🚀 go-pugleaf

vhs@rhein-main.de (Volker Herminghaus-Shirai) writes:

>I need to design a retrieval system for ~15TB of data, of which ~5TB are
>retrieved with a high frequency (25 requests/second) and an access time
>of avg. 30 seconds (max. 120 seconds). Retrieval lacks any locality, so the
>5TB are real random-access. The rest of the data is still accessed at a
>rate of 5 requests/second. Here, too, access is completely random.
>So far I've come up with the following ideas:

>1) Use large (10GB) WORM jukeboxes with many pickers (~40),
>   many drives (~150) and a *real* intelligent scheduler. No cache.

>2) Use 1) for the less-often-retrieved data, and several 625MB-OD
>   jukeboxes with even more drives and a less intelligent scheduler.

>I figure that doing some 25 media changes a second would be a real
>performance killer. Due to the response time restrictions (30 seconds)
>there is hardly a chance to build a decent queue for pending requests
>and schedule them appropriately.
>So what we really need is 5TB of something like a huge, if slow, hard disk.
>Is there any such kind of device? Any other ideas? Any help?

>Any kind of information is greatly appreciated (hints, pointers, ideas, etc.)

>--
>Volker Herminghaus-Shirai (vhs@rhein-main.de)

Seeing all other expensive solutions being proposed, have you looked at
Creo's optical tape units? Capacity is around a terabyte I believe, and 
access time is said to be 1 minite for the whole tape. The drives are 
however in the $250k range I believe. Since Creo is in town here send
e-mail if you need an address.

Peter Van Epp / Operations and Technical Support 
Simon Fraser University, Burnaby, B.C. Canada