Tuesday, January 24, 2012
Are we going to see Hadoop on Tape?
I found some fascinating video and pdf with a contrarian analysis of how Tape vendors might see higher growth rates than Disk vendors over the next few years. Given the Hadoop I/O pattern of sequentially reading gigabytes of data, followed by sequentially writing gigabytes of data, will there be a point at which (at least some subset of) Hadoop jobs run directly off tape?
Subscribe to:
Post Comments (Atom)
1 comment:
Some extra info, a year old IBM announcement of 35TB uncompressed date on a single tape: http://www.itjungle.com/tfh/tfh012510-story10.html .
Post a Comment