Any caluculation ro formula to size redo log

Hi
I would like to know is there any formula or caluclation in sizing the redo log. apart from the general funda and talk.:)

you might need to know first of what are the factors
or configuration parameters that greatly influence
the performance of your redo logs.
redo log buffers
LOG_CHECKPOINT_INTERVAL or
LOG_CHECKPOINT_TIMEOUT
multiplexing redo logs
Archiving redo logs
+
maybe hardware configuration factors(RAIDs,mirroring)

For me, the size of a redo logs will depends in how a
database is being utilized. Most of the db config setting
LOG_CHECKPOINT_TIMEOUT to 0 will allow the archiving of redo logs
at LOG SWITCHES/CHECKPOINT(if could remember it right). If you have a long processes in your db, and you have a huge size set up in your redo log files, it might take a lot
of time for SMON to recover your db during instance recovery if CHECKPOINT is set high. Again, it all depends in your database activities. You probably needs redo logs to be archived more often in away that it won't hurt additional IO cost.
IMHO, there could be no definite calculation of redo log size, but there is a calculation for LOG_CHECKPOINT_INTERVAL.

Hmm,

You need a formula ...

Y = x*x*x + x*x+x+1

I hope this helps you.

Tamil

This is what I do...

Start with 100M redologs
See how often they switch in peak periods of activity in database, adjust the size so that the log switch occurs every 30 min.

adjust the size so that the log switch occurs every 30 min.
___________

I am trying to understand why we need to make a switch for every 30 min. For your DB it may be best way. How can we generalize the time intervel. Any good reason ??

If you're running a standby server (not DataGuard) or just simply shipping the archlogs to another location, the interval between log switches determines how much work you lose in case of total destruction of your server.

So (e.g.) I force a log switch every 10 minutes during working hours - that's how much OLTP the user management decided they could afford to lose and re-enter.

Quote:

---

Originally posted by raghud
I am trying to understand why we need to make a switch for every 30 min. For your DB it may be best way. How can we generalize the time intervel. Any good reason ??

---

No, that's just a rule of thumb that doesn't work for every (arguably any) environment. As long as your arch can keep up with your lgwr and you don't get "checkpoint not complete" messages, your logs are probably fine. My logs switch every couple minutes. I'd have to have a 20G log file if I wanted it to switch every 30 minutes and then it would pound the crap out of my log_archive_dest for a couple minutes.

Quote:

---

Originally posted by tamilselvan
Hmm,

You need a formula ...

Y = x*x*x + x*x+x+1

---

If you're sending them over a network, this may contain more useful formulae:
http://cm.bell-labs.com/cm/ms/what/s...hannon1948.pdf

Quote:

---

Originally posted by DaPi
If you're sending them over a network, this may contain more useful formulae:
http://cm.bell-labs.com/cm/ms/what/s...hannon1948.pdf

---

Given the circular nature of redo logs, I thought the formula was
PI*R*R ;)

Seniors Guys be patience !

Good explanation from "marist89". Some times there may be basic questions from the users. Any way knowledgable DBAs do not need help, they can solve their problems. But these kind of sites (user groups )are very helpful for the Jr.DBA's .