Scaling MySQL

[GridNote]

How do you "scale" mysql in the grid ? From my reading it seems that you must have similar resources available if your mysql instance dies. So if I'm deploying a mysql instance that needs 70% of a dual core processor, then I will have to have another physical server with similar available resources free to be able support a HA environment. So the MySQL instance is not distributed across my grid, but lives on a physical server, how does this differ from have mysql on 2 physical servers?

[PeterNic]

Let's assume you have an 8 server grid. You can use 7 servers worth of resources, while keeping one for reserve (N+1 redundancy). No matter which of the 7 servers dies, the 8th server will now run the appliances of the failed server.

Compare this to reserving two servers for each appliance, where you will need 14 servers to ensure HA (2N redundancy).

Regards,
-- Peter

[PeterNic]

And what was the question about scaling (the topic in your question was scaling but the content of the question seemed to be about HA)?

[GridNote]

The 8 server grid would require all servers to be equal. In an example of a 4 server grid. 2 x dual xeon , 2 x p4 2.4 . For a high resource db site , I would be forced to deploy my MySQL virtual instance on once of the dual xeon and leave the other unused to be able to support an HA environment.

As far as the HA I was thinking of cluster/grid computing were you have the CPU across all servers as resources for any job.

[PeterNic]

GridNote,

Distributing a single VM or a single process over several commodity boxes will actually reduce performance significantly due to the network latency; in addition, most regular software we all use simply doesn't scale this way. We try to be very clear in our literature that a single appliance or OS process is not spread across multiple servers. If you heard it from 3Tera, our partners or read it in our documents, please point me to the source so it gets fixed.

On the example grid -- I'd go with 3 dual xeon servers -- gives you the same number of CPUs and more flexibility. As you increase the number of servers in the grid, keeping them uniform makes more sense, and the reserve for the N+1 failover becomes a smaller percentage of the grid.

Best regards,
-- Peter

[jayglate]

How do you "scale" mysql in the grid ? From my reading it seems that you must have similar resources available if your mysql instance dies. So if I'm deploying a mysql instance that needs 70% of a dual core processor, then I will have to have another physical server with similar available resources free to be able support a HA environment. So the MySQL instance is not distributed across my grid, but lives on a physical server, how does this differ from have mysql on 2 physical servers?

Hi,


This something we over here at FortressITX are working on and should have a prototype ready by 2nd week of sept.

[PeterNic]

jayglate

I love to hear it. Let us know how it comes out.

Regards,
-- Peter