J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedings of the Symposium on CommunicationsArchitectures and Protocols (SIGCOMM), (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....mechanisms that serialize access to shared objects (such as messages, message queues, protocol connection records, and demultiplexing tables) used when processing protocols in parallel. A number of process architectures have been proposed as the basis for parallelizing communication subsystems [1, 2, 3, 4]. There are two fundamental types of process architectures: task based and message based. Task based process architectures are formed by binding one or more PEs to units of protocol functionality (such as presentation layer formatting or transport layer segmentation reassembly, acknowledgment ....

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedings of the Symposium on CommunicationsArchitectures and Protocols (SIGCOMM), (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....up, contention in proxies quickly degrades the overall application performance. Parallel protocol stack processing has not been carefully studied for clustered multiprocessors that target parallel applications. Results from researchers often assume that networks are reliable. Related work ( 6] [8], and [5] have concentrated on increasing throughput rather than reducing contention to reduce latency. Commodity LAN technologies are getting faster. Welsh et al. 14] have studied how to provide fast user level communication on Fast Ethernet and ATM. Their results are impressive, but to ....

N. Jain, M. Schwartz, and T.R. Bashkow. Transport Protocol Processing at Gbps Rates. In SIGCOMM, pages 188--199, Philadelphia, PA, September 1990.

....corresponds to the natural protocol layering abstraction. However, it appears that the overheads and specialized hardware requirements preclude these approaches in a general purpose (RISC based) end host environment. Instead, many studies have considered parallelism among distinct messages [6, 20, 25, 26, 31, 34, 35, 49, 55, 64, 65, 66]. While this approach does not reduce latency for individual packets, it does increase throughput and decrease latency within and among sessions. We now turn our attention to these studies. In the first study of message level parallelism, Jain, Schwartz and Bashkow consider the design of a ....

....and decrease latency within and among sessions. We now turn our attention to these studies. In the first study of message level parallelism, Jain, Schwartz and Bashkow consider the design of a dedicated front end multiprocessor performing message level parallel computation of OSI level 4 [35]. Lower layers are handled by dedicated channels, both on input and output. The authors decompose the transport level protocol functionality, identify data structures involved (focusing on context records and an ACK queue ) and discuss techniques for reducing contention and preserving ordering. ....

Niraj Jain, Mischa Schwartz, and Theordore R. Bashkow. Transport protocol processing at Gbps rates. In Proceedings of the ACM SIGCOMM Conference on Communications, Architectures, Protocols and Applications, pages 188--199, Philadelphia, PA, September 1990. ACM. 38

....regardless of their connection or where they are in the protocol stack, achieving speedup both with multiple connections and within a single connection. The disadvantage is that it requires locking shared state, most notably the protocol state at each layer. Systems using this approach include [5, 24, 29, 31, 50]. A set of connections forms the unit of concurrency in connection level parallelism [21, 60, 65, 67] Speedup is achieved using multiple connections, which can potentially be processed in parallel. The advantage of this approach is that it exploits the natural concurrency between connections. ....

Jain, N., Schwartz, M., and Bashkow, T. R. Transport protocol processing at Gbps rates. In SIGCOMM Symposium on Communications Architectures and Protocols, pages 188--199, Philadelphia, PA, Sept. 1990. ACM.

....greater overlap between useful computation and communication processing. 31 Multiprocessor front ends: Multiprocessor front ends may be designed using specialpurpose or general purpose hardware. Special purpose designs facilitate efficient interaction with the host and the network interface unit [82, 83], while general purpose designs must explicitly coordinate accesses to these interfaces [14, 138, 182] More processing power in the front end can improve the quality of service provided to the applications, by reducing queuing delays within the communication subsystem. Network interfaces for ....

....in Chapter 3 from the proposed architectural approaches to manage communication parallelism, as outlined below. With the advent of high speed networks, uniprocessor protocol implementations (on the host or the network interface) are often unable to keep up with, or fully utilize, the network [83], resulting in performance bottlenecks within the communication subsystem. This performance degradation is all the more pronounced for multiprocessor hosts due to a higher rate of message generation and consumption. The pursuit of application level gigabits per second has prompted recent efforts ....

[Article contains additional citation context not shown here]

N. Jain, M. Schwartz, and T. R. Bashkow, "Transport protocol processing at GBPS rates," in Proc. of ACM SIGCOMM, pp. 188--199, September 1990.

....CPUs often penalize interrupt driven network communication by exhibiting high context switching overhead. This overhead results from the cost of flushing pipelines, storing and retrieving register windows, and invalidating instruction data caches and virtual memory translation lookaside buffers [5]. Ameliorating the throughput preservation problem and providing very high data transfer rates to applications and protocols requires an integrated solution that involves redesigning transport system architectures [6] To help system developers and researchers navigate through the transport ....

....flow up and down between protocol layers. ffl Vertical Process Architectures: Figure 3 (2) illustrates a vertical process architecture that implements the same protocol graph as Figure 3 (1) In this vertical approach, a separate process is associated with each incoming and outgoing message [2, 5]. Each process escorts its message through the protocol graph within a single address space, delivering the message down to a network interface or up to an application. Note that there are other ways to organize vertical process architectures such as connectional parallelism shown in Figure ....

[Article contains additional citation context not shown here]

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in SIGCOMM Symposium on Communications Architectures and Protocols, (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....operating system architectures that are predominant are based on either monolithic or micro kernel architectures. Numerous methods of exploiting the parallelism of multiple processor machines and their limitations operating under monolithic operating systems have been reported in the literature [8, 6, 14]. In depth analysis of protocol implementations in uniprocessor machines operating under both monolithic and micro kernel architectures have been also reported [11, 2, 7, 10, 9, 12] Finally, a generic technique, referred to as application layer framing (ALF) and integrated layer processing (ILP) ....

N. Jain, M. Schwartz, and T. Bashlow. Transport protocol processing at Gbps rates. ACM SIGCOMM'90, pages 188--208, 1990.

....does not by itself change drastically the speed of the bicycles. 2. 1 How to Improve From the observations made in the second half of the eighties about the transport performance on top of LANs, detailed studies of mechanisms and implementations have been done [CJR 89] GKW 89] Mei 91] JSB 90] Zitt 91] and ended up in two different schools of thought. The first school advocated the design of new transport protocols such as NETBLT [CLZ 87] VMTP [ChW 89 ] SNR NRS 90] and XTP [Ches 89] Wha 89] PE 91] The second school is putting the burden of the poor transport performance to ....

.... protocol mechanism such as buffer handling and interaction with the operating system [CJR 89] MaB 91] There exist of course many choices of implementation from VLSI to pure software, more or less integrated in the operating system, using additional dedicated computing power or not [GKW 89] JSB 90] Zitt 91] The syntaxes of the PDU are not neutral with respect to the implementation results. Variable fields may save bandwidth but are more difficult to siliconize . The place of the CRC field in the PDU is also important if hardware implementation with insertion and removal on the fly ....

Jain N., Schwartz M., Bashkow T.R., Transport Protocol Processing at GBPS Rates, SIGCOMM 90, Philadelphia, Sep. 24-27, 1990, Computer Communications Review, 1990, Vol. 20, nr 4, pp. 188-199

....highway does not by itself change drastically the speed of the bicycles. How to Improve From the observations made in the second half of the eighties about the transport performance on top of LANs, detailed studies of mechanisms and implementations have been done [CJR 89] GKW 89] Mei 91] JSB 90] Zitt 91] and ended up in two different schools of thought. The first school advocated the design of new transport protocols such as NETBLT [CLZ 87] VMTP [Che 88] ChW 89 ] SNR [NRS 90] and XTP [Ches 89] Wha 89] PE 91] The second school is putting the burden of the poor transport ....

.... protocol mechanism such as buffer handling and interaction with the operating system [CJR 89] MaB 91] There exist of course many choices of implementation from VLSI to pure software, more or less integrated in the operating system, using additional dedicated computing power or not [GKW 89] JSB 90] Zitt 91] The syntaxes of the PDU are not neutral with respect to the implementation results. Variable fields may save bandwidth but are more difficult to siliconize . The place of the CRC field in the PDU is also important if hardware implementation with insertion and removal on the fly ....

JAIN N., SCHWARTZ M., BASHKOW T.R., Transport Protocol Processing at GBPS 22 Rates, SIGCOMM 90, Philadelphia, Sep. 24-27, 1990, Computer Communications Review, 1990, Vol. 20, nr 4, pp. 188-199

....does not by itself change drastically the speed of the bicycles. 2. 1 How to improve From the observations made in the second half of the eighties about the transport performance on top of LANs, detailed studies of mechanisms and implementations have been done [CJR 89] GKW 89] Mei 91] JSB 90] Zit 91] and ended up in two different schools of thought. The first school advocated the design of new transport protocols such as NETBLT [CLZ 87] VMTP [ChW 89 ] SNR [NRS 90] and XTP [Ches 89] Wha 89] PE 92] The second school is putting the burden of the poor transport performance to ....

.... protocol mechanism such as buffer handling and interaction with the operating system [CJR 89] MaB 91] There exist of course many choices of implementation from VLSI to pure software, more or less integrated in the operating system, using additional dedicated computing power or not [GKW 89] JSB 90] Zit 91] The syntaxes of the PDU are not neutral with respect to the implementation results. Variable fields may save bandwidth and provide flexibility for extension but they are more difficult to siliconize . The place of the CRC field in the PDU is also important if hardware ....

JAIN N., SCHWARTZ M., BASHKOW T.R., Transport Protocol Processing at GBPS Rates, SIGCOMM 90, Philadelphia, Sep. 24-27, 1990, Computer Communications Review, 1990, Vol. 20, nr 4, pp. 188-199

....have been a drastic improvement of the network performance in terms of access data rate as well as in terms of bit error rate and packet error rate. In depth studies of mechanisms and implementations regarding the transport performance on top of LANs have been done [CJR 89] GKW 89] Mei 91] JSB 90] Zit 91] and have ended up in two different schools of thought. The first school advocated the design of new transport protocols such as NETBLT [CLZ 87] VMTP [ChW 89] SNR [NRS 90] and XTP [Che 89] Wha 89] PEI 92] The second school is putting the burden of the poor transport performance on ....

JAIN N., SCHWARTZ M., BASHKOW T.R., Transport Protocol Processing at GBPS Rates, SIGCOMM 90, Philadelphia, September 24-27, 1990, in: Computer Communications Review, 1990, vol. 20, nr 4, pp. 188-199.

....by multiple instruction sets, running on multiple processors. Three levels of parallelism are generally identified for a communication process, which are given by the degree of granularity (i.e. coarse grain, medium grain and fine grain) used when introducing parallelism into protocol processing [JSB90], Haa91] Zit91] GNI92] 26 Typically, coarse grain processing involves high level functions or tasks assigned to individual processors such as, for instance, in the case of connection dedicated to processor approach. Despite its simplicity (which is advantageous for opening up the ....

....it through the (transport) protocol stack (Fig. 2 3) This is similar in fact to increasing the processing power (for transport protocol) in the above mentioned layerpipeline model. Performance improvements are obtained, regardless of traffic intensity, pattern or distribution over connections [JSB90], GNI92] TW92] However, careful implementation of the protocol is required because of the inherent sequentiality of packets in the data stream. Sophisticated solutions are required to cope with diverse error and fault conditions [JSB90] Also, a serious drawback of this method is given by the ....

[Article contains additional citation context not shown here]

Jain, N., Schwartz, M. and Bashkow, T.R.,"Transport Protocol Processing at Gbps Rates," in Proceedings of ACM SIGCOMM'90, Philadelphia, Pennsylvania, pp. 188 - 199, September 1990.

....notwithstanding any copyright notation thereon. low error rate communications [4] Parallelism can be exploited in either of these approaches. Multiple, separate experiences in applying parallelism to communication protocols have indicated good performance improvements of network transport systems[14, 11, 7, 9, 5, 6, 8, 1, 2, 3, 12]. As shared memory multiprocessor desktop workstations and server platforms are more commonplace, the use of parallelization at the network nodes is a real option. Parallelization also could be used in conjunction with other approaches to dealing with the communication bottlenecks. This paper ....

Niraj Jain, Mischa Schwartz, and Theodore R. Bashkow. Transport protocol processing at GBPS rates. In ACM SIGCOMM, pages 188--199, Philadelphia, 1990.

....as memory and process management) and hardware devices (such as highspeed network controllers) to support diverse applications running on diverse local, metropolitan, and wide area networks. Application performance is significantly affected by the process architecture of the transport system [1, 2, 3]. A process architecture binds certain communication protocol entities (such as layers, tasks, connections, and or messages) together with logical and or physical processing elements. This paper describes a flexible framework called ADAPTIVE that supports, among other things, development and ....

....and (4) Message Parallelism, which associates a process per message. The ADAPTIVE framework also facilitates experimentation with the various process architecture alternatives. Several studies have compared the advantages and disadvantages of these process architectures via qualitative analysis [2, 12]. However, few studies have quantitatively compared the performance of the alternative process architectures via controlled, empirical experimentation. In particular, existing research that measures the performance of process architectures focuses on only one or two approaches [7, 9, 13, 14] ....

[Article contains additional citation context not shown here]

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedings of the Symposium on Communications Architectures and Protocols (SIGCOMM) , (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....in Section 4. Opportunities for Parallelism. The simple protocol described above offers the following opportunities for parallelism: ffl Jain et al. demonstrated conceptually that multiple packets in the transmission window can be processed independently to significantly increase throughput[Jain 90]. This can be done without significant changes to our current protocol implementation. ffl It is intuitive that multiple connections can be processed in parallel, since there are no direct dependencies between different connections. Therefore, the different threads associated with objects could ....

Niraj Jain, Mischa Schwartz and Theodore Bashkow, "Transport Protocol Processing at GBPS Rates," ACM Computer Communications Review, Vol. 20 (September 1990), pp. 188-199.

....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 ....

Jain, N., Schwartz, M., and Bashkow, T. R. Transport protocol processing at Gbps rates. In ACM SIGCOMM Symposium on Communications Architectures and Protocols, pages 188--199, Philadelphia, PA, Sept. 1990. ACM.

....both increase the bandwidth and decrease the latency of multiprocessor communication. In functional parallelism, an individual packet concurrently visits multiple processors (e.g. 16, 19] In layer parallelism, packets visit multiple processors in a pipelined fashion (e.g. 33] Packetlevel [3, 8, 11, 13, 14, 23, 28, 32, 33, 34] and connectionlevel [8, 11, 28, 32, 34, 40] parallelisms enable concurrency at higher levels of granularity. In general, some form of network parallelism is generally necessary on multiprocessor machines, since the alternative would restrict aggregate network access to the bandwidth capacity of ....

Niraj Jain, Mischa Schwartz, and Theordore R. Bashkow. Transport protocol processing at Gbps rates. In Proc. ACM SIGCOMM, pages 188--199, Philadelphia, PA, September 1990. ACM.

....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 ....

....of the x kernel 1 Depending on the coherence state of the referenced cache line. 2 In functional parallelism, an individual packet concurrently visits multiple processors [4, 15, 17] In layer parallelism, packets visit multiple processors in a pipelined fashion [7, 24, 26] Packet level [3, 6, 9, 12, 13, 20, 23, 25, 26, 27] and connection level [6, 9, 23, 25, 27] parallelisms enable concurrency at higher levels of granularity. 3 We use the terms thread and process interchangeably. 3, 20] the STREAMS implementation in Plan 9 [23] and the ASX framework [26, 27] A related form of parallelism is found in the ....

Niraj Jain, Mischa Schwartz, and Theordore R. Bashkow. Transport protocol processing at Gbps rates. In Proceedings of the ACM SIGCOMM Conference on Communications, Architectures, Protocols and Applications, pages 188--199, Philadelphia, PA, September 1990. ACM.

....and Internet reference models) with either model. A process architecture is a logical model that may or may not be implemented using multiple processing elements (PEs) Several different approaches have been proposed to map Task based and Message based process architectures onto multiple PEs [18, 19, 10, 20, 21]. Figures 1 and 2 illustrate four models of process architecture parallelism: Layer Parallelism and Task Parallelism are Task based process architectures; Connectional Parallelism and Message Parallelism are Message based process architectures. As described in Section 4.2, the ADAPTIVE system ....

....protocol layers. In particular, most existing protocol families (such as OSI and TCP IP) possess relatively few protocol layers. Message based Process Architectures: Message based process architectures associate OS processes with messages and connections rather than with protocol layers or tasks [4, 20]. ffl Connectional Parallelism: Connectional Parallelism is a Message based process architecture that dedicates a separate PE for each active connection. Figure 2 (1) illustrates this approach, where connections C 1 ; C 2 ; C 3 , and C 4 are bound to separate PEs that process all messages ....

[Article contains additional citation context not shown here]

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedings of the SIGCOMM Symposium on Communications Architectures and Protocols, (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....management and demultiplexing [10, 11, 33] 5. Design new lightweight and adaptive protocols that are tailored for high speed, low error, and low delay network environments [19, 20, 34, 35, 36] 6. Develop alternative transport system architectures based on (a) vertical process architectures [7, 37], b) parallel processing of protocol functions [3, 4, 19] c) flexible protocol stacks that require fewer layers and or are dynamically assembled [6, 25, 26, 30] and (d) modular and extensible transport system software that supports these flexible protocol stacks [2, 18, 38] The ADAPTIVE ....

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedings of the Symposium on Communications Architectures and Protocols (SIGCOMM) , (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....architectures exhibiting high context switching overhead that penalizes interrupt driven network communication. This overhead results from the cost of flushing pipelines, invalidating CPU instruction data caches and virtual memory translationlookaside buffers, and managing register windows [5]. Alleviating the throughput preservation problem and providing very high data rates to applications requires the modification of conventional transport system architectures [6] To help system researchers navigate through the transport system design space, this paper presents a taxonomy of six ....

....onto processes. For example, Connectional Parallelism typically demultiplexes all messages bound for the same connection onto the same process, whereas Message Parallelism may demultiplex messages onto any suitable process. In general, various scheduling disciplines such as round robin [5], adaptive load balancing, and cache affinity [37] preserving techniques may be used when selecting a suitable process. ffl Connectional Parallelism Connectional Parallelism is a relatively coarse grained message based process architecture that associates a separate process with every open ....

[Article contains additional citation context not shown here]

J. Jain, M. Schwartz, and T. Bashkow, "Transport Protocol Processing at GBPS Rates," in Proceedingsof the SIGCOMM Symposium on Communications Architectures and Protocols, (Philadelphia, PA), pp. 188--199, ACM, Sept. 1990.

....through interprocess communication. Parallelism gains can be achieved mainly through pipelining effects, as shown in [9] Packet level parallelism associates processing with each individual packet, achieving speedup both with multiple connections and within a single connection. Examples include [3, 10, 13, 23]. The remainder of the paper is structured as follows: Section 2 provides background on connection level parallelism. Section 3 discusses our implementation of connection level parallelism, and describes our experiments. In section 4 we present our results. Finally, section 5 summarizes the ....

N. Jain, M. Schwartz, and T. R. Bashkow. Transport protocol processing at Gbps rates. In SIGCOMM Symposium on Communications Architectures and Protocols, pages 188--199, Philadelphia, PA, Sept. 1990. ACM.

No context found.

Jain, N., Schwartz, M., Bashkow, T. [1990] Transport Protocol Processing at GBPS Rates. ACM SIGCOMM 90, pp. 188-199.

No context found.

Jain, N., Schwartz, M., Bashkow, T. R. "Transport Protocol Processing at GBPS Rates", Proceedings of Sigcomm-90, pp. 188-199.

No context found.

, N. Jain, M. Schwartz and T. R. Bashkov, "Transport Protocol Processing at GBPS Rates.", Computer Communications Review, Vol. 20 (4), 1990, pp. 188-199.

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