This thesis comprises of an in-depth investigation on issues related to high performance I/O architectures, including I/O caching, disk architectures as well as web servers.
The first part of this thesis presents a novel high-performance, low-cost disk architecture called DCD, Disk Caching Disk. DCD consists of a three-level storage hierarchy: a RAM buffer and a log disk and a data disk. Trace-driven simulation of DCD under the office/engineering workload demonstrates up to two orders of magnitude performance improvement for small writes when compared to existing disk systems. Furthermore, DCD is very reliable, and works in device or device driver level. As a result, it can be applied directly to current file systems without the need of changing underlying operation systems.
The second part of this thesis presents a new cache architecture called RAPID-Cache for Redundant, Asymmetrically Parallel, and Inexpensive Disk Cache. RAPID-Cache is a highly-reliable and inexpensive write cache for high-performance storage systems. A typical RAPID-Cache consists of two redundant write buffers on top of a disk system. One of the buffers is a primary cache made of RAM or NVRAM and the other is a backup cache containing a two level hierarchy: a small NVRAM buffer on top of a log disk. Our analysis and trace-driven simulation results show that the RAPID-Cache has significant reliability/cost advantages over conventional single NVRAM write caches and has great cost advantages over dual-copy NVRAM caches.
The third part of this thesis presents an in-depth study on I/O issues related to the Apache web server. Very few results have been published that quantitatively study the server behavior. Using standard benchmarks and several OS kernel tracing facilities, we quantitatively identified the server performance bottlenecks. We then proposed and implemented 7 techniques that improve the performance of Apache by 61%. Finally, our results suggest that operating system support for directly sending data from the file system cache to the TCP/IP network can further improve the Web server performance dramatically.
