Sharepoint Forum

Ask Question   UnAnswered
Home » Forum » Sharepoint       RSS Feeds

Network configuration

  Asked By: Srikant    Date: Oct 27    Category: Sharepoint    Views: 1034

Does anyone have, or know of, a post regarding ideal network configuration for a medium farm? My question is around the number of NIC's - should there be multiple NIC's, each dedicated pipes to other farm components, or does that type configuration offer no benefit to the function / speed of the application?

I have 2 seperate farms, with different numbers of users accessing each so I'm open to any suggestions and input.



10 Answers Found

Answer #1    Answered By: Peter Peterson     Answered On: Oct 27

As far as I have seen, there isn't much benefit between having each component with a direct connection to the other ones. So long as all "application" components  (web, search, index) are all on the same LAN segment, they should communicate pretty well. Keep an eye on the router's config though, to make sure you aren't blocking a needed port or protocol. My current farm(s) communicate well across the network, and I only utilize one NIC per device (of course we've got fiber to the SAN for the database server). Unless you are designing a site that will have hundreds of thousands of users  accessing all the time, your standard gigabit ethernet card will handle all the traffic you need. Plus, avoid NLB at all costs with SPSv2...if you use a lot of custom web parts it tends to get a bit messy (from personal experience). That is why I don't use it anymore. :D Besides, a single front end web server can handle one heck of a lot of traffic anyway.

Answer #2    Answered By: Kalyan Pujari     Answered On: Oct 27

So, you're below statement causes some concern as this seems to be the recommended configuration  from Microsoft. What kind of issues did you encounter with NLB? We're in the process of setting up a medium  farm and our two WFE's are configured to use NLB.

Answer #3    Answered By: Laura Walker     Answered On: Oct 27

The content and web parts (mostly just the non-standard web parts) became somewhat unstable. Sometimes they would work right...most times not. Even after duplicative installs on all relevant web servers. Some of the issues were network  related (if you know anything about the Army networks...they're FUBAR), but even after those were compensated for we continued to have issues. If you absolutely have to have some form of load-balancing...I would strongly suggest not using the free stuff that comes with Server 2003. Look for a hardware/software package alternative, if you can.

Answer #4    Answered By: Gopal Jamakhandi     Answered On: Oct 27

Mostly just wouldn't push updates to the other servers or allow me to use non-OOTB web parts (even though they were loaded on both WFEs. Now I just don't see the need to bother because a WFE can handle around 10K-20K simultaneous requests anyway.

Answer #5    Answered By: Kristina Cox     Answered On: Oct 27

A common requirement isn't just for throughput but to allow for HA environments as well.

Answer #6    Answered By: Delbert Frederick     Answered On: Oct 27

We write software that sits on top of WSS and we run that as well as a number  of
other web part (Office web parts, some from the WSS resource Kit, etc) and we
haven't seen this problem. I'm not saying it didn't happen in your environment.
I'm just saying the blanket statement "don't use NLB with WSS" is probably not

Answer #7    Answered By: Kalyan Pujari     Answered On: Oct 27

Given...but I am just tossing my 2 cents in. I know I'm not the best SA in the world too...and time restraints really effect how much troubleshooting I'm able to put into a project. He didn't go too far in depth about his planned topology or his requirements. As far as I know, he may be able to get NLB to function  perfectly within his network. So it could be even more problems with our network  (which doesn't surprise me either).

Answer #8    Answered By: Allison Stewart     Answered On: Oct 27

OK...probably shouldn't have used such a broad statement as that.

"Be wary of using NLB with WSS. It has been know to cause problems in some network  environments."
Hehe...hope I didn't give people any panic attacks.

Answer #9    Answered By: Emmett Hyde     Answered On: Oct 27

I’m for it. I’ve been running NLB for about a year now with absolutely no problems (well ok, I DID forget to update both front ends once) going into a clustered SQL backend. A few times there have been small problems which we “thought” were caused by the NLB but troubleshooting taught us otherwise.

The traffic isn’t going to be major
You are fairly confident that you don’t need the resundancy (this being my biggest concern)

Answer #10    Answered By: Michelle White     Answered On: Oct 27

If you are just looking for good redundancy and backup...I would recommend AvePoint's DocAve and SPDR. It is easily one of the most comprehensive backup/disaster recovery tools I've seen.

Didn't find what you were looking for? Find more on Network configuration Or get search suggestion and latest updates.