Re: On/Near-line Archiving - Suggestions?
Posted in 1994
> > We are developing a system to archive historical data from our > Informix databases. The objectives we are shooting for are: > > * On-line or Near-line access of historical data. > * An initial capacity of 34GB with growth rates of 10GB per year. > * 2-3 concurrent users accessing historical data. > > We are considering building one or more databases with a dbspace per > platter [in cooked filesystems] in a multi-drive optical jukebox. > > If you have any comments, experiences or suggestions you could offer, > please contact me. > > Our environment: > > * RS/6000's with Informix OnLine/I-Star 5.0x and 4.1x tools. > > -- > Dave Newell _/ _/ _/_/_/ _/_/_/ _/_/_/ > dbn@alert.com/(303)488-7781 _/ _/ _/ _/_/ _/ _/ _/ > Mgr/Systems & Database Admin. _/_/_/ _/ _/ _/_/ _/ > Alert Centre, Inc. _/ _/ _/_/_/ _/_/_/ _/ _/ _/ > Dave (and anybody else who cares), We have an archive program which a) reads the schema and determines the names of all tables which possess certain fields then b) determines which rows have been tagged as 'OBSOLETE' then c) marks each one and notifies the owner of the data that their data is about to go (waits 5 days) and then unloads all of said tables into a temp directory (keeping track of the schema version this data uses) then does a compress on all of the unload files then does an 'ar' and creates a library of all the data. When it becomes necessary to restore said data, the schema is tested against the current schema - if any field changes are unresolvable automatically then only an administrator can restore it, but it then generates either 'LOAD' statements for tables whose schemas have not changed, or DBLOAD command files for those tables which have changed - and need special processing and then loads all of the data back in. This data is tagged for deletion after 'n' days (at restore time). The space requirement is about 6%-9% of the original data. The archive subsystem has been running for about 16 months with no loss of data. We have had some minor problems with schema changes - where the operators cannot restore automatically because of schema changes and have to find an admin - but the system has worked very well. We also just made the restore multiuser. If you like I still have the specs - which more clearly document the thinking behind the archive system. cheers j. _____________________________________________________________________________ Jack Parker | Hewlett Packard, BSMC Boise, Idaho, USA| "Discover America, jparker@hpbs2561.boi.hp.com | Get lost on a rally." (208) 396-5388 (W) (208) 384-1623 (H) | _____________________________________________________________________________ Any opinions expressed herein are my own and not those of my employers. _____________________________________________________________________________