Jun Wang - US grants

An I/O Buffer Cache Architecture for Remote Direct Memory Access
Abstract

The project will investigate a low-cost buffer cache solution for improving performance of iSCSI end systems over Remote Direct Memory Access (RDMA). The use of RDMA protocols in modern storage networking architectures can reduce CPU overhead and performance loss due to the multiple memory-copy problem for end systems. An iSCSI-based IP storage architecture over RDMA can provide cost-effectiveness and high-performance.

There are significant challenges in smoothing out the disparities between iSCSI protocol and RDMA model, and reducing performance loss due to inefficient memory utilization. This research addresses these problems by conducting the following activities: 1) developing a unified buffer cache hardware/software architecture for iSCSI applications over RDMA protocol suite on both client and server sides by reducing the number of data packet translations across iSCSI and RDMA boundaries; and 2) developing propagate update protocols to resolve cache RDMA read and write failure problems.

The increasing demand for Exa-byte-scale storage capacity by high end computing applications requires a
higher level of scalability and dependability than that provided by current file and storage systems.
The proposal deals with file systems research for metadata management of scalable cluster-based parallel and distributed file storage systems in the HEC environment. It aims to develop a scalable and adaptive metadata management (SAM2) toolkit to extend features of and fully leverage the peak performance promised by state-of-the-art cluster-based parallel and distributed file storage systems used by the high performance computing community. There is a large body of research on data movement and management scaling, however, the need to scale up the attributes of cluster-based file systems and I/O, that is, metadata, has been underestimated. An understanding of the characteristics of metadata traffic, and an application of proper load-balancing, caching, prefetching and grouping mechanisms to perform metadata management correspondingly, will lead to a high scalability. It is anticipated that by appropriately plugging the scalable and adaptive metadata management components into the state-of-the-art cluster-based parallel and distributed file storage systems one could potentially increase the performance of applications and file systems, and help translate the promise and potential of high peak performance of such systems to real
application performance improvements.
The project involves the following components:
1. Develop multi-variable forecasting models to analyze and predict file metadata access patterns.
2. Develop scalable and adaptive file name mapping schemes using the duplicative Bloom filter array technique to enforce load balance and increase scalability
3. Develop decentralized, locality-aware metadata grouping schemes to facilitate the bulk metadata operations such as prefetching.
4. Develop an adaptive cache coherence protocol using a distributed shared object model for client-side and server-side metadata caching.
5. Prototype the SAM2 components into the state-of-the-art parallel virtual file system PVFS2 and a distributed storage data caching system, set up an experimental framework for a DOE CMS Tier 2 site at University of Nebraska-Lincoln and conduct benchmark, evaluation and validation studies.