Since web workloads are known to vary dynamically with time, in this paper, we argue that dynamic resource allocation techniques are necessary to provide guarantees to web applications running on shared data centers. To address this issue, we use a system architecture that combines online measurements with prediction and resource allocation techniques. To capture the transient behavior of the application workloads, we model a server resource using a time-domain description of a generalized processor sharing (GPS) server. This model relates application resource requirements to their dynamically changing workload characteristics.

Abstract: We present an architecture and prototype implementation of a performance management system for cluster-based web services. The system supports multiple classes of web services traffic and allocates server resources dynamically so to maximize the expected value of a given cluster utility function in the face of fluctuating loads. The cluster utility is a function of the performance delivered to the various classes, and this leads to differentiated service. In this paper we will use the average response time as the performance metric. The management system is transparent: it requires no changes in the client code, the server code, or the network interface between them. The system performs three performance management tasks: resource allocation, load balancing, and server overload protection. We use two nested levels of management mechanism. The inner level centers on queuing and scheduling of request messages. The outer level is a feedback control loop that periodically adjusts the scheduling weights and server allocations of the inner level. The feedback controller is based on an approximate first-principles model of the system, with parameters derived from continuous monitoring. We focus on SOAP-based web services. We report experimental results that show the dynamic behavior of the system. 1.

reconfigurability of the nodes in the data-center environment. This technique enables the nodes in the datacenter environment to efficiently adapt their functionality based on the system load and traffic pattern. While reconfigurability is a widely used technique for clusters, the datacenter environment poses several interesting challenges for the design and implementation of such a scheme. In our approach, we use the advanced features of InfiniBand such as Remote Direct Memory Access (RDMA) operations and network based atomic operations to tackle these challenges in an efficient manner without requiring any modifications to the existing data-center applications. Our experimental results show that the reconfigurability scheme provides a significantly higher performance (up to a factor of 2.5 improvement in the throughput) with the same resources or provides a similar performance while using fewer resources (up to half the nodes) as compared to a rigidly configured data-center. More importantly, our scheme takes advantage of the one-sided communication primitives offered by InfiniBand making it resilient and well-conditioned to the load on the servers as compared to two-sided communication protocols such as TCP/IP (sockets).

With the dramatic explosion of online information, the Internet is undergoing a transition from a data communication infrastructure to a global information utility. PDAs, wireless phones, webenabled vehicles, modem PCs and high-end workstations can be viewed as appliances that "plugin " to this utility for information. The increasing diversity of such appliances calls for an architecture for performance differentiation of information access. The key performance accelerator on the Internet is the caching and content distribution infrastructure. While many research efforts addressed performance differentiation in the network and on web servers, providing multiple levels of service in the caching system has received much less attention.

With the increasB us of "Virtual Machines (VMs as vehicles thatist.O1 applications running on the se. hos: it is neces sce to devis techniques that enable multipleVMs tos hare underlying resP---P.B both fairly and e#ciently.To that end, one common approach is to deploy complexresex.0 management techniques in the hos0PO infras.B11---5P.ss55.s0P , inthis paper, we advocate the us ofs5O1:---.B1:---0:. in theVMs themsem es bas- on feedback about resPR:P us05 and availability. ConsRR0. tly, we define a "Friendly" VM (FVM) to be a virtual machine that adjus--- its demand forsr.05 res5101.B s o that they are both e#ciently and fairly allocated to competing FVMs.0[ h properties areens222 us5[ one of many provably convergent control rules s uch as AIMD.By adoptingthis dis tributed application-bas--- approach to res50P2 management, itis not necesR.B to makeasP0:.B:2--- about the underlying resderly nor about the requirements of FVMs competing for thes resR---.B::--- odemonsB::--- the elegance ands implicity of our approach, wepres[ t a prototype implementation of our FVM framework inUs50O[ de Linux (UML)---an implementation thatcons[0O ofles than 500lines of code changes to UML.Wepres5 t an analytic, control-theoretic model of FVM adaptation, which es------:.B0R5 convergence andfairnes propertiesRR2. B properties areals backed up with experimental res0[R usR0 our prototype FVM implementation. 1.

Abstract — Current computing systems depend on adaptation mechanisms to ensure that they remain in quiescent operating regions. These regions are often defined using efficiency, fairness, and stability properties. To that end, traditional research works in scalable server architectures and protocols have focused on promoting these properties by proposing even more sophisticated adaptation mechanisms, without the proper attention to security implications. In this paper, we exemplify such security implications by exposing the vulnerabilities of admission control mechanisms that are widely deployed in Internet end systems to Reduction of Quality (RoQ) attacks. RoQ attacks target the transients of a system’s adaptive behavior as opposed to its limited steady-state capacity. We show that a well orchestrated RoQ attack on an end-system admission control policy could introduce significant inefficiencies that could potentially deprive an Internet end-system from much of its capacity, or significantly reduce its service quality, while evading detection by consuming an unsuspicious, small fraction of that system’s hijacked capacity. We develop a control theoretic model for assessing the impact of RoQ attacks on an end-system’s admission controller. We quantify the damage inflicted by an attacker through deriving appropriate metrics. We validate our findings through real Internet experiments performed in our lab.

Control methods are being used increasingly for uncertainty management and QoS in modern web server systems. Previous approaches have suggested combined feedforward and feedback control strategies, using queuing theory for feedforward delay prediction. While queuing theory allows one to predict delay as a function of arrival and service rates, the prediction applies only to longterm averages, and is therefore insensitive to sudden load changes. Unfortunately, Internet load is very bursty, leaving room for predictor improvement. The main contribution of this paper is an extension of the combined feedforward/feedback framework in which the queuing model is replaced with a predictor that instead uses instantaneous measurements to predict future delays. The proposed strategy is evaluated in simulation and by experiments on an Apache web server. It is shown that the new approach performs better than the combined queuing model based feedforward and feedback control presented in earlier papers. 1.

In the past few years several researchers have proposed and configured data-centers providing multiple independent services, known as shared data-centers. For example, several ISPs and other web service providers host multiple unrelated web-sites on their data-centers allowing potential differentiation in the service provided to each of them. Such differentiation becomes essential in several scenarios in a shared data-center environment. In this paper, we extend our previously proposed scheme on dynamic reconfigurability to allow service differentiation in the shared datacenter environment. In particular, we point out the issues associated with the basic dynamic reconfigurability scheme and propose two extensions to it, namely (i) Dynamic Reconfiguration with Prioritization and (ii) Dynamic Reconfiguration with Prioritization and QoS. Our experimental results show that our extensions can allow the dynamic reconfigurability scheme to attain a performance improvement of up to five times for high priority websites irrespective of any background low priority requests. Also, these extensions are able to significantly improve the performance of low priority requests when there are minimal or no high priority requests in the system. Further, they can achieve a similar performance as a static scheme with up to 43% lesser nodes in some cases.

Efficiently capturing the resource usage in a shared server environment has been a critical research issue in the past several years. With the amount of resources used by each application becoming more and more divergent and unpredictable, the solution to this problem is becoming increasingly important. In the past, several researchers have come up with a number of techniques which rely on coarse-grained monitoring of resources in order to avoid the overheads associated with fine-grained monitoring. In this paper, we propose a low-overhead efficient fine-grained resource monitoring scheme using the advanced Remote Direct Memory Access (RDMA) operation provided by RDMA-enabled interconnects such as InfiniBand (IBA). We evaluate the relative benefits of our approach against traditional approaches in various environments (including micro-benchmarks as well as real applications such as an auction server based on the RUBiS benchmark and the Ganglia distributed monitoring tool). Our results indicate that our approach for fine-grained monitoring can significantly improve the overall system utilization, thereby resulting in up to 25 % improvement in the number of requests the cluster-system can admit. 1

We present an architecture and prototype implementation of a performance management system for cluster-based web services. The system supports multiple classes of web services traffic and allocates server resources dynamically so to maximize the expected value of a given cluster utility function in the face of fluctuating loads. The cluster utility is a function of the performance delivered to the various classes, and this leads to differentiated service. In this paper we will use the average response time as the performance metric. The management system is transparent: it requires no changes in the client code, the server code, or the network interface between them. The system performs three performance management tasks: resource allocation, load balancing, and server overload protection. We use two nested levels of management mechanism. The inner level centers on queuing and scheduling of request messages. The outer level is a feedback control loop that periodically adjusts the scheduling weights and server allocations of the inner level. The feedback controller is based on an approximate first-principles model of the system, with parameters derived from continuous monitoring. We focus on SOAP-based web services. We report experimental results that show the dynamic behavior of the system. 1