Vivek S. Pai

CoMon: a mostly-scalable monitoring system for PlanetLab

CoMon is an evolving, mostly-scalable monitoring system for PlanetLab that has the goal of presenting environment-tailored information for both the administrators and users of the PlanetLab global testbed. In addition to passively reporting metrics provided by the operating system, CoMon also actively gathers a number of metrics useful for developers of networked systems. Using CoMon, PlanetLab administrators and users can easily spot problematic machines, where the problem may arise from the machine itself, local configuration/environment problems, or the workload running on the machine. Furthermore, users can easily observe many properties of all of the experiments running across multiple PlanetLab nodes, facilitating not only their own experiment monitoring and debugging, but also helping scale the task of finding PlanetLab problems.In this paper we describe CoMons design and operation, including what kinds of data are gathered, the scale of the processing involved, and the approaches we have taken to keep CoMon running. Our goal is not only to illustrate the kinds of problems faced in this environment, but also to invite others to participate, either by experimenting with the data generated by CoMon, or by building on the CoMon system itself.

Using PlanetLab for network research: myths, realities, and best practices

PlanetLab is a continuously-evolving global network research testbed that is simultaneously used by hundreds of researchers for diverse tasks, ranging from short-term self-contained experiments among PlanetLab nodes to continuously-running Web-accessible services with tens of thousands of non-PlanetLab users. While PlanetLab cannot provide a perfectly-customized environment for every experiment, it has been changing over time, and the base of knowledge of how to best utilize it has also been growing. As a result, many of the early observations researchers made about PlanetLab would change if rechecked today. In this paper, we discuss these issues and explain whether they remain, have been addressed via PlanetLabs evolution, or can be avoided by the use of best practices. Where possible, we provide quantitative evidence showing the realities of PlanetLab and possible research avenues to further broaden the opportunities for using PlanetLab in network research.

The effectiveness of request redirection on CDN robustness

It is becoming increasingly common to construct network services using redundant resources geographically distributed across the Internet. Content Distribution Networks are a prime example. Such systems distribute client requests to an appropriate server based on a variety of factors---e.g., server load, network proximity, cache locality--in an effort to reduce response time and increase the system capacity under load. This paper explores the design space of strategies employed to redirect requests, and defines a class of new algorithms that carefully balance load, locality, and proximity. We use large-scale detailed simulations to evaluate the various strategies. These simulations clearly demonstrate the effectiveness of our new algorithms, which yield a 60--91% improvement in system capacity when compared with the best published CDN technology, yet user-perceived response latency remains low and the system scales well with the number of servers.

Towards understanding modern web traffic

As the nature of Web traffic evolves over time, we must update our understanding of underlying nature of todays Web, which is necessary to improve response time, understand caching effectiveness, and to design intermediary systems, such as firewalls, security analyzers, and reporting or management systems. In this paper, we analyze five years (2006-2010) of real Web traffic from a globally-distributed proxy system, which captures the browsing behavior of over 70,000 daily users from 187 countries. Using this data set, we examine major changes in Web traffic characteristics during this period, and also investigate the redundancy of this traffic, using both traditional object-level caching as well as content-based approaches.

Evaluating the impact of simultaneous multithreading on network servers using real hardware

This paper examines the performance of simultaneous multithreading (SMT) for network servers using actual hardware, multiple network server applications, and several workloads. Using three versions of the Intel Xeon processor with Hyper-Threading, we perform macroscopic analysis as well as microarchitectural measurements to understand the origins of the performance bottlenecks for SMT processors in these environments. The results of our evaluation suggest that the current SMT support in the Xeon is application and workload sensitive, and may not yield significant benefits for network servers.In general, we find that enabling SMT on real hardware usually produces only slight performance gains, and can sometimes lead to performance loss. In the uniprocessor case, previous studies appear to have neglected the OS overhead in switching from a uniprocessor kernel to an SMT-enabled kernel. The performance loss associated with such support is comparable to the gains provided by SMT. In the 2-way multiprocessor case, the higher number of memory references from SMT often causes the memory system to become the bottleneck, offsetting any processor utilization gains. This effect is compounded by the growing gap between processor speeds and memory latency. In trying to understand the large gains shown by simulation studies, we find that while the general trends for microarchitectural behavior agree with real hardware, differences in sizing assumptions and performance models yield much more optimistic benefits for SMT than we observe.

Extensible kernels are leading OS research astray

Argues that ongoing research in extensible kernels largely fails to address the real challenges facing the operating systems (OS) community. Instead, these efforts have become entangled in trying to solve the safety problems that extensibility itself introduces into OS design. We propose a pragmatic approach to extensibility, where kernel extensions are used in experimental settings to evaluate and develop OS enhancements for demanding applications. Once developed and well understood, these enhancements are then migrated into the base OS for production use. This approach obviates the need for guaranteeing the safety of kernel extensions, allowing the OS research community to re-focus on the real challenges in OS design and implementation. To provide a concrete example of this approach, we analyze the techniques used in experimental HTTP servers to show how proper application design combined with generic enhancements to OSs can provide the same benefits without requiring application-specific kernel extensions.

Towards understanding developing world traffic

While many projects aim to provide network access to the developing world or improve existing network access, relatively little data exists regarding the behavior of traffic in these environments, especially in regards to the characteristics of traffic in the developing world. In this paper, we provide a first glimpse into the traffic gathered by a worldwide proxy network, and try to observe differences in first-world and developing-world traffic characteristics. What sets this work apart from similar research is the scope and level of detail -- we capture more than 3TB of content representing one weeks browsing by 348K users across 190 countries. Capturing the content, rather than just access logs, also allows us to perform similarity analysis at the content level.

Understanding and characterizing PlanetLab resource usage for federated network testbeds

Global network testbeds are crucial for innovative network research. Built on the success of PlanetLab, the next generation of federated testbeds are under active development, but very little is known about resource usage in the shared infrastructures. In this paper, we conduct an extensive study of the usage profiles in PlanetLab that we collected for six years by running CoMon, a PlanetLab monitoring service. We examine various aspects of node-level behavior as well as experiment-centric behavior, and describe their implications for resource management in the federated testbeds. Our main contributions are threefold: (1) Contrary to common belief, our measurements show there is no tragedy of the commons in PlanetLab, since most PlanetLab experiments exploit the systems network reach more than just its hardware resources; (2) We examine resource allocation systems proposed for the federated testbeds, such as bartering and central banking schemes, and show that they would handle only a small percentage of the total usage in PlanetLab; and (3) Lastly, we identify factors that account for high resource contention or poor utilization in PlanetLab nodes. We analyze workload imbalance and problematic slices in PlanetLab, and describe the implications of our measurements for improving overall utility of the testbed.

The origins of network server latency & the myth of connection scheduling

We investigate the origins of server-induced latency to understand how to improve latency optimization techniques. Using the Flash Web server [4], we analyze latency behavior under various loads. Despite latency profiles that suggest standard queuing delays, we find that most latency actually originates from negative interactions between the application and the locking and blocking mechanisms in the kernel. Modifying the server and kernel to avoid these problems yields both qualitative and quantitative changes in the latency profiles -- latency drops by more than an order of magnitude, and the effective service discipline also improves.We find our modifications also mitigate service burstiness in the application, reducing the event queue lengths dramatically and eliminating any benefit from application-level connection scheduling. We identify one remaining source of unfairness, related to competition in the networking stack. We show that adjusting the TCP congestion window size addresses this problem, reducing latency by an additional factor of three.

Better flash access via shape-shifting virtual memory pages

Today, many system designers try to fit the entire data set of an application in RAM to avoid the cost of accessing magnetic disk. However, for many data-centric applications this is not an option due to the capacity and high