Abhinav Kamra

Yaksha: a self-tuning controller for managing the performance of 3-tiered Web sites

Managing the performance of multiple-tiered Web sites under high client loads is a critical problem with the advent of dynamic content and database-driven servers on the Internet. This paper presents a control-theoretic approach for admission control in multitiered Web sites that both prevents overload and enforces absolute client response times, while still maintaining high throughput under load. We use classical control theoretic techniques to design a proportional integral (PI) controller for admission control of client HTTP requests. In addition, we present a processor-sharing model that is used to make the controller self-tuning, so that no parameter setting is required beyond a target response time. Our controller is implemented as a proxy, called Yaksha, which operates by taking simple external measurements of the client response times. Our design is noninvasive and requires minimal operator intervention. We evaluate our techniques experimentally using a 3-tiered dynamic content Web site as a testbed. Using the industry standard TPC-W client workload generator, we study the performance of the PI admission controller with extensive experiments. We show that the controller effectively bounds the response times of requests for dynamic content while still maintaining high throughput levels, even when the client request rate is many times that of the servers maximum processing rate. We demonstrate the effectiveness of our self-tuning mechanism, showing that it responds and adapts smoothly to changes in the workload.

The effect of DNS delays on worm propagation in an IPv6 Internet

It is a commonly held belief that IPv6 provides greater security against random-scanning worms by virtue of a very sparse address space. We show that an intelligent worm can exploit the directory and naming services necessary for the functioning of any network, and we model the behavior of such a worm in this paper. We explore via analysis and simulation the spread of such worms in an IPv6 Internet. Our results indicate that such a worm can exhibit propagation speeds comparable to an IPv4 random-scanning worm. We develop a detailed analytical model that reveals the relationship between network parameters and the spreading rate of the worm in an IPv6 world. We also develop a simulator based on our analytical model. Simulation results based on parameters chosen from real measurements and the current Internet indicate that an intelligent worm can spread surprising fast in an IPv6 world by using simple strategies. The performance of the worm depends heavily on these strategies, which in turn depend on how secure the directory and naming services of a network are. As a result, additional work is needed in developing detection and defense mechanisms against future worms, and our work identifies directory and naming services as the natural place to do it.

CountTorrent: ubiquitous access to query aggregates in dynamic and mobile sensor networks

We study the problem of aggregate querying over sensor networks where the network topology is continuously evolving. We develop scalable data aggregation techniques that remain efficient and accurate even as nodes move, join or leave the network. We present a novel distributed algorithm called CountTorrent, that enables fast estimation of certain classes of aggregate queries such as COUNT and SUM. CountTorrent does not require a static routing infrastructure, is easily implemented in a distributed setting, and can be used to inform all network nodes of the aggregate query result, instead of just the query initiator as is done in traditional query aggregation schemes. We evaluate its robustness and accuracy compared to previous aggregation approaches through simulations of dynamic and mobile sensor network environments and experiments on micaz motes. We show that in networks where the nodes are stationary, CountTorrent can provide 100% accurate aggregate results even in the presence of lossy links. In mobile sensor networks where the nodes constantly move and hence the network topology changes continuously, CountTorrent provides a close (within 10 -- 20%) estimate of the accurate aggregate query value to all nodes in the network at all times.

Fair adaptive bandwidth allocation: a rate control based active queue management discipline

Active queue management disciplines such as RED and its extensions have been widely studied as mechanisms for providing congestion avoidance, differentiated services and fairness between different traffic classes. With the emergence of new applications with diverse Quality-of-Service requirements over the Internet, the need for mechanisms that provide differentiated services has become increasingly important. We propose fair adaptive bandwidth allocation (FABA), a buffer management discipline that ensures a fair bandwidth allocation amongst competing flows even in the presence of non-adaptive traffic. FABA is a rate control based AQM discipline that provides explicit fairness and can be used to partition bandwidth in proportion to pre-assigned weights. FABA is well-suited for allocation of bandwidth to aggregate flows as required in the differentiated services framework. Since FABA can be extended to scenarios such as aggregate, hierarchical and weighted flows, it can serve as a useful method for enforcing service level agreements at the edge of the network. We study and compare FABA with other well known queue management disciplines and show that FABA ensures fair allocation of bandwidth across a much wider range of buffer sizes at a bottleneck router. Further, FABA is shown to give high values of fairness coefficient for diverse applications such as FTP, Telnet and HTTP. FABA uses randomization and has an O(1) average time complexity, and, is therefore scalable. The space complexity of the proposed algorithm is O(B) where B is the buffer size at the bottleneck router. We argue that though FABA maintains per active-flow state, through O(1) computation, reasonably scalable implementations can be deployed which is sufficient for network edges.

Controlling the performance of 3-tiered web sites: modeling, design and implementation

E-Commerce is rapidly becoming an everyday activity as consumers gain familiarity with shopping on the Internet. The infrastructure behind E-Commerce Web sites is typically composed of a three-tiered architecture, consisting of a front-end Web server, an application server and a back-end database. Two problems are frequently encountered with deploying such Web sites. First is overload, where the volume of requests for transactions at a site exceeds the site’s capacity for serving them and renders the site unusable. Second is responsiveness, where the lack of adequate response time leads to lowered usage of a site, and subsequently, reduced revenues. This paper presents a method for controlling multiple-tiered Web site performance, both by bounding response times and preventing overload. Our approach uses a self-tuning proportional integral (PI) controller for admission control, enabling overload protection and bounding response time based on an administrator-based policy (e.g., 90 percent of the requests should see a response time of less than 100 milliseconds). By using a self-tuning controller, our system automatically adapts to variation in load and requires only two parameter settings. Our method requires no changes to the operating system, Web server, application server or database. This allows rapid deployment and use of pre-existing components. We present an implementation of our controller in a proxy, called Yaksha. We evaluate our system with standard software components used in multiple-tiered e-Commerce Web sites, namely Linux, Apache, Tomcat, and MySQL. We drive the system using the industry-standard TPC-W [2] benchmark, and demonstrate that Yaksha achieves both stable behavior during overload and bounded response times. Our results show that a properly designed and implemented controller be used in a complex environment, such as multi-tiered Web sites.

Fair Adaptive Bandwidth Allocation: A Rate Control Based Active Queue Management Discipline

We propose Fair Adaptive Bandwidth Allocation (FABA), a buffer management discipline that ensures a fair bandwidth allocation amongst competing flows even in the presence of non-adaptive traffic. FABA is a rate control based active queue management discipline that provides explicit fairness and can be used to partition bandwidth in proportion to pre-assigned weights. FABA is well-suited for allocation of bandwidth to aggregate flows as required in the differentiated services framework. We study and compare FABA with other well known queue management disciplines and show that FABA ensures fair allocation of bandwidth across a much wider range of buffer sizes at a bottleneck router. FABA uses randomization and has an O(1) average time complexity, and, is therefore scalable. The space complexity of the proposed algorithm is O(B) where B is the buffer size at the bottleneck router. We argue that though FABA maintains per active-flow state, through O(1) computation, reasonably scalable implementations can be deployed which is sufficient for network edges and ISPs.

Modelling resource management for multi-class traffic in mobile cellular networks

Future cellular mobile networks will be limited by the number of channels available in each cell. On the other hand, new broadband applications like video telephony will demand tight quality of service guarantees that must be met by the network at all times. Thus, advanced mechanisms for allocating these channels to incoming calls with different quality of service level will be of utmost importance. In this paper we introduce a new analytical model for cell channel allocation to multiclass traffic. Being based on Markov chains, the new model exploits the multi-class property and reduces the state space dramatically, thus enabling the solution of previously unsolvable problem classes. We additionally describe CECALL, a simulator implementing several different strategies for allocating cell channels to multiclass traffic, handoff pre-reservation and degradation of low-level call classes. The results of the analytical model are used for explaining important simulation results.