I don't think that the dollars are as important as the availability.
This is a client-server application (not a web service which would be
helped by a load balancer) that has the potential to be used by hundreds
of users across campus, and is considered to be "near-critical".
We can pretty much go in two directions - one with high availability,
and the other with fail-over (automatic). I'm hoping for comments on
both. The emails so far on Windows Server 2003 clustering include some
great information. Windows Server 2008 Clustering is said to be
officially supported by this vendor.
Has anybody done any of this with any of the VMWare solutions? I'm
interested in comments on their VMWare Infrastructure products too.
-Nick Kwiatkowski
MSU Telecom Systems
________________________________
From: MSU Network Administrators Group [mailto:[log in to unmask]] On
Behalf Of Vasquez, Timo
Sent: Friday, March 07, 2008 11:49 AM
To: [log in to unmask]
Subject: Re: [MSUNAG] Server Uptime/Failover
To support johns observation with clustering you are using a load
balance service as well with in the cluster. So there is division
amongst the connections and if a server in the cluster become
unavailable the client affected would have to reconnect to the target
cluster IP or app name.
This is the difference between high availability versus Failover. You
can purchase hardware for your servers that will create a seemless
failover for any app you run on cluster environment that cost is a bit
much, but it depends on how you measure ROI. If you are in an
environment that does measure ROI or TCO then know that the hardware
solution for FAILOVER will give you a great position when it comes to
the 9's (99.x.x.x) and your apps reliability.
Here are two I know about that are not so pricy-
http://www.kemptechnologies.com/demo.shtml?gclid=CKKWu5m4-5ECFQLwPAodnTo
Tiw ,
http://www.barracudanetworks.com/ns/products/balancer_overview.php?a=goo
gle-load_balancer&gclid=CLPdufK4-5ECFQJEPAodF1k3jg
If you need that high availability improvement and you don't have a big
budget Indeed the windows clustering can help you make a webfarm/appfarm
that can be reliable and scalable. I am sure there are some scripts out
there and code snippets to keep apps from experiencing connection loss
delays.
Good luck.
Timo Vasquez- D.S.S. Team Member
Michigan State University
Administrative Information Services
[log in to unmask] <mailto:[log in to unmask]>
517-353-4420 ext 249
________________________________
From: MSU Network Administrators Group [mailto:[log in to unmask]] On
Behalf Of schwarzj
Sent: Friday, March 07, 2008 11:07 AM
To: [log in to unmask]
Subject: Re: [MSUNAG] Server Uptime/Failover
I agree with Mike I think you are looking at Microsoft clustering. My
biggest complaint with Microsoft Clustering is that it is not totally
seamless when failover occurs. Then again I am not sure if there is a
failover clustering method that is. This link
http://support.microsoft.com/kb/171277 explains a little bit about
failover time. In my case it just an annoyance and most of the time
users don't even know a failover occurred, because we are talking about
seconds of time here. Unless, they are in one of my JAVA based apps at
the time of failover and they get kicked out which requires a restart of
the program. I am not sure about DB failover times, but I thought I
would just add that the redundancy is great it just comes with some
snags.
Here also is one of the more useful guides that I used to setup my
cluster.
http://www.microsoft.com/downloads/details.aspx?FamilyID=A5BBB021-0760-4
8F3-A53B-0351FC3337A1&displaylang=en
John
From: MSU Network Administrators Group [mailto:[log in to unmask]] On
Behalf Of Kwiatkowski, Nicholas
Sent: Friday, March 07, 2008 8:29 AM
To: [log in to unmask]
Subject: [MSUNAG] Server Uptime/Failover
Nag List,
Looking for your opinion on how to increase one of our applications
uptime. Fist a little background:
Our vendor recently came out with a new version of a previously
rock-solid application/service. Prior versions were internal to our
phone system, actually sitting on a programmable circuit-pack within the
switch. Pretty much it boiled down to that as long as the PBX was up,
this application stayed up as well. While this particular application
is not deemed 'CRITICAL' by our group (that designation is really only
for life-saving/emergency services), it is pretty much deemed
near-critical.
With this most recent release, the vendor has decided to externalize
the application, having it sit on a Windows 2000/2003 server. That
offers us many advantages, such as being able to throw more powerful
equipment at it, and allowing it to integrate with our existing backup
solution. However, as this application got externalized, it no longer
has the resiliency / uptime that it once had. It is no longer the case
that this application would work just as long as the PBX was up.
My question to the NAG list is this: How does one create a server
environment (in Windows) that allows for automatic failover should this
equipment or software fail? I am really looking for a solution that
would allow the application on Server-A to run, and should it fail,
Server-B would pick up without manual intervention. We would also need
something that would be able to share a common IP address (as we don't
want to re-home all the clients manually if/when the failure occurs).
The application is a standard Windows Services written in C++. It
uses Sybase as it's DB in the background. It uses TCP/IP sockets to
communicate to the end-clients. My standard solution of using an HTTP
Load balancer or Java Application clustering won't seem to work in this
case.
I am more personally in-tune with the Linux/Unix world as far as this
goes, and haven't really been keeping up on it in the Windows side of
the house. The vendor has suggested looking at EMC AutoStart, however
in talking with EMC, they won't officially support Sybase DB's, and
their solution may cause corruption to the DB.
Thanks for the help!
-Nick Kwiatkowski
MSU Telecom Systems, P&E Group
