Tablespace/Datafile configurations

Hi all, Our database is currently 4GB and we are projecting it to grow to 40GB end of next year. What is the best way to allocate the tablespace/datafile config for this.

Current config is : we have 4 mount points and
1 - system
2 - Redo
3 - Data - with single datafile of maxsize 8GB
4 - Index - with single datafile of maxsize 8GB

Would it be better to add more datafiles to the existing tablepsace mount points or is it better to distribute the datafiles acorss the mount points in a round-robin fashion ?

Thanks.

spread all files across as many spindles as you can. Evenly distribute your I/O and if you can keep redo on their own separate very fast file system

How many physical disks are in the server?
Did you use RAID?

Tamil

Quote:

---

Originally posted by tamilselvan
How many physical disks are in the server?
Did you use RAID?

Tamil

---

What's RAID isn't that some kind of insect killer? That's not very nice messing with people like that Tamil.

:D Wow im surprised Oradoc u also have RAID in ur place, but i prefer
BAYGON. its effective:p

Hi Dears.
Just curious about this I/O striping accross spindles performance
strategy.

Say I have two servers A & B.(same O/S win2k).

Server A have six (6) hard disks of capacity 20G each = total 120G.
Server B have one (1) hard disks of capacity 120g = total 120g.

Assuming that I filled up both servers disk spaces will data, which
one will perform better? Considering that those 20g-size disk is
much slower than the 120g-size disk.


Thanks

Quote:

---

Originally posted by tamilselvan
How many physical disks are in the server?
Did you use RAID?

Tamil

---

We are using NetApp servers and there are 5 disks.

Quote:

---

Originally posted by rshivagami
We are using NetApp servers and there are 5 disks.

---

IMHO, the more relavent question is how are your LVMs setup across those disks? Also, what's your RAID level?

Quote:

---

Originally posted by kris123
:D Wow im surprised Oradoc u also have RAID in ur place, but i prefer
BAYGON. its effective:p

Hi Dears.
Just curious about this I/O striping accross spindles performance
strategy.

Say I have two servers A & B.(same O/S win2k).

Server A have six (6) hard disks of capacity 20G each = total 120G.
Server B have one (1) hard disks of capacity 120g = total 120g.

Assuming that I filled up both servers disk spaces will data, which
one will perform better? Considering that those 20g-size disk is
much slower than the 120g-size disk.


Thanks

---

All things being equal the array of 20Gb disks would perform better .. more bandwidth, more spindles, more heads etc.

Of course if the 20Gb drives were very old and crappy then things might be different, but in general "more & smaller" beats "fewer & bigger".

Since you have 5 disks only, my suggestion is use RAID 0 - Striping alone. Simple striping across 5 disks at about 64K or 128K per disk (stripe unit size) is probably the best way of minimizing I/O contention.
Create fewer mount points. Other wise, later you will find that one mount point will be more heavily used than others.

Tamil

Quote:

---

Originally posted by rshivagami
Hi all, Our database is currently 4GB and we are projecting it to grow to 40GB end of next year. What is the best way to allocate the tablespace/datafile config for this.

Current config is : we have 4 mount points and
1 - system
2 - Redo
3 - Data - with single datafile of maxsize 8GB
4 - Index - with single datafile of maxsize 8GB

Would it be better to add more datafiles to the existing tablepsace mount points or is it better to distribute the datafiles acorss the mount points in a round-robin fashion ?

Thanks.

---

Your current config looks just fine. But in the end its the actual I/O that needs to be monitor. If you see improper I/O distribution then might need to move data files from 1 disk to the other or your subsequent addition of data files will be considered. And assuming that your mount points are for each disk. But this is only effective when no RAIDs implementation, do you have one?