No abstract available.

Cloud computing demands cheap, always-on, and reliable storage. We describe Pahoehoe, a key-value cloud storage system we designed to store large objects cost-effectively with high availability. Pahoehoe stores objects across multiple data centers and provides eventual consistency so to be available during network partitions. Pahoehoe uses erasure codes to store objects with high reliability at low cost. Its use of erasure codes distinguishes Pahoehoe from other cloud storage systems, and presents a challenge for efficiently providing eventual consistency. We describe Pahoehoe's put, get, and convergence protocols-convergence being the decentralized protocol that ensures eventual consistency. We use simulated executions of Pahoehoe to evaluate the efficiency of convergence, in terms of message count and message bytes sent, for failure-free and expected failure scenarios (e.g., partitions and server unavailability). We describe and evaluate optimizations to the naïve convergence protocol that reduce the cost of convergence in all scenarios.

We present two new cooperative caching algorithms that allow a cluster of file system clients to cache chunks of files instead of directly accessing them from origin file servers. The first algorithm, called C-LRU (Cooperative-LRU), is based on the simple D-LRU (Distributed-LRU) algorithm, but moves a chunk's position closer to the tail of its local LRU list when the number of copies of the chunk increases. The second algorithm, called RobinHood, is based on the N-Chance algorithm, but targets chunks cached at many clients for replacement when forwarding a singlet to a peer. We evaluate these algorithms on a variety of workloads, including several publicly available traces, and find that the new algorithms significantly outperform their predecessors.

Distributed systems are notoriously difficult to implement and debug. One important tool for understanding the behavior of distributed systems is tracing. Unfortunately, effective tracing for modern distributed systems faces several challenges. First, many interesting behaviors in distributed systems only occur rarely, or at full production scale. Hence we need tracing mechanisms which impose minimal overhead, in order to allow always-on tracing of production instances. Second, for high-speed systems, messages can be delivered in significantly less time than the error of traditional time synchronization techniques such as network time protocol (NTP), necessitating time adjustment techniques with much higher precision. Third, distributed systems today may generate millions of events per second systemwide, resulting in traces consisting of billions of events. Such large traces can overwhelm existing trace analysis tools. These challenges make effective tracing difficult. We present techniques that address these three challenges. Our contributions include (1) a low-overhead tracing mechanism, which allows tracing of large systems without impacting their behavior or performance (0.14 ¿s/event), (2) a post hoc technique for producing highly accurate time synchronization across hosts (within 10/ts, compared to between 100 ¿s to 2 ms for NTP), and (3) incremental data processing techniques which facilitate analyzing traces containing billions of trace points on desktop systems. We have successfully applied these techniques to two distributed systems, a cooperative caching system and a distributed storage system, and from our experience, we believe our techniques are applicable to other distributed systems.

No abstract available.

We describe a new class of utility-maximization scheduling problem with precedence constraints, the disconnected staged scheduling problem (DSSP). DSSP is a nonpreemptive multiprocessor deadline scheduling problem that arises in several commercially-important applications, including animation rendering, protein analysis, and seismic signal processing. DSSP differs from most previously-studied deadline scheduling problems because the graph of precedence constraints among tasks within jobs is disconnected, with one component per job. Another difference is that in practice we often lack accurate estimates of task execution times, and so purely offline solutions are not possible. However we do know the set of jobs and their precedence constraints up front and therefore some offline planning is possible.Our solution decomposes DSSP into an offline job selection phase followed by an online task dispatching phase. We model the former as a knapsack problem and explore several solutions to it, describe a new dispatching algorithm for the latter, and compare both with existing methods. Our theoretical results show that while DSSP is NP-hard and inapproximable in general, our two-phase scheduling method guarantees a good performance bound for many special cases. Our empirical results include an evaluation of scheduling algorithms on a real animation-rendering workload; we present a characterization of this workload in a companion paper. The workload records eight weeks of activity on a 1,000-CPU cluster used to render portions of the full-length animated feature film Shrek 2 in 2004. We show that our improved scheduling algorithms can substantially increase the aggregate value of completed jobs compared to existing practices. Our new task dispatching algorithm LCPF performs well by several metrics, including job completion times as well as the aggregate value of completed jobs.

Despite the importance of storage in enterprise computer systems, there are few adequate tools to design and configure a storage system to meet application data requirements efficiently. Storage system design involves choosing the disk arrays to use, setting the configuration options on those arrays, and determining an efficient mapping of application data onto the configured system. This is a complex process because of the multitude of disk array configuration options, and the need to take into account both capacity and potentially contending I/O performance demands when placing the data. Thus, both existing tools and administrators using rules of thumb often generate designs that are of poor quality.This article presents the Disk Array Designer (DAD), which is a tool that can be used both to guide administrators in their design decisions and to automate the design process. DAD uses a generalized best-fit bin packing heuristic with randomization and backtracking to search efficiently through the huge number of possible design choices. It makes decisions using device models that estimate storage system performance. We evaluate DAD's designs based on traces from a variety of database, filesystem, and e-mail workloads. We show that DAD can handle the difficult task of configuring midrange and high-end disk arrays, even with complex real-world workloads. We also show that DAD quickly generates near-optimal storage system designs, improving in both speed and quality over previous tools.

Disk drives are now available with capacity and price per capacity comparable to nearline tape systems. Because disks have superior performance, density and maintainability characteristics, it seems likely that they will soon overtake tapes as the backup medium of choice. The authors outline the potential advantages of a backup system composed of high-capacity disk drives and describe what implications such a system would have for backup software.