opinion: File Systems internal drive vs disk array

Since I am the lone DBA in my company, I would like to bounce some decisions i need to make. I am setting up a 2 node RAC on HP-UX and my question is about placement on the NON raw files and diretcory structures. Specifically, whether to place certain directory file systems on the HP boxes internal drives or on the external disk enclosure. The internal drives will be R0+ 1. The external disk enclosure will have a mixed bag of raw devices and file systems, all RAID0+1.

So, in your opinion, where would you put the following file systems?
Server internal drive or External disk array?

/oracle (for node 1) oracle binaries (instances will NOT share)
/oracle (for node 2) oracle binaries (instances will NOT share)
/archive_thread1 – archive logs for thread1
/archive_thread2 – archive logs for thread2
/backup_thread1 - misc backups NOT rman
/backup_thread2 - misc backups NOT rman
/backup - RMAN backups to disk prior to going to tape
/export
/data - This is basically the output of UTL_FILE dir

Much thanks for your opinion!
Frank

Everything on array. Always. Makes it much simpler to swap out or add a box. I configure all servers with mirrored boot drives for the OS & Swap and have everything else on the array (except for /etc/oratab).

If the binaries are going on the array then I might as well share them. True?

If the archive logs are on the array, perhaps it should be one file system instead of two? I will then name the archive logs with the thread number if in the same file system. You agree?

boot, os and swap will indeed be mirrored.

TYVM!

Quote:

---

Originally posted by gopi
If the binaries are going on the array then I might as well share them. True?

---

Sure.

Quote:

---

If the archive logs are on the array, perhaps it should be one file system instead of two? I will then name the archive logs with the thread number if in the same file system. You agree?

---

depends on the archive redo log activity. I wouldn't want my archiver to wait if the disk was really busy. Also, I would keep a closer eye on the free space in the log_archive_dest as I would end up hanging two instances if it filled up. If I were putting them on the same filesystem, I would probably put them in seperate directories.

RAC BAHH Hot standby all the way.
take a look see at this article. (someone posted this before and it's a very good read)

Quote:

---

Originally posted by OracleDoc
RAC BAHH Hot standby all the way.
take a look see at this article. (someone posted this before and it's a very good read)

---

This is indeed a very good article!

Not every shop needs RAC but I think we are one that does. Hot standby would failover very well but we need the load balancing and horizontal scalability too. The app we are cloning is very hybrid (OLTP and DSS). We can use the load balancing features and also hard wire certain monster jobs to the instance of our choice. (of course we would have to manually failover anything hard wired should that instance fail.)

I don't know gopi, I'm just an advocate for less moving parts the better. These days with arrays and caches, load balancing at the database server level shouldn't be an issue. Setup the standby and just let it run.
Now if you've got an application server to deal with that's a different story. Being able to have more than one app server that the clients can connect to would be nice for load balancing purposes.