cs6235-10.1.1.158.6724 - Real-Time Systems, 24, 171202,...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Real-Time Systems, 24, 171±202, 2003 # 2003 Kluwer Academic Publishers. Manufactured in The Netherlands. Utilization-Based Admission Control for Scalable Real-Time Communication BYUNG-KYU CHOI bkchoi@mtu.edu Department of Computer Science, Michigan Technological University, Houghton, MI 49931-1295, USA DONG XUAN xuan@cis.ohio-state.edu Department of Computer and Information Science, The Ohio State University, Columbus, OH 43210-1277, USA RICCARDO BETTATI bettati@cs.tamu.edu Department of Computer Science, Texas A&M University, College Station, TX 77843-3112, USA WEI ZHAO zhao@cs.tamu.edu Department of Computer Science, Texas A&M University, College Station, TX 77843-3112, USA CHENGZHI LI c14v@cs.virginia.edu Department of Computer Science, University of Virginia, Charlottesville, VA 22904-4740, USA Abstract. We propose a utilization-based schedulability and admission control mechanism for distributed systems with workload aggregation to achieve scalability. We use the differentiated services ( diffserv ) architecture to describe and illustrate our approach. Scalability of admission control is achieved by determining off-line safe levels of server utilization. Scalability during the connection lifetimes is provided by the aggregation mechanisms (for example, class-based scheduling) provided by the diffserv architecture. Our evaluations show that our system's admission probabilities are very close to those of signi®cantly more expensive approaches, which do not allow for workload aggregation. At the same time, admission control overhead during ¯ow establishment is very low. Keywords: admission control, end-to-end deadlines, schedulable utilization, QoS guarantees, and differentiated services 1. Introduction This paper presents a methodology to apply utilization-based admission control to provide hard real-time communication in a scalable fashion. By hard real-time we mean that packets are delivered from their sources to the destinations within their end-to-end deadlines. Packets delivered beyond their end-to-end deadlines are considered useless. By scalable we mean that we want to support a large number (say, many thousands) of connections over a network with many nodes (in our context nodes are switches or routers), while at the same time allowing for high resource utilization. The integrated services (IS) architecture of the IETF (Braden et al., 1994) is a good example to illustrate the types of problems encountered in admission control for large systems. In IS, each connection is strictly controlled both by admission control at connection establishment time and by packet scheduling during the lifetime of the connection. At establishment time, the necessary resources for the new connection are
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
allocated, and during the lifetime, the connection is policed to ensure that abnormal behavior of a connection does not affect other connections. This necessitates that information about every connection is kept by each node along the path for admission control and packet forwarding purposes.
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 10/10/2010 for the course CS 6235 taught by Professor Staff during the Fall '08 term at Georgia Institute of Technology.

Page1 / 32

cs6235-10.1.1.158.6724 - Real-Time Systems, 24, 171202,...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online