From: Chuck Reeves (creeves_at_windows.microsoft.com)
Date: Wed Feb 25 2004 - 15:15:08 PST
The paper "Manageability, Availability and Performance in Porcupine: A
Highly Scalable Internet Mail Service" was written by a number
researchers from the University of Washington in 1999.
This was the coolest system we've read about this semester.
The document describes a clustered email system, named Porcupine. I
genuinely appreciated the resilence of the design. Specifically, self
management. The ease with which Porcupine was able to introduce new
servers into the cluster and it's ability to handle component failures
with a reasonable level of degradation was impressive. In the
description of the inner workings of the system objects (mail fragment,
mail map, user map, cluster membership list,...) I was consistently
impressed at how simple the interaction between these components seemed
to be. Having no exposure to "Three Round Membership Protocol" or a
Lamport clock, the description of the membership service was a bit
difficult to understand, but the protocol seemed minimal.
The performance vs availability tradeoffs discussed surrounding
affinity-based scheduling and the spread-limiting load balancer were
paticularly insightful.
Chuck Reeves, creeves_at_microsoft.com
Microsoft | Windows | Directory Services
This archive was generated by hypermail 2.1.6 : Wed Feb 25 2004 - 15:15:18 PST