This paper has benefited from the detailed and perceptive comments of our reviewers, especially our shepherd Hank Levy. We thank Randy Katz and Eric Anderson for their detailed readings of early drafts of this paper, and David Culler for his ideas on TACC's potential as a model for cluster programming. Ken Lutz and Eric Fraser configured and administered the test network on which the TranSend scaling experiments were performed. Cliff Frost of the UC Berkeley Data Communications and Networks Services group allowed us to collect traces on the Berkeley dialup IP network and has worked with us to deploy and promote TranSend within Berkeley. Undergraduate researchers Anthony Polito, Benjamin Ling, and Andrew Huang implemented various parts of TranSend's user profile database and user interface. Ian Goldberg and David Wagner helped us debug TranSend, especially through their implementation of the rewebber

permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works, must be obtained from the IEEE. Contact: Manager, Copyrights and Permissions /

(NOW) project demonstrates a new approach to largescale system design enabled by technology advances that provide inexpensive, low latency, high bandwidth, scalable interconnection networks. This paper provides an overview of the hardware and software architecture of NOW and reports on the performance obtained at each layer of the system: Active Messages, MPI message passing, and benchmark parallel applications. 1

We describe the design and evaluation of a 192-processor Windows NT cluster for high performance computing based on the High Performance Virtual Machine (HPVM) communication suite. While other clusters have been described in the literature, building a 58 GFlop/s NT cluster to be used as a general-purpose production machine for NCSA required solving new problems. The HPVM software meets the challenges represented by the large number of processors, the peculiarities of the NT operating system, the need for a production-strength job submission facility, and the requirement for mainstream programming interfaces. First, HPVM provides users with a collection of standard APIs like MPI, Shmem, Global Arrays with supercomputer class performance (13 μs minimum latency, 84 MB/s peak bandwidth for MPI), efficiently delivering

this paper we present the hardware design of a high-speed ATM firewall that does not require the termination of an end-to-end connection in the middle. We propose a novel firewall design philosophy, called Quality of Firewalling (QoF), that applies security measures of different strength to traffic with different risk levels and show how it can be implemented in our firewall. Compared with the traditional firewalls, this ATM firewall performs exactly the same packet-level filtering without compromising the performance and has the same "look and feel" by sitting at the chokepoint between the trusted ATM LAN and untrusted ATM WAN. It is also easy to manage and flexible to use.

The use of workstations connected by a fast Local Area Network (LAN) to form a so called Network of Workstations (NOW) is a very appealing idea to implement a low-cost parallel processing platform. Interprocess communication is the most difficult feature for such a system to implement with an acceptable level of performance. Standard protocols and mechanisms implemented at Operating System (OS) level usually do not provide satisfactory performance in a NOW architecture, especially with respect to the communication performance offered by the raw interconnection hardware. Two main solutions to such efficiency issue have been proposed so far, namely: standard OS mechanisms relying on simplified communication protocols, and user-level protected access to the raw communication hardware. We show that a third way, namely efficient OS mechanisms supporting an Active Message communication layer, can not only offer higher level communication primitives in a multiprogrammed environment but also o...

this paper is supported in part by DARPA orders #E313 and #E524 through the US Air Force Rome Laboratory Contracts F30602-96-1-0286 and F30602-97-2-0121. Support from Microsoft, Intel Corporation, Hewlett-Packard, 18 M. Lauria, S. Pakin, A. Chien / Efficient Layering: MPI over FM

Traditional host-resident protocol stacks are burdensome and often fail to keep pace with today's high-speed network data movement. With the PANIC system (Protocols Aboard Network Interface Cards), we explore shifting all or part of the protocol processing to the network interface card (NIC). Our system allows us to deploy user-level protocols, or portions thereof, across a collection of machines. We have implemented a first prototype of PANIC over Myrinet, and experiments show the feasibility and efficiency of this approach. Key Words: Network interface, protocol composition, networks-of-workstations, Myrinet. 1 Introduction As the performance provided by networking technologies dramatically increases, solutions for high-performance finegrained distributed computing start to emerge. Computing based on clusters, or on networks of workstations, greatly increases the performance of a variety of applications at low costs [1, 2, 3, 4]. The performance of such clusters relies heavily on l...

This work quantifies how persistent increases in processor speed compared to I/O speed reduce the performance gap between specialized, high performance messaging layers and general purpose protocols such as TCP/IP and UDP/IP. The comparison is important because specialized layers sacrifice considerable system connectivity and robustness to obtain increased performance. We first quantify the scaling effects on small messages by measuring the LogP performance of two Active Message II layers, one running over a specialized VIA layer and the other over stock UDP as we scale the CPU and I/O components. We then predict future LogP performance by mapping the LogP model's network parameters, particularly overhead, into architectural components. Our projections show that the performance benefit afforded by specialized messaging for small messages will erode to a factor of 2 in the next 5 years. Our models further show that the performance differential between the two approaches will continue to erode without a radical restructuring of the I/O system. For long messages, we quantify the variable per-page instruction budget that a zero-copy messaging approach has for page table manipulations if it is to outperform a single-copy approach. Finally, we conclude with an examination of future I/O advances that would result in substantial improvements to messaging performance.

This paper has benefited from the detailed and perceptive comments of our reviewers, especially our shepherd Hank Levy. We thank Randy Katz and Eric Anderson for their detailed readings of early drafts of this paper, and David Culler for his ideas on TACC's potential as a model for cluster programming. Ken Lutz and Eric Fraser configured and administered the test network on which the TranSend scaling experiments were performed. Cliff Frost of the UC Berkeley Data Communications and Networks Services group allowed us to collect traces on the Berkeley dialup IP network and has worked with us to deploy and promote TranSend within Berkeley. Undergraduate researchers Anthony Polito, Benjamin Ling, and Andrew Huang implemented various parts of TranSend's user profile database and user interface. Ian Goldberg and David Wagner helped us debug TranSend, especially through their implementation of the rewebber