Need a good hardware cluster solution

Discussion in 'Windows Networking' started by Clayton Sutton, Dec 16, 2004.

  1. Hi everyone,

    The company that I work for is wanting to move to a Windows 2003 Server and
    Exchange 2003 clustered environment. I know that Windows 2003 (Standard)
    will do a "Network Load Balancing" and the Enterprise Edition will do both
    "Network Load Balancing" and "High Availability" clustering but not BOTH.
    If you want to do BOTH "Network Load Balancing" and "High Availability" you
    need a third party solution. That's what I'm looking for, anyone have any
    ideas? Also, any white papers on Windows and Exchange clustering would be
    great too. Thanks for any input.

    Clayton Sutton, Dec 16, 2004
    1. Advertisements

  2. check out specifically their Big-IP solutions.
    =?Utf-8?B?QnJ5YW4gRSBGYWlyY2hpbGQ=?=, Dec 16, 2004
    1. Advertisements

  3. Hi,

    You don't need a third-party solution to use NLB and server clusters. For
    example, you can use NLB for front-end Exchange servers (which can run the
    Standard or Enterprise Edition of Exchange) and you can cluster back-end
    Exchange servers (running the Enterprise Edition of Exchange) using the
    Windows Cluster Service found in Windows 2000 Advanced Server, Windows 2000
    Datacenter Server, Windows Server 2003 Enterprise Edition or Windows Server
    2003 Datacenter Edition.

    Information on planning, deploying and managing Exchange 2003 clusters can
    be found at:

    Hope this helps.

    Scott Schnoll
    This posting is provided "AS IS" with no warranties, and confers no
    rights. Please do not send email directly to this alias. This alias is for
    purposes only.
    Scott Schnoll [MSFT], Dec 16, 2004
  4. Yea, but the front end servers would not be "High Availability". And the
    back end servers would not have NLB-ing. Your solution does assume that a
    Microsoft/software solution is the "RIGHT" solution. I'm looking to see if
    there is a hardware solution that will do "BOTH" ("High Availability" and
    Load Balencing). Just doing my homework right now so I don't look dumb when
    going into my meeting with my boss next week.

    Clayton Sutton, Dec 20, 2004
  5. Papers on clustering Exchange 2003 (High Availability Guide link here plus
    a myriad of other papers)
    As for a clustered solution, there are some "cluster in a box" products.
    Compaq put one out a few years ago and it worked pretty well. HP seems to
    have continued this.

    Stratus makes servers designed for high-availability. Literally, you can
    yank a processor out of the box and it will keep on humming (not

    HA does not always mean a cluster investment...most of the time, but not
    always. As for load balancing...realistically, if you cluster, you want to
    look at an active-active-passive solution...and that is some $$.

    Many companies would like five 9's...but when they see the price behind it,
    it is pricey up-front. I have researched five 9's solutions for several of
    my past employers as well as the cost of downtime. Most of the smaller
    firms (1000 or less employees) could tolerate some unexpected downtime.
    Realistically, with the Compaq hardware I had, there were very few hardware
    problems and most of those were self-induced (letting the server room get
    over 95 degress (long long story). Honestly, there were more problems with
    people sending 100MB attachments (long story, but it had to be allowed) and
    not understanding why it did not get there in 15 seconds like a small 1k
    message did.

    Bob Christian, Dec 20, 2004
  6. Clayton Sutton

    Scott Lowe Guest


    I could be wrong (wouldn't be the first time), but I don't think there
    is a solution that will do both, especially when it comes to Exchange
    Server 2003. There are some third-party clustering solutions (Veritas
    Cluster Server comes to mind), but I don't know if those behave in the
    same way as Microsoft's own Cluster Server. What you need to keep in
    mind is that Exchange is a different beast than a typical "web
    application," in which you could use a hardware-based load balancer
    (F5's Big-IP, Foundry's ServerIron, etc.) to create a "load balanced
    fault-tolerant" environment. With Exchange, we have mail databases
    that must be shared/common between servers, stateful MAPI connections
    (not stateless HTTP connections), etc.

    Perhaps others with greater experience or knowledge can shed some
    additional light on the topic.

    Scott Lowe, Dec 20, 2004
  7. Why not? One server goes down, another is able to take the load. No
    downtime for users.
    Right. So what you might want to do is look at your business requirements.
    If you require two BE servers, then use a three node cluster with two active
    Part of doing your home work is to understand the business needs and the
    scaling of your solution. I don't think your current thoughts of using 3rd
    party tools will necessarily meet your requirements, especially when it
    comes to support.

    So, with that said, why do you think you need both NLB _and_ server
    Russ Kaufmann [MCT], Dec 20, 2004
  8. Well, while I admit I don't know a lot about clustering I thought that if I
    used NLB and didn't get "High Availability" (failover). I thought it just
    did "Load Balancing" so if one server went down the you were down. Then
    what's the difference between NLB and "High Availability"? Why can you use
    Windows 2003 Standard for one and you "HAVE" to have Enterprise + for the
    Isn't NLB also clustering?

    Just trying to get my arms around these concepts, thanks for all your input.

    Clayton Sutton, Dec 20, 2004
  9. High availability is a term that is often confusing for many people.
    Basically, a high availability solution is continuously available despite
    the failure of individual components and even the failure of complete
    systems. NLB Clustering and Server Clustering both provide high

    NLB provides high availability in that with an NLB cluster, if a node fails,
    surviving nodes take on the load of the failed node. NLB is available on all
    Windows Server 2003 platforms.

    Server clustering also provides high availability in that if a node fails,
    the application/processes will fail over to a surviving node. Server
    clustering requires Enterprise Editions of software.
    Russ Kaufmann [MCT], Dec 21, 2004
  10. And just to show how varible that definition may be, "continuously
    available" refers to "non-stop computing" and MS clusters certainly
    aren't that.

    Specialized hardware is needed to survive failures of "individual
    components" (disk controllers, motherboards, CPUs, etc.). Think of
    machines like Stratus and hardware schemes like Marathon Technologies.
    This type of hardware is "fault tolerant", not "fault resistant"
    (which is what MS clusters are). Fault tolerant hardware "fails out" a
    component. MS clusters "fail over". That's a big difference if you
    really need continuity of service.

    MS clusters have a single point of failure: the share-nothing disks.
    Lose one of those and you might as well be running on a single

    The "continuously available" falls apart when it comes to the time it
    takes to "fail over" a node of a MS cluster.
    Rich Matheisen [MVP], Dec 22, 2004
  11. Thanks Rich!

    That's the answer to my original post! I was looking for hardware solutions
    that will do a better job then a Windows solution. Providing both a "Load
    Balancing" AND "Continuously Available", "Fault Tolerant" solution!

    Clayton Sutton, Dec 22, 2004
  12. Clayton Sutton

    Steve Riley Guest

    By all means, if you've got a business requirement for continuous
    availability, then go ahead and deploy the appropriate technology. I
    hope you're prepared for the significant expense this involves. Make
    sure you've done the appropriate business analysis first.
    Steve Riley

    12/21/2004 19:49:51
    Steve Riley, Dec 22, 2004
  13. I believe 'continuously available' is Russ' definition (after all, they were
    his words <g>). We do not consider clustering to be a continuously
    available solution (or what might be described as a fault tolerant
    solution). Clustering provides high availability, which is a level of
    availability approaching 100%.

    Both NLB and the Windows Cluster Service provide high availbility. If you
    build fault tolerance into your NLB/server cluster design, you can achieve
    very high uptime (99.99% or more).
    Scott Schnoll
    This posting is provided "AS IS" with no warranties, and confers no
    rights. Please do not send email directly to this alias. This alias is for
    purposes only.
    Scott Schnoll [MSFT], Dec 22, 2004
  14. Thanks for the input Scott,

    What do you mean by: "If you build fault tolerance into your NLB/server
    cluster design"? Using a product like... (what?)


    Clayton Sutton, Dec 22, 2004
  15. Agreed, but I was just providing a "generic" definition here.
    Russ Kaufmann [MCT], Dec 22, 2004
  16. :) Yes, it was a "generic" definition to help Clayton understand better
    what is meant by HA in context with NLB and Server Clustering.
    Russ Kaufmann [MCT], Dec 22, 2004
  17. You may want to clarify that then, as the definition is not accurate for NLB
    or server clusters.
    Scott Schnoll
    This posting is provided "AS IS" with no warranties, and confers no
    rights. Please do not send email directly to this alias. This alias is for
    purposes only.
    Scott Schnoll [MSFT], Dec 22, 2004
  18. As Rich pointed out, we can do many things to improve uptime through fault
    tolerance. For example, if your servers have dual power supplies, RAID
    configurations for hard drives, dual connections to dual Fiber switches,
    mirrored RAM, and so on and so on. You can then combine the fault tolerance
    within a server to the next level by having multiple servers in either an
    NLB cluster or a server cluster. Which technology you choose will most
    likely be dependent on the application requirements.

    From my years of working with lock step products like Marathon, I really
    haven't seen much improvement in uptime of an application over a Microsoft
    cluster. When you start getting to 99.99%, you can rest pretty easy.
    Russ Kaufmann [MCT], Dec 22, 2004
  19. I think I just did, but thanks, I will expand here. Just to be clear, this
    is not a Microsoft take on the world of HA.

    Clayton, "continuous availability" is not necessarily what I meant when it
    comes to the truest definition of the term, but it is a term that I have
    found resonates with many people and they begin to understand what HA means.

    Don't worry that Microsoft uses terms a little differently, but if you
    decide to do some googling, you will find the terms are often used
    interchangeably by many major players in the industry. For example, one
    definition states:

    "In information technology, high availability refers to a system or
    component that is continuously operational for a desirably long length of
    time. Availability can be measured relative to "100% operational" or "never
    failing." A widely-held but difficult-to-achieve standard of availability
    for a system or product is known as "five 9s" (99.999 percent)


    Obviously, "never failing" just isn't possible over extremely long periods.
    We all need to understand that HA includes not just the design of the
    hardware/software solution, but it also includes the backup/restore
    solution, and it includes failover processing. Some experts will also
    contend that a true HA environment includes a well documented development,
    test, and production migration process with in-depth documentation. There is
    much to achieving HA, however, it simply comes down to application
    availability through processes, software, and hardware implementations.

    If you use NLB to provide application availability to your users over the
    Internet for your web based app, then that is fantastic. It helps keep the
    application available to your users. The same can be said for server
    clustering, however, you need to take into account the non-availability
    during the actual failover of your application. Sometimes, it is a matter of
    seconds, in other cases it can be several minutes. In all cases, a
    clustering solution will significantly drive down non-availability and
    increase the uptime of your application as run on your servers.
    Many experts state that, for any application or system to be highly
    available, the parts need to be designed around availability and the
    individual parts need to be tested before being put into production. As an
    example, if you are using 3rd party products with your Exchange environment
    that have not been properly tested, you may find that they are a weak link
    that results in loss of availability. Implementing an Exchange server
    cluster will not necessarily result in HA.
    Russ Kaufmann [MCT], Dec 22, 2004
  20. It's not a product. It's all in your design. Things like RAID arrays, SMP
    processors, ECC memory, redundant power supplies, redundant networks, etc.
    You can have an HA solution with NLB and server clusters. As an example,
    have a look at the Exchange Server 2003 High Availability Guide at
    Scott Schnoll
    This posting is provided "AS IS" with no warranties, and confers no
    rights. Please do not send email directly to this alias. This alias is for
    purposes only.

    Scott Schnoll [MSFT], Dec 22, 2004
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.