Clusters Opinion

  • Hello All,

    Thanks for your time and hopefully this DOES NOT turn into a cluster vs log shipping post. So here is my situation

    I have many shared servers(5) that have approx 300 databases on them. I would like to consolidate this so that i can keep my high availability yet make life simplier for me in the long run. Now bear with me in that i cant put more then 80 databases on any one server due to older infrastructure reason(too long a story to type but it is what it is). So i am thinking that if i can move all that over to the newer infrastructure and then cluster 2 or 3 machines i gain because i release 2 or 3 servers AND keep a good high availability as well. I do have some experience with clusters but i like to get some input from others because there's more then one way to get it right. OH BTW..the cluster would be on 2003 OS and either 2008 or 2005 SQL all Ent Editions. BTW... Log Shipping is really not an option. This set up must stay up because at any given time i amy have 15-20 different companies using the cluster or databases (unless maintenance time of course).

    Any thoughts would be greatly appreciated

    Thanks

    -D-

    DHeath

  • How are you achieving high availability in the old infrastructure?

    I'm not sure what you mean by 'shared server'

  • We currently run two active/active/active SQL clusters (MS 2003/SQL 200 SP4) We are getting ready to stand up two new SQL 2005 Clusters. The biggest problem we had initially was the SAN shared resource that we utilized for SQL Server. We were running on ISCSI SATA, we found that the I/O throughput was horrible on the SATA drives and eventually went with a SAS ISCSI SAN. If you are utilizing fiber this will not be an issue for you but make sure the shared SAN does not bottle neck your SQL servers.

    On another note if I had it to do over I would go with VM to achieve my HA solution. This requires less hardware and my experience with MSCS is it is great when it is working but it is a real pain when it breaks.

  • Thanks for the looks... 🙂 How are we achieving the availability now. Well needless to say... we just hope the servers stay up and not have any problems. And so far the servers are pretty good with the maintenance windows and all few problems. So just trying to "update" that if i can or maybe its ok to just leave them alone and go with the (if it aint broke dont fix it idea)

    "Shared Server" hhmmmmmm will give it a shot..These servers belong to my team... we are the "owners" of them(hardware) but i have clients that dont want to have a database server so i allow them to use the "shared server". These servers could have any number of different companies using the database aspects of the server and i have full responsibility of their backups and restores and the fact that they will access to their servers at all times. In a nutshell that explains what a "shared server is" I have them for SQL7, SQL2000,SQL2005,SQL2008(in the process). This allows for some organizations much flexibility. Their are tons of reasons they would need a shared server but thats besides the point.

    Hopefully that gives you a clearer picture of the situation. If not my apologies

    -D-

    Thanks Again

    DHeath

  • High availability usually implies some form of redundancy, rather than a standalone server that rarely encounters problems of a hardware or software nature.

    My understanding is that you have 5 servers with 300 databases on them for various different customers. Consolidation of these would involve monitoring their usage to determine what memory, processor and disk space requirements they have. Armed with this knowledge you could spec 2 - 3 servers to house them all.

    If you want high availability you should employ clustering (server level failover with shared data array), or database mirroring ( database level failover with independent, redundant data). To be honest with 300 databases, I would be looking at clustering. it would be considerably less grief than individually mirroring 300 databases.

    If budget is not an issue you should purchase the 2 - 3 servers mentioned above in duplicate along with a shared data array (SAN).

    This would give you a cluster (or clusters) operating in an Active / Passive scenario. Many would argue that if you 'put a bit more memory in the servers' you could split the database load and cluster more databases. This would be an Active / Active scenario. This requires careful planning because in the event of a problem with one of the Nodes all the databases will failover to run on one node. If there is not enough memory or processor available all databases may become virtually unusable, giving you a very expensive self-defeating object.

    hth 😉

    mark

  • Thanks to all for the time in your comments and suggestions. JBaxter your insight is much appreciated. Mark your ideas are right along the lines i was thinking previous to the post but wanted to get a lil bit of justification on my thought process. And to all other DBA's ...SSC ROCKS!!! thanks a lot everyone. 😀

    -D-

    DHeath

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply