P. T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer, 26(5):12--23, May 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....distributions to determine network performance. While other studies have looked at comparing various routing algorithms over different patterns, most studies limit the selection functions used [14 16] the range of algorithms evaluated [7, 10, 17] or the destination distributions considered [11, 18]. 3 Distribution Description NodeUniform Uniform random selection of destination node MatrixTranspose Source (x; y) selects destination (y; x) BitComplement Destination node id is the bit complement of the source id BitReversal Destination node id is the bit reversal of the source id Table ....

P. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, pp. 12--23, May 1993.

....distributions to determine network performance. While other studies have looked at comparing various routing algorithms over different patterns, most studies limit the selection functions used [14 16] the range of algorithms evaluated [7, 10, 17] or the destination distributions considered [11, 18]. Distribution Description NodeUniform Uniform random selection of destination node MatrixTranspose Source (x, y) selects destination (y,x) BitComplement Destination node id is the bit complement of the source id BitReversal Destination node id is the bit reversal of the source id Table 1: ....

P. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnec- tion networks," IEEE Computer', pp. 12 23, May 1993.

....the physical bandwidth among several messages [11] However, it has been shown that virtual channels are expensive, increasing node delay considerably [9] So, the number of virtual channels per physical channel should be kept small. An alternative approach consists of using adaptive routing [14]. However, deadlocks may appear if the routing algorithms are not carefully designed. A deadlock occurs in an interconnection network when no message is able to advance toward its destination because the network buffers are full. As routing decisions must be taken in a few nanoseconds in wormhole ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993.

.... It is very common for the adaptive routing function to contain a escape subfunction, which is usually a deterministic routing function guaranteeing a escape route to every message if the adaptive options are not available [5] VCs and adaptive routing have been extensively studied in the past [2, 6, 8, 13]. Most of the research has been based on simulations using synthetic workloads, or at most traces of real applications. In addition, adaptive routing has seen a few actual implementations, like in the Cray T3E network [14] Many of the evaluations using synthetic workloads concluded that both VCs ....

P. Gaughan and S. Yalamanchili. "Adaptive Routing Protocols for Hypercube Interconnection Networks". IEEE Computer, May 1993.

....naturally supports many parallel algorithms. Several research and commercial hypercube machines were also built (e.g. the cosmic cube, NCUBE, and Intel iPSC) Various fault tolerant routing schemes have been proposed for injured hypercubes for reliability critical applications [16] 17] 18] [19], 20] Hypercube structure has also been considered as a virtual lightwave network topology. However, it is not regarded as a practical solution since its nodal degree increases with the network size, which necessitates additional expensive nodal interfaces (including transceivers) and a larger ....

P.T. Gaughan and S. Yalamanchilli, "Adaptive Routing Protocols for Hypercube Interconnection Networks," IEEE Computer, Vol. 26, No. 5, pp. 12-24, May 1993.

....the remaining flits of the message get stranded in the network. Adaptive wormhole routing techniques have been introduced to handle the problem of message blocking. Gaughan and Yalamanchili have used a pipelined circuit switching mechanism with backtracking for fault tolerant routing systems [5]. Virtual channels have been employed for adaptive routing [4] A number of adaptive and faulttolerant wormhole routing techniques have been developed for mesh topologies [2, 3, 6, 11 and 12] that have been reviewed recently [13] We present a new fault tolerant wormhole routing technique that is ....

....message transfer. Fault free path establishment is a critical process and we employ DRB approach to find a fault free path. 3.1 Fault free Path Establishment A pioneer flit is routed by using the DRB approach to establish a fault free and economical path. In contrast to the backtracking protocol [5], packet data flits can follow the pioneer flit immediately to overlap message transfer and path establishment processes. The source node sends the pioneer flit to its two successors as explained in Fig. 3. The successors acknowledge and if the acknowledgement is received from both successors, the ....

[Article contains additional citation context not shown here]

GAUGHAN, P. T., and YALAMANCHILI, S.: `Adaptive routing protocols for hypercube interconnection networks', IEEE Computer, 1993, 26 (5), pp. 12-23

....this communication pattern each node sends messages only to the node that has the reversed bit string as identifier. The negative impact of the internal conflicts can be alleviated by adaptive algorithms. Several examples already exists for the more common topologies, as the class of k ary n cubes [91, 7, 62, 13, 17, 27, 38, 48, 57, 64, 104]. Alternatively, interconnection topologies can be provided with many redundant paths and simpler routing algorithms [44] But this kind of solution is more expensive in terms of wiring complexity. External conflicts are the result of a poor program development or compilation process. An external ....

....non monotonic allocation strategies have been proposed [48] 13] These algorithms assume = 2, as in the dynamic reversal case, but a 4 Initialized with 0. CHAPTER 3. NETWORK ROUTING 49 message can be routed along adaptive channels after using the deterministic ones. Experimental results [62] show that these algorithms are less sensitive to congestion and can reach a performance close to the network capacity. Duato has shown that, for adaptive routing, it is not necessary to eliminate all the cyclic dependencies, provided that every packet can always find a path towards its ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. Computer, 26(5), May 1993.

.... that was considered for use in a 16 PE NYU Ultracomputer prototype because it allowed short wires on a backplane connecting 4 Theta 4 switch boards that were all wired identically [16] For networks with more than one path from source to destination, adaptive routing may improve performance [48], at some cost in complicating logic at the switching nodes. We will not consider such networks here. 1.2.3 Switching strategy Maximum bandwidth and minimum latency are determined by the number of wires in the network, the way the wires are connected, and the number of hops a message must take, ....

Patrick T. Gaughan and Sudhakar Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer, pages 12--24, May 1993.

....is to find an appropriate routing scheme. In general, routing is a more critical design factor for wormhole switched networks than for buffered networks. Accordingly, a large number of algorithms have been pro16 16 posed [114] Adaptive routing is a central issue for wormhole switched networks [5,6,35,40,49, 50,88,104,105]. The objective is either to avoid blocking or to recover from it. Another way to deal with blocking is in terms of virtual channels as proposed by [33, 34, 37] The idea is to equip the switches with parallel flit buffers, also called multi lane buffers. Then a suspended packet can be decoupled ....

GAUGHAN, P., AND YALAMANCHILI, S. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer, 5 (1993), 12--23.

....5 presents our experimental methodology and simulation results. Section 6 summarizes this paper. 2 Background and Related Work High performance interconnection networks in tightly coupled multiprocessors can be achieved by using wormhole [6, 5] or cut through switching [15] adaptive routing [12], and multiple virtual channels [4] The Cray T3E [25] and SGI Origin [18] machines use a combination of these techniques for their interconnection networks. In these systems communication occurs by sending packets of information that are routed independently through the network. Each packet is ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for hypercube Interconnection Networks. IEEE Computer, pages 12--22, May 1993.

....to compute the optimal sequence of lateral links to be used in the path to the destination node. Such a sequence can be used in the header of the message, making the algorithm usable with different communication mechanisms (e.g. packet switching, circuit switching and virtual cut through routing) [11]. 4 Diameter The diameter of the n SCC graph can be calculated from its antipodes. An antipode is the farthest node from a given node along the shortest path [10] In particular, we consider the antipodes of an identity node hi 1 ; 1 i when deriving the diameter of the SCC graph. We recall that ....

P. T. Gaughan and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks, " Computer, Vol. 26, No. 5, May 1993, pp. 12-23.

....communication operations, its development has been along a different direction than the conventional development of wormhole routed systems. In the past, wormhole routed systems started with e cube routing [12] Recently in the literature, various adaptive routing schemes have been proposed [5, 8, 11, 13, 15, 16, 19, 24, 31, 34]. Though today s commercial and experimental systems (Intel Paragon, Cray T3D, Ncube 3, Stanford DASH) use only e cube routing, a natural extension for these machines in the near future will be to go for adaptive routing. 1 In order to implement Hamiltonian path based routing, these machines need ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE Computer, 26:12--23, May 1993.

.... It is very common for the adaptive routing function to contain a escape subfunction, which is usually a deterministic routing function guaranteeing a escape route to every message if the adaptive options are not available [5] VCs and adaptive routing have been extensively studied in the past [2, 6, 8, 13]. Most of the research has been based on simulations using synthetic workloads, or at most traces of real applications. In addition, adaptive routing has seen a few actual implementations, like in the Cray T3E network [14] Many of the evaluations using synthetic workloads concluded that both VCs ....

P. Gaughan and S. Yalamanchili. \Adaptive Routing Protocols for Hypercube Interconnection Networks". IEEE Computer, May 1993.

....are still very few commercial systems which use it, and some researchers have raised questions about its effectiveness [96] It is our belief that the parallel computing community has hesitated to adopt adaptive routing because the perceived need is not great enough to justify the added costs. In [37], Gaughan and Yalamanchili provide a taxonomy for adaptive routing protocols. Figure 2.16 shows this taxonomy, with the terminology adjusted to match our work. The taxonomy is split into two main classes of protocols, progressive and backtracking. In a progressive protocol, a blocked packet may ....

....but we do not focus on this approach. 30 Full Partial Full Partial Full Partial Adaptive Routing Progressive Algorithms Backtracking Algorithms Full Partial Minimal Minimal Non minimal Non minimal Figure 2. 16: A taxonomy for adaptive routing algorithms, due to Gaughan and Yalamanchili [37]. Adaptive algorithms may be minimal or non minimal, with non minimal methods providing more adaptivity (and the potential for more fault tolerance) at the expense of longer paths. A minimal adaptive algorithm is fully adaptive if all minimal paths are available to every packet. Otherwise, it ....

[Article contains additional citation context not shown here]

P. Gaughan and S. Yalamanchili, Adaptive Routing Protocols for Hypercube Interconnection Networks, IEEE COMPUTER, 26(5):12--23, May 1993.

....tree effects on network performance in direct and indirect networks have been proposed. These concepts can be divided into four classes: 1) combining techniques [7] 2) flow control techniques [8] 3) enhanced switch box and router designs [4, 9, 10] and (4) adaptive routing protocols [11]. Hardware combining methods do not work for most routing protocols in mesh networks. Flow control techniques, like feedback or discarding networks [8] often result in decreased performance under uniform traffic. Most enhanced router designs for wormholerouting direct networks (e.g. 9] are ....

....performance under uniform traffic. Most enhanced router designs for wormholerouting direct networks (e.g. 9] are unable to control the traffic flow under nonuniform traffic patterns. In direct networks, adaptive routing protocols are able to route messages around congested network areas [11] however, under hot spot traffic patterns, the overall network might become congested, so that adaptive routing is unable to find lesser congested routes. To the knowledge of the author, no concepts for the traffic control (rather than just traffic routing) in wormhole routing mesh networks are ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, Vol. 26, No. 5, May 1993, pp. 12-23.

.... The use of virtual channels can increase throughput considerably by dynamically sharing the physical bandwidth among several messages [8] However, it has been shown that virtual channels are expensive, increasing node delay [6] An alternative approach consists of using adaptive routing [18]. However, deadlocks may appear if the routing algorithms are not carefully designed. A deadlock occurs in an interconnection network when no message is able to advance toward its destination because the network buffers are full. As routing decisions must be taken in a few nanoseconds in wormhole ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993.

....between them unless a nonminimal path around the fault is permitted. There are a variety of choices for minimal and nonminimal routing. For example, a nonminimal routing algorithm may or may not permit a message to backtrack along the path it has already taken. Gaughan and Yalamanchili [35] present a comprehensive overview of the different adaptive routing protocols that exist. Whether minimal or nonminimal, adaptive routing algorithms can be further differentiated by the fraction of shortest paths they allow. Partially adaptive routing algorithms do not allow all packets to use any ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE Computer, 26(5):12--23, May 1993.

....while nonminimal routing algorithms do not require messages to use only shortest paths. Minimal routing algorithms provide higher throughput for high message traffic and are generally simpler to implement. Nonminimal routing algorithms are useful for fault tolerance. Gaughan and Yalamanchili [15] present a good overview of adaptive routing protocols. Whether minimal or nonminimal, adaptive routing algorithms can be further differentiated by the fraction of shortest paths they allow. Partially adaptive routing algorithms do not allow all messages to use any shortest path. Fully adaptive ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE Computer, 26(5):12--23, May 1993.

....by the router are busy, the header is blocked until one of those channels is freed; the flow control within the network blocks the trailing flits. Currently, wormhole is the switching technique that achieves the lowest message latency. See [31] for a detailed analysis of wormhole switching and [15] for a study of adaptive routing protocols. Virtual cut through [23] was initially proposed for computer networks. It also reduces latency by pipelining packet transmission across successive channels. When the packet header arrives at a node, it is routed without waiting for the rest of the ....

P.T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Comput. Mag., vol. 26, no. 5, pp. 12--23, May 1993.

....while nonminimal routing algorithms do not require messages to use only shortest paths. Minimal routing algorithms provide higher throughput for high message traffic and are generally simpler to implement. Nonminimal routing algorithms are useful for fault tolerance. Gaughan and Yalamanchili [15] present a good overview of adaptive routing protocols. Whether minimal or nonminimal, adaptive routing algorithms can be further differentiated by the fraction of shortest paths they allow. Partially adaptive routing algorithms do not allow all messages to use any shortest path. Fully adaptive ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE Computer, 26(5):12--23, May 1993.

.... nodes involves several issues, such as data transmission mechanisms and routing requirements [2] Most common interconnection topologies provide multiple physical paths for routing messages between two nodes: messages can either follow a path deterministically chosen or be routed adaptively [3]. Adaptive routing, in general, leads to improved performances, especially in the case of non uniform traffic conditions. The goal of the routing algorithm presented in the paper, called Virtual Path (VP) is to provide efficient communications for applications composed of long lasting entities ....

P. Gaughan, S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks", IEEE Computer, Vol. 26, No. 5, May 1993.

.... an option for some gather scatter related operations on Connection Machine systems [46] Minimal and non minimal deterministic, adaptive, routing algorithms, have been proposed by many researchers [9, 11, 12, 14, 15, 23, 32, 38, 37] A survey of adaptive routing for binary cubes can be found in [13]. Deterministic, adaptive routing is used on the Connection Machine models CM 1, CM 2, and CM 200 [17] The Chaos router [24] is a randomized, adaptive router, which allows messages to follow random shortest paths from source to destination, but misroutes messages when congestion occurs. A chip ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE COMPUTER, 26(5):12--23, May 1993.

....operations on Connection Machine systems [33] Adaptive routers which do not employ randomization have been proposed by Dally and Aoki [4] Duato [6] Konstantinidou [17] Ngai and Seitz [24] and Pifarr e et al. 26] among others. A survey of adaptive routing for binary cubes can be found in [7]. Deterministic, adaptive routing is used on the Connection Machine models CM 1, CM 2, and CM 200 [9] The Chaos router [18] is a randomized, adaptive router, which allows messages to follow random shortest paths from source to destination, but will misroute messages when congestion occurs. A ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. COMPUTER, 26(5):12--23, May 1993.

....and scattering) The cases where a source node wishes to transmit one or more than one messages are distinguished. The interconnection network under consideration is the multidimensional torus network, which has been proven to be a flexible topology for the interconnection of processors [5, 14]. An n dimensional, k ary multidimensional torus, denoted by MT n;k , has N = k n processors, each one labeled by an n digit number in radix k arithmetic. Two processors are connected if their labels differ in exactly one digit by j mod k, j 2 f Gamma1; 1g, i.e. processor v n Gamma1 : v i 1 v ....

....to derive solutions for the single node and multinode reduction over an associative operator, and for the single node and multinode gathering problems by inversing the transmission of the messages. A survey on adaptive communication algorithms on the multidimensional torus network can be found in [14]. Another collection of communication algorithms on the MT n;k network under a variety of communication models can be found in [11] The method of spanning graph construction has been previously used to design communication algorithms on other interconnection networks, such as the binary hypercube ....

[Article contains additional citation context not shown here]

P.T. Gaughan and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks ", Computer, vol. 26, no. 5, pp. 12-23, 1993.

....placed on the choice of virtual channels on that physical channel. Hence, not all fully adaptive routing algorithms are equivalent. Some fully adaptive routing algorithms allow more adaptiveness than others by placing fewer restrictions on the choice of virtual channels. Gaughan and Yalamanchili [8] present a good overview of adaptive routing protocols. Since each virtual channel needs a separate buffer and the virtual channels are multiplexed across the physical channel, the number of virtual channels required by an adaptive routing algorithm gives a good approximation of the hardware cost ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. IEEE Computer, 26(5):12--23, May 1993.

....and destination. Latency and contention can be reduced by using these multiple paths. Routing algorithms are either minimal or nonminimal. Minimal routing allows only shortest paths to be chosen, while nonminimal routing does not require packets to use only shortest paths. Gaughan and Yalamanchili [13] present a comprehensive overview of adaptive routing protocols. Whether minimal or nonminimal, adaptive routing algorithms can be further differentiated by the fraction of shortest paths they allow. Partially adaptive routing algorithms do not allow all packets to use any shortest path. Fully ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube InterconnectionNetworks. IEEE Computer, 26(5):12--23, May 1993.

....fully adaptive algorithms a message 9 S D Figure 1.4: An example of a non minimal path between a source (S) and destination (D) on a 2D mesh. may take any shortest path as well as some longer paths. Specific routing algorithms are described later in Chapter 2 and in the literature [DYN97, GY93, NM93] 1.1.5 Router Properties As mentioned earlier, deadlock, livelock, and starvation freedom are three main properties every practical router must possess. We briefly discuss the different methods of achieving deadlock and livelock freedom. All three properties are covered in more detail in ....

P.T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer Magazine, 26(5):12--23, May 1993.

....and scattering) The cases where a source node wishes to transmit one or more than one messages are distinguished. The interconnection network under consideration is the generalized hypercube network, which has been proven to be a flexible topology for the interconnection of processors [5, 14]. An n dimensional, k ary generalized hypercube, denoted by GH n;k , has N = k n processors, each one labeled by an n digit number in radix k arithmetic. Two processors are connected if their labels differ in exactly one digit, i.e. processor v n Gamma1 : v i 1 v i v i Gamma1 : v 0 is ....

....optimal solutions for the single node and multinode reduction over an associative operator, and for the single node and multinode gathering problems by inversing the transmission of the messages. A survey on adaptive communication algorithms on the generalized hypercube network can be found in [14]. The method of spanning graph construction has been previously used to design communication algorithms on other interconnection networks, such as the binary hypercube [16] the multidimensional torus [10] and the star network [6, 7] The remainder of this paper is organized as follows. Notations ....

P.T. Gaughan and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks ", Computer, pp. 12-23, 1993.

....are allowed, provided that there exists a connected channel subset free of cyclic dependencies. To simplify the application of the former theory, two design methodologies have been proposed. Some adaptive routing algorithms designed with the former methodologies have been evaluated by simulation [12, 13, 14, 16, 25, 26]. The results show an important improvement over previous routing algorithms. The above mentioned routing schemes are intended for one to one communications. However, there are several parallel applications that require one to many communications, such as the simulation of computer networks, ....

P.T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Comput. Mag., vol. 26, no. 5, pp. 12--23, May 1993.

....one destination node are involved; in Section 3.4 this restriction will be removed. 3.3. 1 Routing The problem of routing a message through a network to avoid congestion, deadlock, livelock, and starvation has long been studied [25] Many routing algorithms have been proposed for various networks [33, 36, 59]. Each can be defined as a possibly non deterministic function yielding a path between two given nodes. Fortunately, in our less detailed simulation we do not need a routing algorithm. Instead, we assume that optimal routes are always taken; which route is chosen does not matter, since we are not ....

Patrick T. Gaughan and Sudhakar Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. Computer, 26(5):12--23, May 1993.

....Machine systems was discussed in Section 2.5. Adaptive routers which do not employ randomization have been proposed by Dally and Aoki [6] Duato [9] Konstantinidou [38] Ngai and Seitz [57] and Pifarr e et al. 58] among others. A survey of adaptive routing for binary cubes can be found in [15]. Deterministic, adaptive routing is used on the CM 2 200 [18] The Chaos router [2, 39] is a randomized, adaptive router, which allows messages to follow random shortest paths from source to destination, but will misroute messages when congestion occurs. ROMM (Randomized, Oblivious, Minimal, ....

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for Hypercube Interconnection Networks. COMPUTER, 26(5):12--23, May 1993.

....compute the paths that must be followed by the connection establishment probes, and by the control and best effort messages, in their travel through the network. Due to the fact that in PCS the probes travel independently of the data, the Exhaustive Profitable Backtracking (EPB) routing algorithm [42] is deployed when establishing connections. The EPB algorithm performs an exhaustive search of the minimal paths in the network until a valid path is found or the probe backtracks to the source node. On the other hand, best effort messages are routed according to a fully adaptive routing algorithm ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, May 1993.

....virtual cut through and pipelined circuit switching. But, due to the fact that in PCS the probes travel independently of the data, more flexible routing algorithms can be deployed in this case, i.e. backtracking algorithms. An example is the Exhaustive Profitable Backtracking (EPB) algorithm [39], which performs an exhaustive search of the minimal paths in the network until a valid path is found or the probe backtracks to the source node. The PCS connection establishment process will be slightly modified in the MMR. In PCS a probe is sent from the source to the destination, carrying ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, May 1993.

....not affect the QoS received by multimedia flows. As there is no single switching technique suitable for both kinds of traffic, the MMR uses a hybrid approach, where the most suitable switching technique is used for each kind of traffic: a connectionoriented scheme (Pipelined Circuit Switching [10, 11]) for the multimedia flows, and Virtual CutThrough for best effort messages. For both schemes, the flow control unit has the same size and will be referred to as a flit. Flits are synchronously forwarded through the crossbar. Input Buffers The MMR is able to support a large number of connections. ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993. 13

....followed by the probes when establishing a connection and by the control and best effort packets. A deadlockfree fully adaptive routing algorithm proposed for networks with irregular topology [11] 12] will be used to route packets using VCT switching. Exhaustive profitable backtracking (EPB) [6] will be used when establishing connections. A connection is established in PCS by routing a probe from source to destination. This probe contains some control information, the destination address, and information about bandwidth requirements. The destination node returns an acknowledgement to the ....

P. T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993.

No context found.

P. T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer Magazine, May 1993.

....use of the channels they occupy and wasting channel bandwidth. Virtual channels can increase throughput considerably by dynamically sharing the physical bandwidth among several messages [7] Another approach to reduce contention and improve channel utilization consists of using adaptive routing [11]. Adaptive routing algorithms must be carefully designed to avoid deadlocks [8, 9] Virtual channels can be combined with adaptive routing to maximize throughput. Unfortunately, virtual channels and adaptive routing make the router more complex, increasing node delay [4] As a consequence, latency ....

P.T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993.

....contention produced by blocked messages. Physical channels may be split into virtual channels, increasing throughput by dynamically sharing the physical bandwidth among several messages [7] Another approach to reduce contention and improve channel utilization consists of using adaptive routing [12, 10]. Adaptive routing algorithms must be carefully designed to avoid deadlocks [9, 11] Unfortunately, virtual channels and adaptive routing make the router more complex, increasing node delay [4] As a consequence, latency may increase. Another method to increase channel bandwidth consists of using ....

P.T. Gaughan and S. Yalamanchili, "Adaptive routing protocols for hypercube interconnection networks," IEEE Computer, vol. 26, no. 5, pp. 12--23, May 1993.

....network until a message in the buffer has been routed. 6. 1 Fault Free Performance In Figure 12, the latency throughput curves of Two phase routing with detours is compared with those of Duato s Protocol (DP) a WR protocol) 12] and Misrouting, Backtrack with m misroutes (MB m) a PCS protocol) [17] in the fault free network. In the fault free network, TP routing approximates WR by setting the scouting distance, K, to zero. The current design eliminates any positive acknowledgments from being transmitted when SR = 0. Notice that SR = 0 sets K = 0 in every virtual channel crossed. ....

P. T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer Magazine, May 1993.

No context found.

P. T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer, 26(5):12--23, May 1993.

No context found.

P. T. Gaughan and S. Yalamanchili. Adaptive routing protocols for hypercube interconnection networks. IEEE Computer, 26(5):12--23, May 1993.

No context found.

Gaughan, P. T. and Yalamanchili, S. (1993) Adaptive routing protocols for hypercube interconnection networks. IEEE Comput., 26, 12--23.

No context found.

P. T. Gaughan and S. Yalamanchili. Adaptive Routing Protocols for hypercube Interconnection Networks. IEEE Computer, pages 12--22, May 1993.

No context found.

P.T. Gaughan, and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks", IEEE Computer, vol.26, no.5, pp.12-23, May 1993.

No context found.

GYA93: P Gaughan and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks", IEEE Computer, May 1993.

No context found.

GYA93: P Gaughan and S. Yalamanchili, "Adaptive Routing Protocols for Hypercube Interconnection Networks", IEEE Computer, May 1993.

No context found.

P. T. Gaughan and S. Yalamanchili, Adaptive routing protocols for hypercube interconnection networks, IEEE Computer, 26 (5): 12-23, 1993.

No context found.

Patrick T. Gaughan and Sudhakar Yalamanchili, 1993: "Adaptive Routing Protocols for Hypercube Interconnection Networks," Computer, May 1993, pp. 12-23

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