of persistent connections. protocol. Section 7 describes the design of the simulation

research relevant to the design and application of high performance scientific computers. We test our ideas by designing, building, and using real systems. The systems we build are research prototypes; they are not intended to become products. There are two other research laboratories located in Palo Alto, the Network Systems

of these algorithms work well only when the nature of the traffic is consistent. However, in client/server applications, such as HTTP persistent connections that involve multiple request/response in a single TCP connection, these al-gorithms may not be able to function properly. We have proposed Cocktail approach

on simulation experiments, driven by traces collected from connection for each request. Modifications to HTTP an extremely busy Web server, that support the proposed have been proposed that would transport multiple re- HTTP modifications. According to these simulations, the quests over each TCP connection

Several overload admission control architectures have been developed to protect web servers from overload. Some of these architectures base their admission decision on information found in the HTTP header. In this context, persistent connections represent a challenging problem since the HTTP header

, the traffic pattern of the connection, the lack of knowledge of the number of congested gateways, the possibilities of routing changes, or other difficulties in distinguishing propagation delay from persistent queueing delay. The most effective detection of congestion can occur in the gateway itself

This paper describes a mechanism, called "persistent connection," to preserve stream connections after the communicating peer exits and till it restarts. Such connections have many applications: to survive failures that crash one party, network partitions that cut off the two parties

We consider a finite capacity Erlang loss system that alter-nates between active and inactive states according to a two state modulating Markov process. Work arrives to the sys-tem as a Poisson process but is blocked from entry when the system is at capacity or inactive. Blocked jobs cost the owner a fixed amount that depends on whether blockage was due to the system being at capacity or due to the system being inactive. Jobs which are present in the system when it becomes inactive pause processing until the system becomes active again. A Laplace transform expression for the expected undis-counted revenue lost in [0, t] due to blocking is found. Fur-ther, an expression for the total time discounted expected lost revenue in [0,∞) is provided. We also derive a second order approximation to the former that can be used when the computing power to invert the Laplace transform is not available. These expressions can be used to ascribe a value to four alternatives for improving system performance: (i) increasing capacity, (ii) increasing the service rate, (iii) in-creasing the repair rate, or (iv) decreasing the failure rate.

Abstract not found

in heavily utilized paths, as well as in paths with limited capacity (probably due to a lower degree of statistical multiplexing). We finally examine the relation between avail-bw and TCP throughput. A persistent TCP connection can be used to roughly measure the avail-bw in a path, but TCP saturates the path