Existing user-level network interfaces deliver high bandwidth, low latency performance to applications, but are typically unable to support diverse styles of communication and are unsuitable for use in multiprogrammed environments. Often... more
Existing user-level network interfaces deliver high bandwidth, low latency performance to applications, but are typically unable to support diverse styles of communication and are unsuitable for use in multiprogrammed environments. Often this is because the network abstraction is presented at too high a level, and support for synchronisation is inflexible. In this paper we present a new primitive for in-band synchronisation: the Tripwire. Tripwires provide a flexible, efficient and scalable means for synchronisation that is orthogonal to data transfer. We describe the implementation of a non-coherent distributed shared memory network interface, with Tripwires for synchronisation. This interface provides a low-level communications model with gigabit class bandwidth and very low overhead and latency. We show how it supports a variety of communication styles, including remote procedure call, message passing and streaming.
This work considers broadcast protocols made of successive communication rounds in the linear cost model: the time needed to send a message of length L is defined as α + Lτ, where α stands for a start-up time while Lτ represents the cost... more
This work considers broadcast protocols made of successive communication rounds in the linear cost model: the time needed to send a message of length L is defined as α + Lτ, where α stands for a start-up time while Lτ represents the cost of sending L bits of data in a channel with bandwidth 1/τ. In this model, the communication time of any algorithm [Formula: see text] is expressed as the sum [Formula: see text], where [Formula: see text] is the number of rounds and [Formula: see text] the transmission cost of the algorithm. In order to design an efficient algorithm realizing a given communication pattern, it appears that minimizing [Formula: see text] and [Formula: see text] are antinomic goals. We study this trade-off issue for broadcast protocols. Surprisingly, such a general theoretical study has almost never been done. In the literature, only the two opposite issues are actually considered: minimizing the number of rounds in the case of short messages, or minimizing the transmi...
Il n’existe pas en JavaScript une procédure ou méthode pour créer facile-ment/directement les arrays multidimensionnels. Pour arriver à ce faire, il faut passer par des constructions « artificielles ». I. Array Externe avec new Array()... more
Il n’existe pas en JavaScript une procédure ou méthode pour créer facile-ment/directement les arrays multidimensionnels.
Pour arriver à ce faire, il faut passer par des constructions « artificielles ».
I. Array Externe avec new Array() : a. Les arrays internes comme variables b. Arrays dans un Array (inline) : Les Arrays internes avec []. c. Les Arrays internes avec new Array().
II. Array externe avec []:
On peut aussi enrichir l’objet « Array » d’une méthode qui peut créer automatiquement un tableau multidimensionnel, ci-après un tableau bidi-mensionnel (matrice).
In VLSI layout of interconnection networks, routing two-point nets in some restricted area is one of the central operations. The main aim is usually minimization of the layout area, while reducing the number of wire bends is also very... more
In VLSI layout of interconnection networks, routing two-point nets in some restricted area is one of the central operations. The main aim is usually minimization of the layout area, while reducing the number of wire bends is also very useful. In this paper, we consider connecting a set of N inputs on a line to a set of N outputs on a perpendicular line inside a right triangle shaped area, where the order of the outputs is a given permutation of the order of the corresponding inputs. Such triangles were used, for example, by Dinitz, Even, and Artishchev-Zapolotsky for an optimal layout of the Butterfly network. That layout was of a particular permutation, while here we solve the problem for an arbitrary permutation case. We show two layouts in an optimal area of ½ N2 + o(N2), with O (N) bends each. We prove that the first layout requires the absolutely minimum area and yields the irreducible number of bends, while containing knock-knees. The second one eliminates knock-knees, still k...
Performance properties of switching methods used in existing interconnection networks have been mainly examined under the assumption that network traffic follows the Poisson arrival process. However, the Poisson process is unable to... more
Performance properties of switching methods used in existing interconnection networks have been mainly examined under the assumption that network traffic follows the Poisson arrival process. However, the Poisson process is unable to capture the bursty nature of traffic. In this paper we propose the first analytical model for pipelined circuit switching under bursty traffic using the Markov-Modulated Poisson Process (MMPP). Simulation experiments reveal that the proposed model exhibits a good degree of accuracy for various networks sizes and under different traffic conditions.
