Scientific Computing Upgrade Rescheduled for October

[Sandy Philpott <Sandy.Philpott@jlab.org>]

To match today's announcement that the Scheduled Accelerator Down will
be moved to October, these dates will be changed to October 3-4 for the
JASMine/Auger outage, with production testing through October 9.

> During the accelerator down time in September, the Scientific 
> Computing systems will be shut down for planned upgrades. All JASMine 
> and Auger systems, including all mass storage (/mss, /work, /cache) 
> and batch and interactive farm nodes (IFARM*, FARM*), will be 
> unavailable.
>
> The JASMine/Auger outage is scheduled for September 6-7, to install a 
> new Central Dispatcher for JASMine data movement, and to update data 
> mover clients to use the new dispatcher. The system will then undergo 
> production testing through September 12.
>
> This upgrade will increase data throughput for the mass storage 
> system, and paves the way for the upcoming project to compress data 
> and reuse tapes, saving a significant amount of money over purchasing 
> new blank tapes. Another important change is the increase in supported 
> file sizes from 2GB to 10GB.
>
> The tape copy/compression project will begin shortly thereafter. This 
> project reuses tapes previously written at a lower density of 60GB per 
> tape with StorageTek 9940A tape drives. The same media can store 200GB 
> per tape when written with our current 9940B tape drives. 
> Approximately 8000 9940A-formatted tapes will be rewritten with the 
> 9940B drives, freeing over 5000 tapes for reuse at the higher density. 
> At approximately $80 per tape, we're eager to get started!
>
> The mass storage system at JLab currently stores 1.3 petabytes of data 
> in two StorageTek Powderhorn 9310 silos that hold almost 6000 tapes 
> each. With less than 10% of slots currently unused, we would only be a 
> short time away from exceeding capacity. When replaced with all 200GB 
> tapes, the silo capacity reaches just over 2PB. Each year the amount 
> of data written by the experiments increases, with an expected 400 
> terabytes (0.4PB) to be written during this fiscal year. Current 
> planning includes a new silo purchase early in FY08, with some of the 
> oldest data being moved offline in FY07.
>