You should read this page on clusters:
http://dev.mysql.com/doc/mysql/en/MySQL_Cluster_Limitations_in_4.1.html
Replication and clustering is different technology.
Replication works great, but there are no algorithms to prevent the same row being updated on different servers at the same time (in other words, multimaster replication does not have collision detection algorithms like Oracle does). You can however chain servers together so that a server that is a slave is also the master to another server. I'd suggest you should buy & read High Performance MySQL by Derek Balling and Jeremy Zawodny. He discusses all these issues.
Last but not least, there is a big disconnect between "24x7" and "way too expensive for our budget". Say you get three computers replicating so that in the event of the failure of a single server, you still have two up and running. What if your power goes out for a day? Do you have a backup generator? Do you have multiple redundant network feeds? Do you have multiple hot-standby sites around the world to take over in the case of an earthquake/tidal wave/hurricane (even if you did, how long would it take for the DNS updates to percolate around the world)?
The cost of a setup goes up by orders of magnitude as you go past the 99% availability. And remember, hardware requires someone on call or sitting at a desk 24 hours a day, 7 days a week to monitor the hardware, and change things around. When I worked for GTE, we spent over a million dollars on a single server that had redundant nodes (each with their own disks) all connected by a very fast fiber optic cable, and we needed several of them.
Food for thought.
David
<http://www.oreillynet.com/cs/catalog/view/au/1758?x-t=book.view> Mark Papadakis wrote:
Hello all,
After playing with the idea of 'abandoning ship' in favor of IBM DB2 or Oracle, we deiced to stick with mySQL, due its simplicity and investment in time and experience we have put into it.
Our company needs a HA solution for ensuring 24x7 operation for the mySQL server instances. As it is, there are are two solutions available for the problem: Emic Networks's EAC for mySQL and MySQL's Cluster.
The Emic solution seems to work but is way too expensive for our budget (around 4k$ for each 2CPUs node). So we need to either go with NDB or try to get replication to work properly.
Here is a list of questions: o How 'stable' is MySQL cluster (NDB) ? Is it ready for enterprise use? Or even tested?
o Does the memory size limit the data we can manage? If it is a memory based solution it should mean that it can handle of a very limited number of databases/tables/rows, based on the available memory of the nodes.
o Is there some sort of tight integration planned for mySQL cluster and mySQL server ? Perhaps in 5.0 ?
o When is adding/removing nodes on the fly scheduled for implementation? Without such a feature the system would have to be shutdown - therefore not a complete HA solution.
o Has anyone gotten replication to work right? Perhaps Multi-Master replication?
Thank you very much in advance, MarkP
-- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
