Thursday, April 26, 2001
Noon - 1 pm
Hammerschlag Hall D210
Carnegie Mellon University
My Cache or Yours?
Modern high-end disk arrays typically have several gigabytes of cache RAM.
Unfortunately, the array cache is often inclusive: interactions
between client and array cache management policies cause the array cache
to duplicate some of the contents of the client cache. As a result, the
aggregate cache behaves as if it was only as big as the larger of the
client and array caches, instead of as large as the sum of the two.
Our preliminary research explores the potential benefits of making the
array cache exclusive. We implement modified client and array cache
management policies to ensure that data blocks are placed at either the
client or the array, but not both. Exclusivity thus creates a single,
large unified cache. We present the results of a simulation study that show up
to 7.5x speedups by switching from inclusive to exclusive caches for a set of
synthetic and real-life workloads.
This research was conducted with Greg Ganger (Dept. of ECE, CMU) and John
Wilkes (Hewlett-Packard Laboratories), with the support of the Storage
Systems Program at HPL and the members of the Parallel Data Consortium.
Ted is a 4th year Ph.D. student in the Computer Science Department at
Carnegie Mellon University. He completed a BA in Engineering Science from
Oxford University in 1993, and an M.Eng in Computer Science from Cornell
University in 1995. He has been part of several research projects since
coming to CMU, but has settled down to work on dynamic recovery for
survivable storage systems.
For Further Seminar Info: