Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (35 results)

  • Article
  • Peer Reviewed
Technical Reports (2002)
This paper explores the user-perceived Web performance of downloading entire pages, and how various common Web enhancements impact overall page performance. We use Medusa, a non-caching forwarding proxy, to collect user traces and replay them under various configurations of HTTP request optimizations. These optimizations include parallel and persistent connections, DNS caching, and the use of CDNs. We then use Medusa to characterize whole-page performance and measure the impact of request optimizations on downloading entire Web pages. We find that the most effective optimization is parallel connections. Other optimizations provide incremental benefits due to limited opportunity for use or due to limited benefit of the optimization itself. When there is more opportunity for use, we find that optimizations in the former case provide substantial benefit.

Pre-2018 CSE ID: CS2002-0729

    Cover page: Whole Page Performance
    • Article
    • Peer Reviewed
    Technical Reports (2002)
    As applications executing in virtual machine runtime environments such as the Java Virtual Machine and the .NET Common Language Runtime become more prevalent on desktop operating systems, this trend raises the interesting question of the best model for supporting multiple applications in virtual machine environments. A natural model is to execute each application in its own virtual machine process. However, multiple applications could also be executed in a single virtual machine process by leveraging the protection and security features of the programming languages used by the applications. Executing within a single virtual machine process makes it easier to share code and data and communicate among the processes. In this paper, we present an intermediate model for supporting multiple applications in virtual machine environments for desktop operating systems. We argue that applications should take advantage of the protection features and resource management provided by the operating system and execute each application in its own virtual machine process. However, to facilitate code reuse, reduce application initialization time, and facilitate interprocess communication among applications in different virtual machines, we propose extending virtual machine implementations with the use of a shared class cache and an efficient serialization implementation optimized for local machine interprocess communication. To demonstrate and evaluate our approach, we describe the design and implementation of a shared class cache and optimized class serialization in the Java virtual machine, and evaluate our implementation using a set of application and micro-benchmarks.

    Pre-2018 CSE ID: CS2002-0728

      Cover page: Interaction of Virtual Machine with the Operating System
      • Article
      • Peer Reviewed
      Technical Reports (2004)
      In this paper, we analyze the mobility patterns of users of wireless handheld PDAs in a campus wireless network using an 11 week trace of wireless network activity. Our study has three goals. First, we characterize the high-level mobility and access patterns of handheld PDA users and compare these characteristics to previous workload mobility studies focused on laptop users. Second, we develop two wireless network topology models for use in wireless mobility studies: an "evolutionary topology model" based on user proximity and a "campus waypoint" model that serves as a trace-based complement to the random waypoint model. Finally, we use our wireless network topology models as a case study to evaluate ad-hoc routing algorithms on the network topologies created by the access and mobility patterns of users of modern wireless PDAs.

      Pre-2018 CSE ID: CS2004-0780

        Cover page: Access and Mobility of Wireless PDA Users
        • Article
        • Peer Reviewed
        Technical Reports (2006)
        Distributed hash tables are increasingly being proposed as the core substrate for content delivery applications in the Internet, such as cooperative Web caches, Web index and search, and content delivery systems. The performance of these applications built on DHTs fundamentally depends on the effectiveness of request routing within the DHT. In this paper, we show how to use soft state to achieve routing performance that approaches the aggressive performance of one-hop schemes, but with an order of magnitude less overhead on average. We use three kinds of hint caches to improve routing latency: local hint caches, path hint caches, and global hint caches. Local hint caches use large successor lists to short cut final hops. Path hint caches store a moderate number of effective route entries gathered while performing lookups for other nodes. And global hint caches store direct routes to peers distributed across the ID space. Based upon our simulation results, we find that the combination of the hint caches significantly improves Chord routing performance: in a network of 4,096 peers, the hint caches enable Chord to route requests with average latencies only 6% more than algorithms that use complete routing tables with significantly less overhead.

        Pre-2018 CSE ID: CS2006-0862

          Cover page: ShortCuts: Using Soft State To Improve DHT Routing
          • Article
          • Peer Reviewed
          Technical Reports (2006)
          In this paper, we discuss the problem of storage reclamation in DHTs that use redundancy to provide high data availability. Two approaches for performing storage reclamation for lazy repair are active polling and implicit garbage collection. We evaluated the overhead of these two approaches under low churn situation to high. The garbage collection is most effective for high churn peer to peer environments and the active approach is more suitable for low churn systems like educational/corporate environments.

          Pre-2018 CSE ID: CS2006-0863

            Cover page: Resource Reclamation in Distributed Hash Tables
            • Article
            • Peer Reviewed
            Technical Reports (2008)
            In this paper we describe Cumulus, a system for efficiently implementing filesystem system backups over the Internet. Cumulus is specifically designed under a thin cloud assumption--that the remote datacenter storing the backups does not provide any special backup services, but only provides a least-common-denominator storage interface (i.e., get and put of complete files). While Cumulus can thus use virtually any storage service, we show that it still provides efficiencies comparable to integrated approaches.

            Pre-2018 CSE ID: CS2008-0927

              Cover page: Cumulus: Filesystem Backup to the Cloud
              • Thesis
              • Peer Reviewed
              UC San Diego Electronic Theses and Dissertations (2022)

              Serverless computing is a new computing paradigm that aims to simplify building cloud applications. Existing serverless computing is designed for certain types of workloads, such as stateless web request handlers that require little coordination across instances. In this thesis I explore the potential of general purpose serverless computing by providing new features at both the application framework level and the serverless platform level.

              I first describe Sprocket, a parallel video application framework based on serverless computing. Based on my work with Sprocket, I observed several key limits of existing serverless offerings, including lack of cross-instance networking, inefficient resource scheduling, and the cold start problem. I develop three systems for the serverless platform: Particle, Sophon, and FaaSnap, that tackle networking, scheduling, and cold start challenges, respectively. These systems reduce the gap between the requirements of emerging serverless applications and the capabilities of current serverless offerings, paving the way for future general purpose serverless computing.

                Cover page: Framework and Platform Support for General Purpose Serverless Computing
                • Thesis
                • Peer Reviewed
                UC San Diego Electronic Theses and Dissertations (2023)

                Security is an ever growing concern for daily Internet users, especially sincemany facets of a user's daily interactions (banking, commerce, workplace) are now accessed via the Internet. Fortunately, recent technical advancements – such as encrypted web browsing, email spam filtering, and login two factor authentication – have increased the accessibility and practicality of security for users. However, studies show that the majority of exploited attacks take advantage of the human in the loop. Technology and humans are required to work in harmony for security to be effective. As a result, it is crucial that we understand the extent to which users follow best practices, and that we evaluate whether their behaviors in fact help prevent adverse security outcomes. In this dissertation, I argue that large-scale empirical measurement is a practical and effective technique to answer these questions as the basis for prioritizing security practices, and I support this argument with three different projects. First, I use network traffic data and measurement methods to quantify user behavior ``best practices'' and how they relate to an outcome (in this case, compromise). Next I examine how communication about a security policy change can affect an organization by analyzing large-scale organizational data. Finally, I quantify attacker behavior in the “Hack for Hire” market by hiring and monitoring attackers, which provides insight into which defenses to prioritize for better protecting users from these types of attacks. By empirically understanding and prioritizing effective security practices, we can further improve security for users.

                  Cover page: Prioritizing Security Practices via Large-Scale Measurement of User Behavior
                  • Article
                  • Peer Reviewed
                  Technical Reports (2007)
                  To evade blacklisting, the vast majority of spam email is sent from exploited MTAs (i.e., botnets) and with forged "From" addresses. In response, the anti-spam community has developed a number of domain-based authentication systems -- such as SPF and DKIM -- to validate the binding between individual domain names and legitimate mail sources for those domains. In this paper, we explore an alternative solution in which the mail recipient requests a real-time affirmation for each e-mail from the declared sender's MX of record. The "Occam" protocol is trivial to implement, offers authenticating power equivalent to SPF and DKIM and, most importantly, forces spammers to deploy and expose blacklistable servers for each domain they use during a campaign. We discuss the details of the protocol, compare its strengths and weaknesses with existing solutions and describe a prototype implementation in Sendmail.

                  Pre-2018 CSE ID: CS2007-0893

                    Cover page: Slicing Spam with Occam's Razor
                    • Article
                    • Peer Reviewed
                    Technical Reports (2002)
                    When developing fault-tolerant protocols, systems are usually modeled assuming that process failures are independent and identically distributed. In this paper, we present a system model that can represent correlated failures. We show that such a model is useful in that protocols can be made more efficient. Central to our approach is the idea of a core, which is a reliable minimal subset of processes. We present two probabilistic failure models for dependent failures and discuss them in terms of computing cores. For both, finding a smallest minimal core is NP-hard, but one of the two models can be strengthened in a natural way to make computing a minimal core in P.

                    Pre-2018 CSE ID: CS2002-0723

                      Cover page: Coping with Dependent Process Failures
                      Top