Bjorn - you state to use auto_increment in 2-way replication. Others have said that this could cause a crash if the auto_increment field is duplicated. Have experienced any problems with it?
----- Original Message ----- From: "Björn Pålsson" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Sunday, December 08, 2002 3:14 PM Subject: Re: E-Commerce, db-clustering and MySQL? > First, I wish to say that you are making the right descision leaving M$ and > taking a step into MySQL. > > Currently, I'm working on a project involving +500.000 rows and estimated > 30.000 visitors / day on a 2 x dual 2.8 GHz Xeon dedicated for SQL. > Additional servers are for HTTP and other tasks. There will be a lot of > load on the SQL when the users searches our database on either category, > area or on a fulltext keyword. > > The key to managing thousands of visitors is to have a system setup to > minimize the number of querys and managing the traffic between the diffrent > machines. We will use a second gigabit-switch and a secondary net on the > 10.0.0.0 to unload our primary public switch. We also have pre-cached pages > with the most popular products/listings. In our case, the visitor may input > data and therefore he will trigger a re-cache. A separate machine is used > for this purpose as this may happen very often. > > To make a user allowed to input data you must setup the servers in a ring > with each server acting as both master and slave. If you have a clean > system with very few inserts/updates you may choose a simpler solution with > a single master an several slaves f.x. if the only insert is in the > checkout in a webshop. If you have click counters, this simple solution > will fall. > > If you have the ring-solution, you must make shure your script never makes > any contradicting inserts. Use auto_increment if possible when inserting > and/or use your session-id as an identifier. Always use relative updates > when you must update a value. > > Do not write > > SELECT count from article WHERE id = '$id' > > $count++ > > UPDATE count set count = '$count' WHERE id = '$id' > > but use > > UPDATE count = count + 1 > > All these things to think about will force you to dig deep into your script > and if you have a script-in-a-box solution from a third party coder, you > are definately in a difficult situation. If you know your script well, you > will have an easier task. > > Please tell me about your future findings! > > Greets > > >Hello: > > > >We are running a large e-commerce site currently with ASP/MSSQL. The ever > >increasing number of hits, complexity of the site and request for reporting > >is slowly but surely bringing us to the point were the current database > >application is getting tired. > > > >Now we're thinking Linux, MySQL and clustering. > > > >I have been looking around a little and don't like the fact that there are > >no stored procedures and no triggers. However - if we cluster several quad > >processor machines we should be fine without. > > > >Right now it's all theory, and there don't seem to be a lot of articles on > >the web about this. At least I had no luck finding any. > > > >Does anybody have real life experiences with this kind of a setup? Does > >anybody know where to find reports about this? > > > >I'd be greatful for any link or hint! > > > >Dietrich Speer > > > > > > > --------------------------------------------------------------------- > Before posting, please check: > http://www.mysql.com/manual.php (the manual) > http://lists.mysql.com/ (the list archive) > > To request this thread, e-mail <[EMAIL PROTECTED]> > To unsubscribe, e-mail <[EMAIL PROTECTED]> > Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php > > --------------------------------------------------------------------- Before posting, please check: http://www.mysql.com/manual.php (the manual) http://lists.mysql.com/ (the list archive) To request this thread, e-mail <[EMAIL PROTECTED]> To unsubscribe, e-mail <[EMAIL PROTECTED]> Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
