locked
Storage solution suggestion for a clustered server RRS feed

  • Question

  • We have 3 clustered boxes sql server ent 2008 64 bit sp2.

    System Manufacturer: 'HP', System Model: 'ProLiant DL785 G6'.

    8x Six-Core AMD Opteron(tm) Processor 8439 SE

    256GB ram

    What would be a good storage solution that can have raid 10 configuration for tempdb and log and raid5 for data, it needs to be able to have 3 volumes 1tb for tempdb that will point to those 3 servers and 3 volumes for 2-3TB for log file and 3 large volumes 10gb or so for backups

    How much would a box like that should be in estimate .


    Gokhan Varol
    Thursday, February 2, 2012 9:34 PM

Answers

  • Ok, we currently are using SAN for all volumes, Today I was copying a 1TB file withing 2 volumes of the same server I got between 30-60MB/sec speed during copy, not sure if it good or bad.

    I was hoping there would be a way to improve sql server's performance by putting tempdb into a faster something, I did not know how this can be implemented in the clustered environment, if we choose to use something like fusion io which goes into motherboard and tempdb on it , then I guess every time sql server switches between node tempdb will be rebuild (we have 8 files totaling to 1TB of tempdb size). This also requires 3 set of fusionio one for each server, that seemed not good buck for the money. I guess in this case any external box other than san becomes out of question because of the requirements of the cluster.

    Do we have any options to improve IO the performance of the sql servers if we are already using a san ?


    Gokhan Varol

    Work with the SAN vendor to identify the root of the bottleneck and then purchase the required upgrade to alleviate that bottleneck.  

    There are so many different layers in a SAN architecture that it takes a comprehensive review to determine where the bottleneck is.  If you have oversubscribed your FA ports there is no point in adding disks or changing to SSD within the current SAN until you fix that problem.  If your pathing is not setup right and you are only getting 2 of the 4 available paths of throughput that is what you'd need to fix.  I've worked on SAN bottleneck and performance issues that had nothing to do with the disks in the SAN, it was an issue somewhere else in the stack and the fastest way to figure that out is to contact the SAN vendor.

    Using Fusion-IO for tempdb is not supported in a failover cluster.  Can you do it, sure, but it requires hacks that aren't supported for SQL Server 2005/2008/2008R2, and until 2012 goes RTM it isn't supported by Microsoft at all.  If you go that route and you have a problem, support can be refused until you can reproduce the problem in a supported configuration, and if that means your tempdb setup becomes a massive bottleneck, you might not be able to do that.  


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!

    • Marked as answer by GV1973 Saturday, February 4, 2012 12:13 AM
    Friday, February 3, 2012 11:50 PM

All replies

  • We have 3 clustered boxes sql server ent 2008 64 bit sp2.

    System Manufacturer: 'HP', System Model: 'ProLiant DL785 G6'.

    8x Six-Core AMD Opteron(tm) Processor 8439 SE

    256GB ram

    What would be a good storage solution that can have raid 10 configuration for tempdb and log and raid5 for data, it needs to be able to have 3 volumes 1tb for tempdb that will point to those 3 servers and 3 volumes for 2-3TB for log file and 3 large volumes 10gb or so for backups

    How much would a box like that should be in estimate .


    Gokhan Varol

    It is impossible to tell you what would work best here because you haven't provided any storage performance requirements, just size.  I could easily tell you to go buy a really cheap solution that has 1TB SATA disks in it that meets your size and RAID requirements, and it will be slower than the SSD in my laptop because the wrong information was used to create the environment.  You need to monitor your current IOPS and MB/sec requirements for each server and size based on those numbers some projection for future growth.  If you don't know how to do this, consider contacting a couple of the vendors out there like DELL, HP, EMC, IBM, etc.  They all have sizing tools for SQL Server that you run for 24 hours to get your I/O pattern.  Shop around and have a few of them provide you the report and quotes and see what you come up with.

    I can tell you that pricing multi-TB SAN platforms that are going to scale well and perform is not going to be cheap.


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!

    • Proposed as answer by Peja Tao Friday, February 3, 2012 7:30 AM
    Friday, February 3, 2012 2:16 AM
  • I am only trying to get a direction to start. We are already on the san with 3 clustered server nodes with 2 sql servers installed on them. They have their limitation, for some reason they cannot do raid 10, the performance is not the greatest (it's at a remote location to us, never seen it in person). We were thinking for a faster tempdb solution and a cheaper and faster daily backup solution that I guess will have a fiber connection to these 3 servers and will be cluster compatible, meaning pointing to a active server and when the passive became deactive point to the new active server. Therefore the volumes and their attachment to phsical servers follow the sql servers. At the same time assume A is active and B is deactive and C is another active node (we have 2 active nodes in 3 physicals), A backs up to volume BackA and C backs up to volume BackC, we want C to be able to see the volume BackA directly from fiber read only as A can see volume BackC readonly directly from fiber, this way we can restore a database backed up from C with not using much network traffic from server A.

    We thought about fusionio but it seem to be expensiver, not a good idea to have like 10TB volume for backup using fusionio.

    Our it and san team know only the hardware they have been using here, they are not much guidance, and I have not build a raid myself, just wanted to find out where to start looking and what to expect, and for tempdb and immediate drive backup needs what would be the way to go, does this external raid box need to have many different raids setups (3 x tempdb (1-2 tb each) volumes for 3 servers and 3 backup volumes for 3 servers (10TB each) to satisfy something like that, is raid 1 the way to go for the whole setup, drives should be sata or scsi, what should be the connection between the server and the raid box, and can these boxes be totally remote configurable, who has a good support etc.

    Thank you


    Gokhan Varol
    Friday, February 3, 2012 2:54 AM
  • What you are asking for is impossible.  If a LUN is active on a specific node, no other node will have access to that LUN except as a UNC share, so you will incur network overhead for the access.  I would recommend that you look at 10Gbit solutions that can be leveraged for providing a UNC path to the backup path so that you have high throughput from both servers to the location.  Then configured a dedicated network segment strictly for backup and restore activity, and then design the underlying RAID so it can support the 10Gbit connectivity throughput performance wise.  If you are using tempdb so much that it requires 2TB of storage, I would work on reducing the tempdb usage requirements because quite frankly that is incredibly large and will significantly impact performance.  I've worked on customer environments where I've reduced tempdb usage from the 1TB+ range down to a few hundred MB's and the result is incredible performance increases in the environment overall.

    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!


    • Edited by Jonathan KehayiasMVP Friday, February 3, 2012 4:15 AM
    • Proposed as answer by Naomi N Friday, February 3, 2012 4:30 AM
    • Unproposed as answer by GV1973 Friday, February 3, 2012 2:57 PM
    Friday, February 3, 2012 4:05 AM
  • About tempdb space, the workload we have over here is pure adhoc queries users write for each request they get from government, banks etc, these are nationwide queries touching 2-5TB data in 50+ tables, the queries itself use over 60GB memory sometimes, and there are many cases they are creating a local (#temp) table to make the query run faster or run ever and these #temp tables by themselves in some instances are over 500GB each forgetting about sorting and versioning etc, if there are multiple of these requests running the demand increases and 2TB tempdb space is something I threw it to be on the safe side. These are humongous ad-hoc queries against relational tables with over 400-500 columns and filtering and joining is so complex that no index does the job (like find all all the homeowners that has income, kids, which have houses under second foreclosure (thats tracking down a record history to find the link etc)).

    Our san i able to have a read only share to some other servers volume available to the other servers, they changed some company policy and we lost the ability, therefore I know it's possible to see a volume through fiber which some other server writes to through fiber.

    What we also would like to do is to have the log shipping, or daily backing up and restoring differential backups etc between the server not go through network traffic and only effect the io system, this io system will still be backed up by our san team but between machine transfers we are looking for a faster and cheaper on resources solution.

    At the end of the day we want faster tempdb space, faster disk backup restore space which is cluster aware and can be shared between servers (each volume can be pointed to one server for read write and to other servers for read at the same time), this storage should support 3x10TB + 3 x 2TB volumes to start with.

    Thank you


    Gokhan Varol
    Friday, February 3, 2012 3:10 PM
  • Jonathan is right here. Anything that's not formally recognized by the cluster won't be seen by a clustered instance without using a UNC path. I have heard some vendors provide unsupported workarounds to use their solutions, but it would put your cluster at a supportability risk from MS. That's not something I would recommend. So what you want is not possible with your architecture. A WSFC works the way it works. 

    SQL Server 2012 changes some of this with SMB support (so you can put DB files or backups on a network share seen by all nodes) as well as support for a local tempdb on the node itself, but that's it. You still can't do local backups from a clustered instance to a disk that is only attached to a node and not part of the cluster.

    The only way to get faster is to:

    1. Reduce your I/O load by tuning queries and indexes (which doesn't sound easy considering your workload is ad hoc)

    2. Increase the I/O capability of your storage solution which as Jonathan said won't be cheap

     


    Allan Hirt Blog: http://www.sqlha.com/blog Author: Pro SQL Server 2008 Failover Clustering (out now!)
    Friday, February 3, 2012 5:01 PM
  • About tempdb space, the workload we have over here is pure adhoc queries users write for each request they get from government, banks etc, these are nationwide queries touching 2-5TB data in 50+ tables, the queries itself use over 60GB memory sometimes, and there are many cases they are creating a local (#temp) table to make the query run faster or run ever and these #temp tables by themselves in some instances are over 500GB each forgetting about sorting and versioning etc, if there are multiple of these requests running the demand increases and 2TB tempdb space is something I threw it to be on the safe side. These are humongous ad-hoc queries against relational tables with over 400-500 columns and filtering and joining is so complex that no index does the job (like find all all the homeowners that has income, kids, which have houses under second foreclosure (thats tracking down a record history to find the link etc)).

    This sounds like a design problem is the underlying issue and you are going to hit a point, I've seen it happen multiple times, where you can't put hardware in to compensate for the design issue.  You should look at normalizing the structure to better support the requirements of the environment.  It is possible to work with multi-billion row data sets, and TBs of data and have good response times and a small tempdb impact, even for adhoc requests if the database is designed properly and good coding practices are applied to the Adhoc code.

    Our san i able to have a read only share to some other servers volume available to the other servers, they changed some company policy and we lost the ability, therefore I know it's possible to see a volume through fiber which some other server writes to through fiber.

    OK I shouldn't have said impossible, I should have said unsupported in a Failover Cluster resource group for SQL like that.  You can have multiple servers accessing a shared LUN, that is how Scalable Shared Databases (http://technet.microsoft.com/en-us/library/ms345392.aspx) work in SQL.  The same presentation of LUN could be done, but that doesn't make it a supported implementation.

    What we also would like to do is to have the log shipping, or daily backing up and restoring differential backups etc between the server not go through network traffic and only effect the io system, this io system will still be backed up by our san team but between machine transfers we are looking for a faster and cheaper on resources solution.

    How delayed is your log shipping or data updates?  If you are only loading data at fixed intervals, a SAN snapshot might be a better solution for the secondary system since it is already in the same SAN.

    At the end of the day we want faster tempdb space, faster disk backup restore space which is cluster aware and can be shared between servers (each volume can be pointed to one server for read write and to other servers for read at the same time), this storage should support 3x10TB + 3 x 2TB volumes to start with.

    Thank you


    Gokhan Varol
    The same recommendation still applies, you need to start working with the vendors out there to see what options they have that are going to meet your specific needs.  Look for something that is going to provide you the ability to use SSDs and have a 8Gb/sec FC connection or a 10Gb/sec iSCSI connection with multiple paths so that you can leverage the throughput optimally.  All of the vendors have competitive solutions that would meet your needs.

    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!


    Friday, February 3, 2012 5:05 PM
  • The only volume I'd like to mount to multiple servers is the one that will be used to enable backup files to be read through the fiber instead of the network. Today when we are restoring a 3-5 TB database from network share it slows down both servers, we'd like to isolate the slow down to a single server. tempdb space and other volumes of sql server will not be shared between servers.

    This question wasn't about changing the application or making it use more tempdb, I was asking some guidance on what kind of hardware to purchase which will work with multiple clusters similar to what san does today but in a smaller scale. The answers I am getting is not towards that intention.


    Gokhan Varol
    Friday, February 3, 2012 6:14 PM
  • The only volume I'd like to mount to multiple servers is the one that will be used to enable backup files to be read through the fiber instead of the network. Today when we are restoring a 3-5 TB database from network share it slows down both servers, we'd like to isolate the slow down to a single server. tempdb space and other volumes of sql server will not be shared between servers.

    This question wasn't about changing the application or making it use more tempdb, I was asking some guidance on what kind of hardware to purchase which will work with multiple clusters similar to what san does today but in a smaller scale. The answers I am getting is not towards that intention.


    Gokhan Varol

    Like I've already said, you need to go to the individual vendors and have them propose which of their solutions is a best fit for what you want.  The problem with the readonly/read write volume not being supported in the failover cluster is a fundamental limitation that you want and isn't a supported solution.  In order to guarantee that your backups don't fail if the cluster resource group is failed over to another node, the LUN for it has to be in the resource group.  In a multiple failover scenario, this could lead to the same disk being presented multiple times, one as a read/write one as a readonly, and that won't work.  It's not that you couldn't do it with a separate standalone systems, the problem is you are in a Failover Cluster and there are issues you have to consider.

    I won't offer any further recommendations or point out possible limitations and misdirection further if that isn't what you want.  Beyond what has already said, I am not a SAN vendor and I am not going to point you to one solution over another, they all can do what you are asking and they all have different benefits and features that you need to look at.  Unless you baseline any given configuration in the vendors lab you won't know if it is going to fit your needs.  If you want a smaller solution that fits your needs, and money is not an obstacle to getting fantastic performance, look at an all SSD based solution that will maximize IOPS and MB/sec per disk installed.  If money is a factor that changes the options significantly and only your SAN vendors are going to be able to tell you what you can or can not do within your specific budget limits, and what is going to fit your performance needs.


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!

    Friday, February 3, 2012 6:44 PM
  • Gokhan,

    So it seems to me that:

    1) You have a very large environment with frequent updates and demanding queries.

    2) Your SAN is not meeting your demand, and you are not in the mood to waste days dealing with vendors.

    So rather than taking criticism from someone telling you shouldn’t be having your problems before they even know what your problem is, how about you take a peek at SQL Magazine's 2011 Community awards for storage:

    http://www.sqlmag.com/content1/topic/2011-sql-server-magazine-editors-community-choice-awards-140830/catpath/awards/showprivate/1/page/8

    Hyper ISE, from XIO storage, seems particularly interesting because they combine expensive SSD and cheap HDD on the same unit. SSD would excel for those random requests from your adhoc queries. And HDD would be suited for our transaction logs and backups.

    Hope this helps.

    • Edited by giraom Friday, February 3, 2012 7:14 PM
    Friday, February 3, 2012 7:08 PM
  • That won't solve the issue of working right in the cluster - which is really what Jon and I are trying to tell you. You can't have your cake and eat it, too - we've told you how the cluster works. You clearly want to do something it's not designed to support, especially with SQL Server 2008/R2. Clustered configurations for SQL Server instances with SQL Server 2008/R2 REQUIRE shared storage. That's usually SAN or iSCSI. If you want to use UNC/network shares for things like backups, it could be possible but that's not what you're looking for either.

     

    Just because you don't like the message doesn't mean we didn't give you the right answer.


    Allan Hirt Blog: http://www.sqlha.com/blog Author: Pro SQL Server 2008 Failover Clustering (out now!)
    • Proposed as answer by Chirag Shah Friday, February 3, 2012 7:49 PM
    Friday, February 3, 2012 7:31 PM
  • Without much of complications, do you guys recommend a vender for a good subsystem for clustered system which will not serve the database and log files but it will only server tempdb and backup restores, log shippings etc. And should we be involved in what raid type etc  that box should have. Is 3 x 2TB tempdb volumes  and 3 x 10 TB backup volumes do require 3 different raid setups or we can bland all 3 backup volumes as a raid 5 into a single raid and have 3 x TB raid 10 for tempdb as 3 seperate raids, I do not enough information to start shopping and I do not like to deal with sales people unless I know the product well. 

    Thanks


    Gokhan Varol
    Friday, February 3, 2012 7:55 PM
  • Without much of complications, do you guys recommend a vender for a good subsystem for clustered system which will not serve the database and log files but it will only server tempdb and backup restores, log shippings etc. And should we be involved in what raid type etc  that box should have. Is 3 x 2TB tempdb volumes  and 3 x 10 TB backup volumes do require 3 different raid setups or we can bland all 3 backup volumes as a raid 5 into a single raid and have 3 x TB raid 10 for tempdb as 3 seperate raids, I do not enough information to start shopping and I do not like to deal with sales people unless I know the product well. 

    Thanks


    Gokhan Varol

    It entirely depends on the vendor and SAN that you pick for the implementation.  Some SANs only offer a large pooled storage implementation like the Xiotech mentioned, an EMC VMAX, Compellant, NetApp FAS series, HP EVAs, most Equilogics, and various other SANs out there.  For those you don't really get much control in  how the RAID is setup so it is useless to try and go down that route if you don't know it will be possible to do.  Different SAN's offer different levels of capabilities, maximum LUN sizing, etc.  If you pick a traditional type of SAN where you can control it, it still may be better to build a single large pool and carve the LUNs from that instead of creating individual RAID groups and isolating things down.  Without knowing the specifics of what storage array and its configuration we are looking at it is impossible to tell.  

    The best thing I can off is that you download a free copy of my book Troubleshooting SQL Server: A Guide for Accidental DBAs and read Chapter 2 which covers storage for SQL Server. Beyond that, I would say that you need to hire a consultant to help you with planning and sizing your environment, and it should be someone that is local that can come onsite with you to look at everything you have currently, and what you need moving forward.  It should ideally be someone that knows SAN architectures well as well as SQL Server.  The easiest way for you to find one of those is to contact the SAN vendor and have one of their technical specialists come on site for a scoping session.


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!

    Friday, February 3, 2012 8:12 PM
  • we do not want a second san, we want to have a smaller and faster box only for those 3 servers for tempdb and backups.
    Gokhan Varol
    Friday, February 3, 2012 8:17 PM
  • we do not want a second san, we want to have a smaller and faster box only for those 3 servers for tempdb and backups.
    Gokhan Varol

    Like we've said, you need to understand how clustering works. If the solution you ultimately implement can be presented PROPERLY to the clustered instance, great. Third party, native, whatever - that doesn't matter. If it can't, why bother? There's no magic bullet here.
    Allan Hirt Blog: http://www.sqlha.com/blog Author: Pro SQL Server 2008 Failover Clustering (out now!)
    Friday, February 3, 2012 9:04 PM
  • we do not want a second san, we want to have a smaller and faster box only for those 3 servers for tempdb and backups.
    Gokhan Varol

    Your requirements spell SAN.  Lets review them and why:

    3 server cluster - No DAS (directly attached storage) out there supports 3 servers, there are some hybrids (that are actually classified as SANs) that allow you to have 2 cluster nodes, but they don't support 3.

    Large capacity requirements with "faster" performance  - Your numbers for capacity with performance can only be met by a SAN, or local DAS with all SSDs in it (which takes out your clustering scenario).  Your numbers equate to 36TB of storage, that's a lot of spindles to have fast performance, there is no smaller about that size of data with good performance even with SSDs.

    I don't want to be mean in saying this, but you obviously don't understand enough about how/why for storage to be planning something like this; it is way over your head at this level of requirements and you REALLY NEED TO BRING SOMEONE ELSE IN if you want to have success.  I am a consultant, but I am not recommending you hire me.  If you were a client of mine I would tell you exactly the same thing I am saying here: you need to bring in a storage consultant that can size this properly.  I have never just gone out and bought a storage configuration for SQL, it takes a lot of planning to figure out what is going to be required to meet the needs of the environment, and the vendors have the right people to do that with you, and they can explain the hows and why's of their implementation.


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!


    Friday, February 3, 2012 9:34 PM
  • Ok, we currently are using SAN for all volumes, Today I was copying a 1TB file withing 2 volumes of the same server I got between 30-60MB/sec speed during copy, not sure if it good or bad.

    I was hoping there would be a way to improve sql server's performance by putting tempdb into a faster something, I did not know how this can be implemented in the clustered environment, if we choose to use something like fusion io which goes into motherboard and tempdb on it , then I guess every time sql server switches between node tempdb will be rebuild (we have 8 files totaling to 1TB of tempdb size). This also requires 3 set of fusionio one for each server, that seemed not good buck for the money. I guess in this case any external box other than san becomes out of question because of the requirements of the cluster.

    Do we have any options to improve IO the performance of the sql servers if we are already using a san ?


    Gokhan Varol
    Friday, February 3, 2012 11:37 PM
  • Ok, we currently are using SAN for all volumes, Today I was copying a 1TB file withing 2 volumes of the same server I got between 30-60MB/sec speed during copy, not sure if it good or bad.

    I was hoping there would be a way to improve sql server's performance by putting tempdb into a faster something, I did not know how this can be implemented in the clustered environment, if we choose to use something like fusion io which goes into motherboard and tempdb on it , then I guess every time sql server switches between node tempdb will be rebuild (we have 8 files totaling to 1TB of tempdb size). This also requires 3 set of fusionio one for each server, that seemed not good buck for the money. I guess in this case any external box other than san becomes out of question because of the requirements of the cluster.

    Do we have any options to improve IO the performance of the sql servers if we are already using a san ?


    Gokhan Varol

    Work with the SAN vendor to identify the root of the bottleneck and then purchase the required upgrade to alleviate that bottleneck.  

    There are so many different layers in a SAN architecture that it takes a comprehensive review to determine where the bottleneck is.  If you have oversubscribed your FA ports there is no point in adding disks or changing to SSD within the current SAN until you fix that problem.  If your pathing is not setup right and you are only getting 2 of the 4 available paths of throughput that is what you'd need to fix.  I've worked on SAN bottleneck and performance issues that had nothing to do with the disks in the SAN, it was an issue somewhere else in the stack and the fastest way to figure that out is to contact the SAN vendor.

    Using Fusion-IO for tempdb is not supported in a failover cluster.  Can you do it, sure, but it requires hacks that aren't supported for SQL Server 2005/2008/2008R2, and until 2012 goes RTM it isn't supported by Microsoft at all.  If you go that route and you have a problem, support can be refused until you can reproduce the problem in a supported configuration, and if that means your tempdb setup becomes a massive bottleneck, you might not be able to do that.  


    Jonathan Kehayias | Principal Consultant, SQLSkills.com
    SQL Server MVP | Microsoft Certified Master: SQL Server 2008
    Author of Troubleshooting SQL Server: A Guide for Accidental DBAs
    Feel free to contact me through My Blog or Twitter. Become a SQLskills Insider!
    Please click the Mark as Answer button if a post solves your problem!

    • Marked as answer by GV1973 Saturday, February 4, 2012 12:13 AM
    Friday, February 3, 2012 11:50 PM