It is well documented that the eeective throughput of TCP can suuer on plain ATM networks. Several research eeorts have aimed at developing additions to ATM networks like Early Packet Discard that avoid TCP throughput degradation. This paper instead investigates improvements to TCP that allow it to perform well on ATM networks without switch-level enhancements, thus avoiding additional complexity and loss of layer transparency in ATM switches. We present an enhanced version of TCP Vegas and show that it performs nearly as well on plain ATM networks as on packet-oriented networks. Moreover, like the original TCP Vegas, our version achieves a signiicant increase in throughput over the BSD 4.4-Lite TCP. We also present an analysis of TCP dynamics over ATM networks, and a simulation based study of the various enhancements in our TCP implementation and their impact on TCP performance over ATM and packet-oriented networks.

This paper presents a resource management framework for providing predictable quality of service (QoS) in Web servers. The framework allows Web server and proxy operators to ensure a probabilistic minimal QoS level, expressed as an average request rate, for a certain class of requests (called a service), irrespective of the load imposed by other requests. A measurement-based admission control framework determines whether a service can be hosted on a given server or proxy, based on the measured statistics of the resource consumptions and the desired QoS levels of all the co-located services. In addition, we present a feedback-based resource scheduling framework that ensures that QoS levels are maintained among admitted, co-located services. Experimental results obtained with a prototype implementation of our framework on trace-based workloads show its effectiveness in providing desired QoS levels with high confidence, while achieving high average utilization of the hardware.

This paper presents a resource management framework for providing predictable quality of service (QoS) in Web servers. The framework allows Web server and proxy operators to ensure a probabilistic minimal QoS level, expressed as an average request rate, for a certain class of requests (called a service), irrespective of the load imposed by other requests. A measurement-based admission control framework determines whether a service can be hosted on a given server or proxy, based on the measured statistics of the resource consumptions and the desired QoS levels of all the co-located services. In addition, we present a feedback-based resource scheduling framework that ensures that QoS levels are maintained among admitted, co-located services. Experimental results obtained with a prototype implementation of our framework on trace-based workloads show its effectiveness in providing desired QoS levels with high confidence, while achieving high average utilization of the hardware.

We present a framework that allows applications to build and customize VM services on the LA microkernel. While the LA microkernel's abstractions are quite powerful, using these abstractions effectively requires higher-level paradigms. We propose the dataspace paradigm which provides a modular VM framework. The modularity introduced by the dataspace paradigm facilitates implementation and permits dynamic configurability. Initial performance results from a prototype are promising

As the World Wide Web experiences increasing commercial and mission-critical use, server systems are expected to deliver high and predictable performance. The phenomenal improvement in microprocessor speeds, coupled with the deployment of clusters of commodity workstations has enabled server systems to meet the continually increasing performance demands in a cost-effective and scalable manner. However, as the volume, variety and sophistication of services offered by server systems increase, effective support for providing differentiated and predictable quality of service has also become important. For example, it is often desirable to differentiate between the resources allocated to virtual web sites hosted on a server system so as to provide predictable performance to individual sites, regardless of the load imposed upon others. Server systems lack adequate support for providing predictable performance to hosted services in terms of metrics that are meaningful to server applications, such as average throughput, response time etc. This is because conventional systems multiplex resources by dealing with system level metrics such as CPU/disk bandwidth, memory pages etc. Maintaining predictable levels of performance in application level metrics, therefore, requires a corresponding mapping to system level metrics that is not supported in conventional systems. High performance server systems based upon cluster-based architectures also lack adequate support for providing differentiated quality of service. This is because providing differentiated quality of service in a cluster-based server system requires global resource management not found in current clusters. This dissertation is concerned with support for both differentiated as well as predictable quality of service in server systems. The specific requirements of server applications and their interactions with the resource management facilities in the operating system software are studied. This leads to a concerted design of a resource management framework for providing effective quality of service in server systems.

This paper proposes and evaluates soft timers, a new operating system facility that allows the efficient scheduling of software events at a granularity down to tens of microseconds. Soft timers can be used to avoid interrupts and reduce context switches associated with network processing without sacrificing low communication delays.More specifically, soft timers enable transport protocols like TCP to efficiently perform rate-based clocking of packet transmissions. Experiments show that rate-based clocking can improve HTTP response time over connections with high bandwidth-delay products by up to 89% and that soft timers allow a server to employ rate-based clocking with little CPU overhead (2-6%) at high aggregate bandwidths.Soft timers can also be used to perform network polling, which eliminates network interrupts and increases the memory access locality of the network subsystem without sacrificing delay. Experiments show that this technique can improve the throughput of a Web server by up to 25%.

In network (e.g., Web) servers, it is often desirable to isolate the performance of different classes of requests from each other. That is, one seeks to achieve that a certain minimal proportion of server resources are available for a class of requests, independent of the load imposed by other requests. Recent work demonstrates how to achieve this performance isolation in servers consisting of a single, centralized node; however, achieving performance isolation in a distributed, cluster based server remains a problem.This paper introduces a new abstraction, the cluster reserve, which represents a resource principal in a cluster based network server. We present a design and evaluate a prototype implementation that extends existing techniques for performance isolation on a single node server to cluster based servers.In our design, the dynamic cluster-wide resource management problem is formulated as a constrained optimization problem, with the resource allocations on individual machines as independent variables, and the desired cluster-wide resource allocations as constraints. Periodically collected resource usages serve as further inputs to the problem.Experimental results show that cluster reserves are effective in providing performance isolation in cluster based servers. We demonstrate that, in a number of different scenarios, cluster reserves are effective in ensuring performance isolation while enabling high utilization of the server resources.

We present a scalable architecture for content-aware request distribution in web server clusters. In this architecture, a level-4 switch acts as the point of contact for the server on the Inernet and distributes the incoming requests to a number of back-end nodes. The switch does not perform any contect-based distribution. This function is performed by each of the back-end nodes, which may forward the incoming request to another back-end based on the requested contect. In terms of scalability, this architecture compares favorably to existing approaches where a front-end node performs contect-based distribution. In our architecture, the expensive operations of TCP connection estabilishment and handoff are distributes among the back-ends, rather than being centralized in the front-end node. Only a minimal additional latency penatly is paid for much improved scalability. We have implemented this new architecture, and we demonstrate its superior scalability by comparing it to a system that performs contect-aware distribution in the front-end, both under synthetic and trace-drive workloads.

We consider cluster-based network servers in which a front-end directs incoming requests to one of a number of back-ends. Specifically, we consider content-based request distribution: the front-end uses the content requested, in addition to information about the load on the back-end nodes, to choose which back-end will handle this request. Content-based request distribution can improve locality in the back-ends' main memory caches, increase secondary storage scalability by partitioning the server's database, and provide the ability to employ back-end nodes that are specialized for certain types of requests.As a specific policy for content-based request distribution, we introduce a simple, practical strategy for locality-aware request distribution (LARD). With LARD, the front-end distributes incoming requests in a manner that achieves high locality in the back-ends' main memory caches as well as load balancing. Locality is increased by dynamically subdividing the server's working set over the back-ends. Trace-based simulation results and measurements on a prototype implementation demonstrate substantial performance improvements over state-of-the-art approaches that use only load information to distribute requests. On workloads with working sets that do not fit in a single server node's main memory cache, the achieved throughput exceeds that of the state-of-the-art approach by a factor of two to four.With content-based distribution, incoming requests must be handed off to a back-end in a manner transparent to the client, after the front-end has inspected the content of the request. To this end, we introduce an efficient TCP handoflprotocol that can hand off an established TCP connection in a client-transparent manner.

This paper studies the performance of BSD-based TCP implementations in Web servers. We find that lack of scalability with respect to high TCP connection rates reduces the throughput of Web servers by up to 25% and imposes a memory overhead of up to 32 MB on the kernel. We also find that insufficient accuracy in TCP’s timers results in overly conservative delays for retransmission timeouts, causing poor response time, low network utilization and throughput loss. The paper proposes enhancements to the TCP implementation that eliminate these problems, without requiring changes to the protocol or the API. We also find that conventional benchmark environments do not fully expose certain significant performance aspects of TCP implementations and propose techniques that allow these benchmarks to more accurately predict the performance of real servers. Keywords—Internet, TCP, Webserver, Timers

This paper studies mechanisms and policies for supporting HTTP/1.1 persistent connections in cluster-based Web servers that employ content-based request distribution. We present two mechanisms for the efficient, content-based distribution of HTTP/1.1 requests among the back-end nodes of a cluster server. A trace-driven simulation shows that these mechanisms, combined with an extension of the locality-aware request distribution (LARD) policy, are effective in yielding scalable performance for HTTP/1.1 requests. We implemented the simpler of these two mechanisms, back-end forwarding. Measurements of this mechanism in connection with extended LARD on a prototype cluster, driven with traces from actual Web servers, confirm the simulation results. The throughput of the prototype is up to four times better than that achieved by conventional weighted round-robin request distribution. In addition, throughput with persistent connections is up to 26% better than without.

This paper studies mechanisms and policies for supporting HTTP/1.1 persistent connections in cluster-based Web servers that employ content-based request distribution. We present two mechanisms for the efficient, content-based distribution of HTTP/1.1 requests among the back-end nodes of a cluster server. A trace-driven simulation shows that these mechanisms, combined with an extension of the locality-aware request distribution (LARD) policy, are effective in yielding scalable performance for HTTP/1.1 requests. We implemented the simpler of these two mechanisms, back-end forwarding. Measurements of this mechanism in connection with extended LARD on a prototype cluster, driven with traces from actual Web servers, confirm the simulation results. The throughput of the prototype is up to four times better than that achieved by conventional weighted round-robin request distribution. In addition, throughput with persistent connections is up to 26% better than without.