The GIMPS Board of Director's has approved purchasing a new server. But first we are looking for anyone with expertise in server hardware, server configuration, or database planning / disaster recovery to examine our plan and suggest improvements.
What follows is a brief outline of our plan to purchase a new server for database only - converting our existing server to web traffic only. 1. Priorities a) Reasonable Cost, b) Data protection, c) Reliability, d) Performance. (Data protection also implies roomy capacity.) 2. Plan Summary a) We will buy a powerful new database server having VM hosting capability and RAID6 as a DB-only tier, convert/upgrade our existing ISP-loaned server to a x64 web-only tier, convert/upgrade the old v4 RAID5 server as a x86 DB backup, add an external RAID5 array device as a third database/file backup store. b) We presently own none of our servers. The old v4 server will be donated and become the first GIMPS asset, the purchased DB server the second, the external RAID5 the third asset. 3a. Proposed hardware http://ecomm.dell.com/dellstore/bask....@mersenne.org 3b. Proposed software Windows Server 2008 Datacenter Edition x64 OS: $120 (for production DB server) SQL Server 2008 x64 or x86: $240 x 3 CPUs = $720 (2x for production DB server, 1x for roll-forward backup DB server ) Windows Server 2008 Web Edition x64 or x86 OS: $19 x 2 servers = $38 (1x for DB server VM image, 1x for ISP-loaner) Windows Server 2003 Std Ed x86: $40 (for roll-forward backup DB server) Windows Remote Desktop Services User CAL: 5 x $6 = $30 Microsoft FrontPage 2003: 1 x $7 4. Disaster Plan a) If ISP-loaned web server is reclaimed by the ISP or breaks, we load a VM web server into the DB server to handles the web tier traffic until we can purchase a replacement, or the ISP loans a suitable alternate server. In practice we should have the VM web server image preloaded on the DB server, but not running. b) If our new primary DB server breaks, we get it fixed ASAP and we are down until it gets fixed. c) The new primary DB server's SQL service will be brought offline for an hour or so every 3 months for an offline file snapshot to the external RAID5 array. These serve also as roll-forward starting points for the backup DB. d) The new primary SQL DB server will regularly and automatically ship its logs to the external RAID5 array. The backup SQL DB server (donated v4 server) will reach over and roll-forward these logs to maintain a mostly-current second SQL DB. e) If our primary SQL DB becomes damaged or lost, we have the backup SQL DB available to copy back onto the new SQL DB server, having lost only the transactions not yet shipped and rolled-forward in the backup DB. f) If our backup SQL DB becomes damaged or lost, we have the snapshots and logs in the external RAID5 from which to reload the most recent DB snapshot and roll-forward to catch it up again. g) If our external RAID5 DB file copies become damaged or lost, we produce a fresh offline snapshot and start the logs shipping onto it again. h) Other data safety options - include shipping the SQL DB snapshots and logs using scripts or other tools to other remote storage and/or SQL server instances. _______________________________________________ Prime mailing list Prime@hogranch.com http://hogranch.com/mailman/listinfo/prime
