Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Currently we have a 10 servers setup as a linux virtual server.

We have:

2 Load Balancers (master / slave) 3 Web Servers 2 MySQL Servers (master / slave) 2 Central File Storage Servers (master / slave) 1 Video Converter (irrelevant).

The people we hired to set it up did not anything about GFS, our preferred Central Storage File System Solution. As we have tight time constraint, we just told them to use whatever they were familiar with, which is NFS.

I have since done some research and it seems that NFS is not ideal for us.

I couldn't find much info online about the significant differences between the 2. As far as I have been able to determine, NFS is more suited to home networks, where GFS is much more suitable for the type of setup we have.

Unfortunately, or fortunately (depends how one looks at it) we are facing a total rebuild of our setup.

Server Fault, please can you help me here!

Is GFS really MUCH better for what we want (99.9% uptime, high trafic flow - lots of reading and writing for the filesystem)?

We to choose the solution that is very reliable and also scaleable. Ideally we can add servers as our traffic grows, without (like I said, ideally) being required to take everything offline to rebuild because we chose the wrong central storage solution...

Thanks for your help

share|improve this question
add comment

2 Answers 2

up vote 0 down vote accepted

NFS in a master/slave situation would work for failover.

GFS would allow you to run dual primary which might cause other issues with your deployment, but, would provide you the extra bandwidth for disk IO if you are splitting traffic between the two. You don't say how your current NFS servers are set - are they sharing the disks over FC? i.e. dual head, single shelf? or is it dual head, dual shelf?

GFS's performance is very good, but, because it has clustered locking, files accessed on multiple nodes at the same time will have worse performance than NFS due to different methods of handling locking.

Adding additional nodes to your cluster if you are sharing your shelf is possible. If you are running dual head, dual shelf, adding more nodes becomes somewhat problematic.

share|improve this answer
add comment

Just to add another point here - NFS is much more scalable, designed to serve a large number of nodes, whereas with GFS you will be limited at some point.

Another question, you mentioned a virtual server - do you have all those systems as VMs, or physical?

share|improve this answer
    
NFS isn't more scaleable. Either can handle the same number of client machines, NFS doesn't have robust clustering, which makes running a dual primary failover system very difficult. GFS locking is fairly intelligent and the performance on files where there is no contention between nodes is extremely quick. Once you have contention, protecting the data is worth the speed hit. Since most systems are read heavy, the ability to run 2+ primary with GFS for writes makes it more scaleable than NFS. At what point do the GFS limitations make NFS more viable? –  user6738237482 Jun 15 '10 at 14:25
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.