A question about servers

  • HI All

    ok here is the senario, i have 4 sql servers on a AD with 220 staff and 2500 student users, i have a contractor wanting to consolidate my sql boxes.

    i have 

    1 on the DMZ

    1 primary

     1 backup

    1 for monitoring mail

     1 used for testing.....

     

    we are a rural education organisation and our support agreement with our server vendor is " next day" for any hardware support, now thats the best we can get becuase of our remote location.

    therefore i have redundant servers (1) just incase we loose our main sql box or any sql box.

     

    our contractor wants to have ONE  sql box with multiple instances of sql server.

     

    what would be your suggestions or concerns  (pro's / con's) on this issue

     

    thanks

     in advance

     Weazzell

     


    Kindest Regards,

    Todd,non est vivere sed valere vita est

  • I would get a new contractor of they want to give you only one SQL server. If you loose the server you loose all the instances too.

  • Server consolidation is a fact of life, but edswartwout is right, moving to one is a bit drastic, and since the presence or absence of a specific contractor is probably not within your authority, you probably need better ideas based on solid reasoning.  Here are some thoughts:
     
    1. The server in the DMZ can be moved inside the network and connections allowed to pass through the firewall, but this opens a security hole.  Personal policy is no transactions initiated outside the firewall (including the DMZ) should be allowed inside the firewall.  All transactions that cross the firewall should be initiated from within the firewall.  This is harsh and sometimes requires new ways of thinking about things, but it is the most secure and it will require at least one SQL Server in the DMZ.
     
    2. Consolidation though instances is a good way to handle consolidation because it isolates the SQL Server engine for each purpose.  Having warm standby is based on availability requirements of your applications.  Is it acceptable for the server to be offline for a few days while next day support arrives, new parts are ordered and brought in, and finally installed?  Further, if you are forced to restore databases from backups, what data loss is acceptable in the case of failure?  With warm standby, you have live data backup being kept up to date and data loss is kept in the 5-10 minute range.  With restoring from backup, depending on your back up strategy, data loss in the event of a restore from back-up could range in the hours or even days.
     
    As always, it depends on the requirements of the application, but these are the things to consider when faced with a consultant who is going to save your organization a ton of money by consolidating.  Consolidation is good, but be careful not to put critical applications at risk.
     
  • Agree with above. Consolidate to 1 if you can, but keep a backup. Use log shipping, etc. to get your data moved to the second server and ready in case youhave an issue. Be sure these two servers are on separate power, network switch, etc. in case of issues and be sure you test and watch the processes to backup your data.

  • I would also consider application requirements. Some applications require specific connection information and not all of them will work with named instances. Also named instances support not all of the features available in the default instance, files are in different locations, registry entries are different. Is your application ready for it?

    Yelena

    Regards,Yelena Varsha

  • What is the rationale given for consolidating to one box? What is broken about the current setup?

    Is that box going to be in the DMZ?

    How will the equivalent functionality be maintained with named instances on a single box?

    Will you have to buy a new server to support this configuration? (Named instances all use the same memory and drive pools so those will have to be split 4 ways. Can the current drive sets/CPU's handle 4 times the load?)

    What is this contractor's disaster recovery plan?

     

    Sounds like a change for the sake of a commission check to me.

     

     

    Richard L. Dawson
    Microsoft Sql Server DBA/Data Architect

    I can like a person. People are arrogant, ignorant, idiotic, irritating and mostly just plain annoying.

  • I agree with Tinker. The slowest part of any database is the disk subsystem and I/O contention will kill performance. Is the contractor also recommending going with separate physical arrays for each of your current separate SQL Server groups?

     

    If I had to guess I would say that the mail monitoring box probably has the highest number of I/O transactions do you really want every other instance waiting for disk access because of e-mail?

     

    Dave

Viewing 7 posts - 1 through 6 (of 6 total)

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