In addition to reporting on implementation details of Mitra, we present performance results that demonstrate the scalability characteristics of the system. We compare the obtained results with theoretical expectations based on the bandwidth of participating disks. Mitra attains between 65% to 100% of the theoretical expectations.
This paper characterizes the impact of disk seeks on the throughput of the system. It describes REBECA as a mechanism that maximizes the throughput of the system by minimizing the time attributed to each incurred seek. A limitation of REBECA is that it increases the latency observed by each request. We quantify this throughput vs latency tradeoff of REBECA and, develop an efficient technique that computes its configuration parameters to realize the performance requirements (desired latency and throughput) of an application.
For applications with large data sets, e.g., video servers, magnetic disks have established themselves as mass storage device of choice. A technique to increase the storage capacity of disks is zoning. A side-effect of zoning is that it introduces a disk drive with variable transfer rates. This paper describes techniques to support a continuous display of video objects using a multi-zone disk drive. As compared to previous approaches, the proposed techniques harness the average transfer rate of the magnetic disk (instead of its minimum transfer rate). In addition, we describe a configuration planner that logically manipulates the zoning information of a disk drive to support the performance criteria of an application.
The contributions of this study are two folds. First, it introduces a framework to store and retrieve "moving sensors" data. The framework advocates physical data independence and software-reuse. Second, we investigate alternative representations for storage and retrieval of data in support of query processing. We quantify the tradeoff associated with these alternatives using empirical data from RoboCup soccer matches.
This paper presents Controlled Buffer Sharing (CBS) as a novel framework that facilitates sharing and supports both a hiccup-free display and VCR operations. It includes a configuration planner and a buffer pool management technique (applied at run time). CBS trades memory for disk bandwidth in order to meet the performance objectives of an application and minimize cost per stream. It uses bridging and merges two displays referencing the same clip when they are $d_t$ blocks apart. One insight of this framework is that $d_t$ is determined by market forces (cost of memory and disk bandwidth) and is independent of a clip's frequency of access. We use both analytical and simulation models to quantify the characteristics of CBS.
Physical-location-independence means a plan will execute as long as a copy of its referenced WSs is available. This concept enables the client proxy objects to continue operation in the presence of both failures and WS migrations that balance system load.
This is joint-work with Tooraj Helmi, a Ph.D. student in the Computer Science department at USC.