Doing some background research for a blog posting on repository
services and the cloud (repositoryman.blogspot.com), I tried to work
out how much it would cost to run eprints.ecs.soton.ac.uk using
Amazon's S3 data store. The answer is "about $250 per annum", based on
the following measurements:
0.1 Gb / month uploads
10 Gb current data storage size
100Gb / month downloads (including spiders)
Ratio of upload bandwidth to download bandwidth: 1:1000
Ration of storage size to monthly downloads: 1:10
Obviously, different repositories at different stages of development
and serving different purposes will have different figures for each of
these headings, but I wondered whether these kinds of extreme ratios
are the norm - for repositories and for web sites? Does anyone have
any evidence or know anything more about this? Cloud services like
Amazon charge fairly small amounts for the storage of data, but
significantly more for its movement across cloud boundaries (I suppose
that would make them "silver linings"!), so the ability to make
estimates about the use of holdings might be useful.
--
Les
|