Person:

Seltzer, Margo

Flash memory has recently become popular as a caching medium. Most uses to date are on the storage server side. We investigate a different structure: flash as a cache on the client side of a networked storage environment. We use trace-driven simulation to explore the design space. We consider a wide range of configurations and policies to determine the potential client-side caches might offer and how best to arrange them. Our results show that the flash cache writeback policy does not significantly affect performance. Write-through is sufficient; this greatly simplifies cache consistency handling. We also find that the chief benefit of the flash cache is its size, not its persistence. Cache persistence offers additional performance benefits at system restart at essentially no runtime cost. Finally, for some workloads a large flash cache allows using miniscule amounts of RAM for file caching (e.g., 256 KB) leaving more memory available for application use.

As the prevalence of blogs, discussion forums, and online news services continues to grow, so too does the portion of this Web content that relates to health and medicine. We propose that everyday, medically-oriented Web content is a valuable and viable data source for medical hypothesis generation and testing, despite its being noisy. In this paper, we present a proof-of-concept system supporting this notion. We construct a corpus comprising news articles relating to the drugs Vioxx, Naproxen and Ibuprofen, that were published between 1998-2002. Using this corpus, we show that there was a significant link between Vioxx and the concept “Myocardial Infarction” well before the drug was withdrawn from the market in 2004. Indeed, within the Vioxx-related content, the concept ranks amongst the top 3.3% in terms of importance. When compared with the Naproxen and Ibuprofen control literatures, the term occurs significantly more frequently in the Vioxx-related content.

The cloud is poised to become the next computing environment for both data storage and computation due to its pay-as-you-go and provision-as-you-go models. Cloud storage is already being used to back up desktop user data, host shared scientific data, store web application data, and to serve web pages. Today’s cloud stores, however, are missing an important ingredient: provenance. Provenance is metadata that describes the history of an object. We make the case that provenance is crucial for data stored on the cloud and identify the properties of provenance that enable its utility. We then examine current cloud offerings and design and implement three protocols for maintaining data/provenance in current cloud stores. The protocols represent different points in the design space and satisfy different subsets of the provenance properties. Our evaluation indicates that the overheads of all three protocols are comparable to each other and reasonable in absolute terms. Thus, one can select a protocol based upon the properties it provides without sacrificing performance. While it is feasible to provide provenance as a layer on top of today’s cloud offerings, we conclude by presenting the case for incorporating provenance as a core cloud feature, discussing the issues in doing so.

We introduce StarFlow, a script-centric environment for data analysis. StarFlow has four main features: (1) extraction of control and data-flow dependencies through a novel combination of static analysis, dynamic runtime analysis, and user annotations, (2) command-line tools for exploring and propagating changes through the resulting dependency network, (3) support for workflow abstractions enabling robust parallel executions of complex analysis pipelines, and (4) a seamless interface with the Python scripting language. We describe a range of real applications of StarFlow, including automatic parallelization of complex workflows in the cloud.

Digital provenance is meta-data that describes the ancestry or history of a digital object. Most work on provenance focuses on how provenance increases the value of data to consumers. However, provenance is also valuable to storage providers. For example, provenance can provide hints on access patterns, detect anomalous behavior, and provide enhanced user search capabilities. As the next generation storage providers, cloud vendors are in the unique position to capitalize on this opportunity to incorporate provenance as a fundamental storage system primitive. To date, cloud offerings have not yet done so. We provide motivation for providers to treat provenance as first class data in the cloud and based on our experience with provenance in a local storage system, suggest a set of requirements that make provenance feasible and attractive.

While there has been a great deal of research on provenance systems, there has been little discussion about challenges that arise when making different provenance systems interoperate. In fact, most of the literature focuses on provenance systems in isolation and does not discuss interoperability – what it means, its requirements, and how to achieve it. We designed the Provenance-Aware Storage System to be a general- purpose substrate on top of which it would be “easy” to add other provenance-aware systems in a way that would provide “seamless integration” for the provenance captured at each level. While the system did exactly what we wanted on toy problems, when we began integrating StarFlow, a Python-based workflow/provenance system, we discovered that integration is far trickier and more subtle than anyone has suggested in the literature. This work describes our experience undertaking the integration of StarFlow and PASS, identifying several important additions to existing provenance models necessary for interoperability among provenance systems.

We describe our experiences importing PASS [16] provenance into PLUS [7]. Although both systems import and export provenance that conforms to the Open Provenance Model (OPM) [14], the two systems vary greatly with respect to the granularity of provenance captured, how much semantic knowledge the system contributes, and the completeness of provenance capture. We encountered several problems reconciling provenance between the two systems and use that experience to specify a Common Provenance Framework, that provides a higher degree of interoperability between provenance systems. In each case, the problems stem from the fact that OPM interoperability is a weaker requirement than query interoperability. Our goal in presenting this work is to generate discussion about differing degrees of interoperability and the requirements thereof.

Processor-level dynamic thermal management techniques have long targeted worst-case thermal margins. We examine the thermal-performance trade-offs in average-case, preventive thermal management by actively degrading application performance to achieve long-term thermal control. We propose Dimetrodon, the use of idle cycle injection, a flexible, per-thread technique, as a preventive thermal management mechanism and demonstrate its efficiency compared to hardware techniques in a commodity operating system on real hardware under throughput and latency-sensitive real-world workloads. Compared to hardware techniques that also lack flexibility, Dimetrodon achieves favorable trade-offs for temperature reductions up to 30% due to rapid heat dissipation during short idle intervals.

The Provenance Aware Storage Systems project (PASS) currently collects system-level provenance by intercepting system calls in the Linux kernel and storing the provenance in a stackable filesystem. While this approach is reasonably efficient, it suffers from two significant drawbacks: each new revision of the kernel requires reintegration of PASS changes, the stability of which must be continually tested; also, the use of a stackable filesystem makes it difficult to collect provenance on root volumes, especially during early boot. In this paper we describe an approach to collecting system-level provenance from virtual guest machines running under the Xen hypervisor. We make the case that our approach alleviates the aforementioned difficulties and promotes adoption of provenance collection within cloud computing platforms.

Upcoming multicore processors, with hundreds of cores or more in a single chip, require a degree of parallel scalability that is not currently available in today’s system software. Based on prior experience in the super-computing sector, the likely trend for multicore processors is away from shared memory and toward shared-nothing architectures based on message passing. In light of this, the lightweight messages and channels programming model, found among other places in Erlang, is likely the best way forward. This paper discusses what adopting this model entails, describes the architecture of an OS based on this model, and outlines a few likely implementation challenges.