How would you manage 500TB of spinning disk, 1PB of tape, and an IT infrastructure that delivers 12 billion (that's billion with a 'b') page views per day, all with a staff of three people? Welcome to Yahoo Storage Operations Manager Steve Curry's world.
In the first of two cases studies presented at Storage Decisions 2003 in Chicago, Curry described Yahoo's migration away from an initial, boot-strapped storage infrastructure that consisted of 160 14-disk JBODs. This inexpensive storage architecture was responsible for running Yahoo's own disk mirroring software and distributed file system.
When it became apparent to Curry that Yahoo's storage needs had quickly outgrown the inexpensive JBOD arrangement, he made the decision to upgrade the storage to include 371 network-attached storage (NAS) filers from Network Appliance.
In a second case study, Curry also discussed "a work in progress" -- the migration of Yahoo's global data mining and global log collection applications, away from a direct-attached storage (DAS) model, to a new model which provides better redundancy, lowers the total storage host count, and improves manageabilty by consolidating many small libraries into a single StorageTek PowderHorn library.
Curry's presentation includes conceptual diagrams that displays the basic architecture behind each consolidation project.
Presentation slides and other links to the full session proceedings are available
This was first published in September 2003