nt cluster

Hi,
We are in the process of installing several microsoft clusters. The basic config is two nodes(two smp cpu's) with a shared storage array. The array has 8 - 18 gig drives and can be configured in a number of different RAID arrangements.

 On one node we will be installing an oracle database, which is a warehouse--batch loads at night only, no transactions during the day... On node number 2 we were probably going to put OAS and a web server.

As I understand the MSCS environment only one node will control a disk. You have to decide which disks are controlled by which node, the other node will only grab the disks if the primary node fails....

We are thinking of giving OAS two disks say #7 and #8, we will have these two disks set up in a mirror config(RAID 1)
That leaves us 6 disks to deal with the oracle database. Obviously spreading out disk I/O is the goal but with 6 disks and the requirement for raid 5 or 1 the actual number of usable disks is less. Since this is a 'pure' warehouse and a relatively simple one that is only 7 X 18 we have plenty of down time to deal with backups and full exports etc...

Do people have any suggestions or documents I can reference on how to lay out the database. Obviously I will need to have all essential files in fault tolerant setups to avoid losing database access from a disk failure So RAID 5 or RAID 1.

In terms of the actual amount of data the data tablespace and the index tablespaces will never exceed 18 gigs each(total would be 36 gigs or one disk each)
We will have the basic system, rbs, temp, user tablespaces There will be a data tablespace(s) and index tablespace(s)

I do not see any need to run in archive log mode. we could possibly lose the addition / deletion of users during the day but there is not a rapid turnover there anyway

Also I typicaly have 3 groups of redo logs(3 sets with 3 members on 3 disks), but I don't see a need for 3 groups in a warehouse?

Some tablespace/datafile layouts I have thought of are:
Mirroring -wasteful but an option

Disk 1,2 are mirrored and I would put the data tablespaces here
Disk 3,4 are mirrored and I would put the index and rbs there
Disk 5,6 are mirrored and would put redos   temp and system and user 

RAID
Could put 1-4 in a raid and throw most of the tablespaces(data,index,sys,) in there
5 and 6 could be for redos,ctls etc... Received on Fri Apr 14 2000 - 00:00:00 CDT