Steven W. Schlosser

Designing computer systems with MEMS-based storage

For decades the RAM-to-disk memory hierarchy gap has plagued computer architects. An exciting new storage technology based on microelectromechanical systems (MEMS) is poised to fill a large portion of this performance gap, significantly reduce system power consumption, and enable many new applications. This paper explores the system-level implications of integrating MEMS-based storage into the memory hierarchy. Results show that standalone MEMS-based storage reduces I/O stall times by 4-74X over disks and improves overall application runtimes by 1.9-4.4X. When used as on-board caches for disks, MEMS-based storage improves I/O response time by up to 3.5X. Further, the energy consumption of MEMS-based storage is 10-54X less than that of state-of-the-art low-power disk drives. The combination of the high-level physical characteristics of MEMS-based storage (small footprints, high shock tolerance) and the ability to directly integrate MEMS-based storage with processing leads to such new applications as portable gigabit storage systems and ubiquitous active storage nodes.

Log-based architectures for general-purpose monitoring of deployed code

Runtime monitoring tools are invaluable for detecting various types of bugs, in both sequential and multi-threaded programs. However, these tools often slow down the monitored program by an order of magnitude or more [4], implying that the tools are ill-suited for always-on monitoring of deployed code. Fortunately, the emergence of chip multiprocessors as a dominant computing platform means that resources are available on-chip to assist in monitoring tasks. In this brief note, we advocate Log-Based Architectures (LBA) that exploit such on-chip resources in order to dramatically reduce the overhead of runtime program monitoring. Specifically, we propose adding hardware support for logging a main programs trace and delivering it to another (otherwise idle) processing core for inspection. A life-guard program running on this other core executes the desired monitoring task.

Tashi: location-aware cluster management

Big Data applications, those that require large data corpora either for correctness or for fidelity, are becoming increasingly prevalent. Tashi is a cluster management system designed particularly for enabling cloud computing applications to operate on repositories of Big Data. These applications are extremely scalable but also have very high resource demands. A key technique for making such applications perform well is Location-Awareness. This paper demonstrates that location-aware applications can outperform those that are not location aware by factors of 3-11 and describes two general services developed for Tashi to provide location-awareness independently of the storage system.

Clotho: decoupling memory page layout from storage organization

As database application performance depends on the utilization of the memory hierarchy, smart data placement plays a central role in increasing locality and in improving memory utilization. Existing techniques, however, do not optimize accesses to all levels of the memory hierarchy and for all the different workloads, because each storage level uses different technology (cache, memory, disks) and each application accesses data using different patterns. Clotho is a new buffer pool and storage management architecture that decouples in-memory page layout from data organization on non-volatile storage devices to enable independent data layout design at each level of the storage hierarchy. Clotho can maximize cache and memory utilization by (a) transparently using appropriate data layouts in memory and non-volatile storage, and (b) dynamically synthesizing data pages to follow application access patterns at each level as needed. Clotho creates in-memory pages individually tailored for compound and dynamically changing workloads, and enables efficient use of different storage technologies (e.g., disk arrays or MEMS-based storage devices). This paper describes the Clotho design and prototype implementation and evaluates its performance under a variety of workloads using both disk arrays and simulated MEMS-based storage devices.

SLIPstream: scalable low-latency interactive perception on streaming data

A critical problem in implementing interactive perception applications is the considerable computational cost of current computer vision and machine learning algorithms, which typically run one to two orders of magnitude too slowly to be used interactively. Fortunately, many of these algorithms exhibit coarse-grained task and data parallelism that can be exploited across machines. The SLIPstream project focuses on building a highly-parallel runtime system called Sprout that can harness the computing power of a cluster to execute perception applications with low latency. This paper makes the case for using clusters for perception applications, describes the architecture of the Sprout runtime, and presents two compute-intensive yet interactive applications.

Materialized community ground models for large-scale earthquake simulation

Large-scale earthquake simulation requires source datasets which describe the highly heterogeneous physical characteristics of the earth in the region under simulation. Physical characteristic datasets are the first stage in a simulation pipeline which includes mesh generation, partitioning, solving, and visualization. In practice, the data is produced in an ad-hoc fashion for each set of experiments, which has several significant shortcomings including lower performance, decreased repeatability and comparability, and a longer time to science, an increasingly important metric. As a solution to these problems, we propose a new approach for providing scientific data to ground motion simulations, in which ground model datasets are fully materialized into octress stored on disk, which can be more efficiently queried (by up to two orders of magnitude) than the underlying community velocity model programs. While octrees have long been used to store spatial datasets, they have not yet been used at the scale we propose. We further propose that these datasets can be provided as a service, either over the Internet or, more likely, in a data center or supercomputing center in which the simulations take place. Since constructing these octrees is itself a challenge, we present three data-parallel techniques for efficiently building them, which can significantly decrease the build time from days or weeks to hours using commodity clusters. This approach typifies a broader shift toward science as a service techniques in which scientific computation and storage services become more tightly intertwined.

MultiMap: Preserving disk locality for multidimensional datasets

MultiMap is an algorithm for mapping multidimensional datasets so as to preserve the datas spatial locality on disks. Without revealing disk-specific details to applications, MultiMap exploits modern disk characteristics to provide full streaming bandwidth for one (primary) dimension and maximally efficient non-sequential access (i.e., minimal seek and no rotational latency) for the other dimensions. This is in contrast to existing approaches, which either severely penalize non-primary dimensions or fail to provide full streaming bandwidth for any dimension. Experimental evaluation of a prototype implementation demonstrates MultiMaps superior performance for range and beam queries. On average, MultiMap reduces total I/O time by over 50% when compared to traditional linearized layouts and by over 30% when compared to space-filling curve approaches such as Z-ordering and Hilbert curves. For scans of the primary dimension, MultiMap and traditional linearized layouts provide almost two orders of magnitude higher throughput than space-filling curve approaches.

Enabling database-aware storage with OSD

The ANSI object-based storage device (OSD) standard is a major step toward enabling explicit application-awareness in storage systems behind a standard, fully- interoperable interface [3]. In this paper, we explore a particular flavor of application-awareness, that of database applications. We describe the design and implementation of a database-aware storage system that uses the OSD interface not only as a means to access data, but also to permit explicit communication between the application and the storage system. This communication is significant, as it enables our storage system to transparently optimize data placement and request scheduling. We demonstrate that OSD makes it practical to improve storage performance in these ways without exposing proprietary disk drive parameters to application code, and without labor-intensive, fragile parameter measurement.

The unique strengths and storage access characteristics of discard-based search

Discard-based searchis a new approach to searching the content of complex, unlabeled, nonindexed data such as digital photographs, medical images, and real-time surveillance data. The essence of this approach is query-specific content-based computation, pipelined with human cognition. In this approach, query-specific parallel computation shrinks a search task down to human scale, thus allowing the expertise, judgment, and intuition of an expert to be brought to bear on the specificity and selectivity of the search. In this paper, we report on the lessons learned in the Diamond projectfrom applying discard-based search to a variety of applications in the health sciences. From the viewpoint of a user, discard-based search offers unique strengths. From the viewpoint of server hardware and software, it offers unique opportunities for optimization that contradict long-established tenets of storage design. Together, these distinctive end-to-end attributes herald a new genre of Internet applications.