processing has recently become an active area of research in both academia [5, 12, 20, 21, 22, 35, 46, 47, 48, 59, 62, 63, 64, 69, 70, 71, 72, 74, 75, 76, 77, 87, 88, 89, 95, 100, 106, 107, 108, 109, 111, 112, 116, 117, 124, 125] and industry [18, 37, 42, 45, 49, 68, 90, 94, 110, 120] Many approaches to parallelism in network protocols have been proposed. We provide a brief taxonomy of parallelism in protocols here; more detailed surveys can be found in [12, 48] In general, we attempt to classify approaches by the unit ....

....are to reap the benefits of parallelized networking, they should perform their own ordering. Using either a connectionless protocol such as UDP or a connection oriented protocol such as TCP with multiple connections, an application must be able to handle outof order delivery. Lindgren et al. [74] make a related argument that the parallel application must be tied closely to the parallel communication system. To illustrate the benefits of using multiple connections, we ran send side and receive side experiments of TCP 1 with MCS locks, without ticketing, using 4KB packets with and without ....

Lindgren, B., Krupczak, B., Ammar, M., and Schwan, K. Parallel and configurable protocols: Experience with a prototype and an architectural framework. In Proceedings of the International Conference on Network Protocols, pages 234--242, San Francisco, CA, Oct. 1993.

....entity (in these cases, the process) is much larger than the time required to entirely reload the referenced memory locations into the cache. In this paper, we explore the benefits of affinity scheduling of parallel networking, an area of research which has recently generated considerable interest [3, 4, 6, 12, 13, 15, 17, 18, 20, 23, 25, 26, 27, 35] and one for which affinity scheduling has not yet been examined. Intuitively, parallel networking is a potential candidate for the technique since packets can be individually scheduled and the time to process a packet is relatively short. However, several aspects of the application domain ....

B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurable protocols: Experience with a prototype and an architectural framework. In International Conference on Network Protocols, pages 234--242, San Francisco, CA, October 1993.

....be introduced without sacrificing performance. System V Streams [Rit84] also supports modularization of protocols, but its model is also hierarchical and relatively coarse33 grained. Several other efforts have concentrated on supporting parallel execution of modular protocols, including [GNI92, LAKS93] While similar to our work in the sense of decomposing protocols along semantic lines, these efforts differ in their emphasis on using parallel execution to improve throughput and latency for high performance scientific applications. They also retain a single level composition model, which we ....

B. Lindgren, M. Ammar, B. Krupczak, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. In Proceedings of International Conference on Network Protocols, Mar 1993.

....avoid this socket crossing cost, we replaced the simulated driver with in memory device drivers for the TCP protocol stacks. The drivers emulate a high speed FDDI interface, and support the FDDI maximum transmission unit (MTU) of slightly over 4K bytes. This is similar to the approaches taken in [3, 13, 20, 34]. The drivers act as senders or receivers, producing or consuming packets as fast as possible, to simulate the behavior of simplex data transfer over an error free network. To minimize execution time and experimental perturbation, the receive side drivers use preconstructed packet templates. They ....

B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurableprotocols: Experience with a prototype and an architectural framework. In Proceedings of the International Conference on Network Protocols, pages 234--242, San Francisco, CA, Oct. 1993.

....when compared to our objects multicast rather than individual properties of multicast, for instance and more oriented toward hierarchical composition and limited data sharing. Several other efforts have also concentrated on supporting parallel execution of modular protocols, including [15, 26]. While similar to our work in the sense of decomposing protocols along semantic lines, these efforts differ in their emphasis on using parallel execution to improve throughput and latency for high performance scientific applications. They also retain a single level composition model, which we ....

B. Lindgren, M. Ammar, B. Krupczak, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. Technical Report GIT-CC-93/22, College of Computing, Georgia Institute of Technology, Atlanta, Georgia, March 1993.

....communication by using PT (Pseudo Terminals) to link Streams on two hosts. 2.4. 7 Parallel Protocol Execution The emphasis of this project is on developing modular implementations of protocols to improve performance and configurability of communication support for gigabit second networks [MS93, LAKS93] Faster performance is achieved by executing the individual modules on the individual processing units of a parallel machine. In addition to speed requirements, these protocols send voice and video transmissions with image and data encryption, which presents even further protocol processing ....

....when compared to our objects multicast rather than individual properties of multicast, for instance and more oriented toward hierarchical composition and limited data sharing. Several other efforts have also concentrated on supporting parallel execution of modular protocols, including [GNI92, LAKS93] While similar to our work in the sense of decomposing protocols along semantic lines, these efforts differ in their emphasis on using parallel execution to improve throughput and latency for high performance scientific applications. They also retain a single level composition model, which we ....

B. Lindgren, M. Ammar, B. Krupczak, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. In Proceedings of International Conference on Network Protocols, March 1993.

....and during operation, to maintain good performance in response to such events as failures and changes in request latencies and utilization. A prototype implementation of the system has been constructed that places special emphasis of parallel execution of the underlying communication protocols [LAKS93] In this model, a protocol consists of a set of protocol objects, each of which performs an isolated protocol processing task. Objects communicate with each other by asynchronous invocations implemented as messages delivered via shared memory mailboxes. Protocol objects cooperate in the ....

....are Adaptive, RAID, the group communication framework [Gol92] and most of the configurable operating systems. A good example of this approach is the concurrency control function in RAID, which can be implemented using different predefined or user implemented modules. The object message model [LAKS93] and class hierarchy based models like Arjuna and the configurable mixed media file system [Maf94] can be viewed as somewhat generalized versions of the function based approach. In the object message model, the operation of the system is divided into functions, each of which is implemented by an ....

B. Lindgren, M. Ammar, B. Krupczak, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. Technical Report GIT-CC-93/22, College of Computing, Georgia Institute of Technology, Atlanta, GE, Mar 1993.

....event execution and a second scheduler runs application threads. Second, we will explore additional topics concerning the efficient support and implementation of distributed shared abstractions, including: 1) the customization and parallelization of network protocols for use in DSA implementations[30], 2) the embedding of DSA objects in a more general system for the construction of objectoriented parallel programs[17] including the development of language level and user interface support for object based parallel programming, and (3) the interfacing of Cthreads and the DSA library with ....

Bert Lindgren, Bobby Krupczak, Mostafa Ammar, and Karsten Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. In International Conference on Network Protocols., 1993. TR# GIT-CC-93/22.

....In contrast, the CI may make allocation decisions based on current communication behavior and on actual levels of resource utilization and availability. This auto configuration attribute of the CI distinguishes our research from other current and past work on configurable communication systems [10, 17, 22, 26, 27, 34, 16]. However, note that auto configuration is also used in recent work that deals with dynamic changes in video traffic mapped to the QoS guarantees offered by ATM networks[30] and that it has proven useful in past research on guaranteeing the predictable behavior of real time applications[3, 8] ....

....the application decides how to handle the observed traffic variation. These characteristics make COMM adapt appropriate for servicing complex systems with dynamic and or hard to accurately stated QoS requirements. Related Work. The COMM adapt prototype is loosely based on the previous work of [17] who only comment on the necessity of runtime adjustments in the mapping of protocol modules to processors. COMM adapt differs from previous work on configurable protocols in its support for constructing protocols that are auto configurable at runtime in their functionality and performance. For ....

B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. Georgia Institute of Technology, GIT-CC-9322, 1993.

.... potentially maintaining a uniform kernel interface (e.g. when not adding any new kernel classes) In our past work, we have developed a significant extension of KTK addressing real time applications[9] and we are now developing a second extension addressing configurable communication protocols[15]. However, in this paper, we are focussing on the use of KTK s mechanisms for dynamic program and kernel configuration. The kernel abstraction configured in this paper is a mutex lock. The program abstraction configured in this paper is a global queue. As with other global abstractions in parallel ....

Bert Lindgren, Bobby Krupczak, Mostafa Ammar, and Karsten Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. Technical Report GIT-CC-93/22, College of Computing, Georgia Institute of Technology, March 1993. To appear in the International Conference on Network Protocols,1993.

....number of connections with high throughput and predictable jitter and (2) distributed real time applications, where multiple streams of data are collected and processed subject to specific and possibly dynamic rates and timing requirements. While previous work in configurable communication systems [6, 8, 10, 13, 14, 18] addresses similar application domains, the maintenance of QoS requirements by on line auto configuration of protocols pursued in our research constitutes a novel contribution. The remainder of this paper first outlines the COMM adapt software architecture and its prototype (see Section 2) ....

....dynamically configuring protocol stacks, but no details are provided on the runtime configuration mechanisms. The ADAPTIVE system[14] enables only application level adaptation decisions. The framework presented in [18] supports QoS requirements by connection time configuration decided by the CI. [8] only comment on the necessity of runtime adjustments in the mapping of protocol modules to processors. The parallel STREAMS implementation in [4] offers only user directed stream configuration, and the policy for runtime scheduling of communication related tasks ignores QoS requirements. We ....

B. Lindgren, B. Krupczak, M. Ammar, and K. Schwan. Parallel and configurable protocols: Experiences with a prototype and an architectural framework. Georgia Institute of Technology, GIT-CC-93-22, 1993.

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