|
|
Posted by Hugo Kornelis on 11/28/57 11:59
On 27 Sep 2006 06:09:06 -0700, tlyczko wrote:
>
>Hugo Kornelis wrote:
>
>> Seperate locations for data and log files are always good.
>> The data file requires lots of random access, i.e. lots of head
>> movement. Luckily, SQL Server buffers data read from the data file, so
>> recently used data doesn't have to be read from disk again. Writes are
>> buffered as well.
>> The log file is mainly written to. These writes have to be completed
>> before SQL Server will signal the client that an operation is finished.
>> So write performance on your log file is critical to performance. Since
>> all log writes are sequential, you can gain a lot of performance if you
>> have the log file on a dedicated disk - the heads hardly have to move
>> and you write performance is it the highest possible rate - increasing
>> the time to completion for yur transactions.
>
>Thank you for explaining...particularly to people like me new to this
>software.
>However, this test DB is only a few meg for now...it'won't grow very
>fast.
>We do however have a production SQL server though with Great Plains and
>perhaps later another SQL-based app.
>Is there any sort of general guideline as to what database size(s) that
>this sort of thing becomes particularly important?? Such as DBs over 4
>GB or over 6 GB for example??
>Thank you, Tom
Hi Tom,
Putting log on a seperate volume increases performance for databases
wiith high data modification rates. This is regardless of size. You can
have databases several TB in size with minimal update activity, or
databases of just a few MB with very high activity - the latter will
proit much more from data-log seperation than the former.
--
Hugo Kornelis, SQL Server MVP
Navigation:
[Reply to this message]
|