On June 28, 2003 05:33 am, Aodhan Cullen wrote:
> 6/27/03 6:37:49 PM, Jeremy Zawodny <[EMAIL PROTECTED]> wrote:
> >> My read/update ratio would be something along the lines of 1:3, 3
> >> updates for every read. So it is highly unusual, and more or less
> >> rules replication out of the picture.
> >
> >I'm unclear why you can't use replication for this. There must be an
> >assumption about what you're doing that we do not share.
> >
> >If you read from the slave and write to the master, why does this not
> >work?
>
> A slave would simply not be able to keep up, replication works really well
> if you have a lot of reads, and a small number of updates. This is reverse
> ways, and needs a different approach.
Actually, I've found MySQL replication to be extremely fast. Our database is
also atypical having over 70% writes to 30% or less reads. However, we have
yet to have a slave (of 2 or 3) not be able to keep up. In fact, I can bring
a slave that has been pasued for a few days (as a snapshot) up to speed again
within an hour or two.
Here's some stats on our setup:
- Queries per second avg: 117.236
- Generating over a 1GB of binlog entries per day.
- Running with ~120 GB of live data now (used to be 200+ but we have
an archival scheme now)
Not sure how this compares to your situation.
--
Guy Davis http://www.guydavis.ca Calgary, Alberta, Canada
--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
