Is it possible to reduce the expected response time ofevery request at a web server, simply by changing the order in which we schedule the requests? That is the question we ask in this paper. This paper proposes a method for improving the performance of web servers servicing static HTTP requests. The idea is to give preference to those requests which are short, or have small remaining processing requirements, in accordance with the SRPT (Shortest Remaining Processing Time) scheduling policy. The implementation is at the kernel level and in-volves controlling the order in which socket buffers are drained into the network.Experiments are executed both in a LAN and a WAN environment. We use the Linux operating system and the Apache and Flash web servers. Results indicate that SRPT-based scheduling of connections yields significant reductions in delay at the web server. These result in a substantial reduction inmean response time, mean slowdown, and variance in response time for both the LAN and WAN environments. Significantly, and counter to intuition, the large requests are only negligibly penalized or not at all penalized as a result of SRPT-based scheduling.

The analysis of workloads is important for understanding how systems are used. In addition, workload models are needed as input for the evaluation of new system designs, and for the comparison of system designs. This is especially important in costly large-scale parallel systems. Luckily, workload data is available in the form of accounting logs. Using such logs from three dierent sites, we analyze and model the job-level workloads with an emphasis on those aspects that are universal to all sites. As many distributions turn out to span a large range, we typically rst apply a logarithmic transformation to the data, and then t it to a novel hyper-Gamma distribution or one of its special cases. This is a generalization of distributions proposed previously, and leads to good goodness-of-t scores. The parameters for the distribution are found using the iterative EM algorithm. The results of the analysis have been codied in a modeling program that creates a synthetic workload based on the results of the analysis. 1

It is common to classify scheduling policies based on their mean response times. Another important, but sometimes opposing, performance metric is a scheduling policy’s fairness. For example, a policy that biases towards short jobs so as to minimize mean response time, may end up being unfair to long jobs. In this paper we define three types of unfairness and demonstrate large classes of scheduling policies that fall into each type. We end with a discussion on which jobs are the ones being treated unfairly. 1

When a moldable job is submitted to a space-sharing parallel computer, it must choose whether to begin execution on a small, available cluster or wait in queue for more processors to become available. To make

Abstract. The evaluation of parallel job schedulers hinges on two things: the use of appropriate metrics, and the use of appropriate workloads on which the scheduler can operate. We argue that the focus should be on on-line open systems, and propose that a standard workload should be used as a benchmark for schedulers. This benchmark will specify distributions of parallelism and runtime, as found by analyzing accounting traces, and also internal structures that create different speedup and synchronization characteristics. As for metrics, we present some problems with slowdown and bounded slowdown that have been proposed recently. 1

... This paper attacks this problem by considering requested time (and its relation with execution time) and the possibility of job cancellation, two aspects of the supercomputer workload that have not been modeled yet. Moreover, we also improve upon existing models for the arrival instant and partition size.

We consider a distributed server system and ask which policy should be used for assigning jobs (tasks) to hosts. In our server, jobs are ot preemptible. Also, the job's service demand is ot known a priori. We are particularly concerned with the case where the workload is heavy-tailed, as is characteristic of many empirically measured computer workloads. We analyze several natural task assignment policies and propose a new one T/IGS (Task Assignment based on Guessing Size). The T/IGS algorithm is counterintuitive in many respects, including load mbalancing, omwork-conserving, and faithless. We find that under heavy-tailed workloads, T/IGS can outperform all task assignment policies known to us by several orders of magnitude with respect to both mean response time and mean slowdown, provided the system load is not too high.

This note briey summarizes some results from two papers: [4] and [23]. These papers pose the following question: Is it possible to reduce the expected response time of every request at a web server, simply by changing the order in which we schedule the requests? In [4] we approach this question analytically via an M/G/1 queue. In [23] we approach the same question via implementation involving an Apache web server running on Linux.

We explore the performance of an M/GI/1 queue under various scheduling policies from the perspective of a new metric: the slowdown experienced by largest jobs. We consider scheduling policies that bias against large jobs, towards large jobs, and those that are fair, e.g., Processor-Sharing. We prove that as job size increases to infinity, all work conserving policies converge almost surely with respect to this metric to no more than 1=(1 \Gamma ae), where ae denotes load. We also find that the expected slowdown under any work conserving policy can be made arbitrarily close to that under Processor-Sharing, for all job sizes that are sufficiently large.

The study and design of computer systems requires good models of the workload to which these systems are subjected. Until recently, the data necessary to build these models---observations from production installations---were not available, especially for parallel computers. Instead, most models were based on assumptions and mathematical attributes that facilitate analysis. Recently a number of supercomputer sites have made accounting data available that make it possible to build realistic workload models. It is not clear, however, how to generalize from specific observations to an abstract model of the workload. This paper presents observations of workloads from several parallel supercomputers and discusses modeling issues that have caused problems for researchers in this area. 1 Introduction We like to think of building computer systems as a systematic process of engineering---we define requirements, draw designs, analyze their properties, evaluate options, and finally construct a w...