SAN v s NAS

Question

I am fairly new to the advanced storage solutions and I thought I had a decent understanding of the technologies. We currently have a Fibre Channel SAN in our environment which is directly connected to a PE2950 via a Fibre HBA. We currently have the entire 3.7TB of space allocated to this single server and created a share on that Volume which we then mapped to 15 - 20 users who process their work. In a technology meeting with management, a new person working within the group stated the following which I am failing to understand. I am hoping someone with more advanced storage experience will be able to either explain why this is so or why he is totally wrong. Thanks in advance.

The best configuration for any ediscovery application on an enterprise or multinode level whether it is for processing, forensics or something else is to load source data to NAS and to deliver output to the same.

I have researched this extensively over the last 3 to 4 years. That said there are specific advantages to maintaining data integrity, access to files and other things where a NAS is a better solution. That said, for pure file storage, regardless of the seemingly favorable comparisons between SAN and NAS arrays, for the kind of work we do NAS for storage is ideal.

Andreas Erson · Answer

SAN = A network dedicated to storage communication. The definition says nothing about the type of communication but most common are FibreChannel and iSCSI. The servers attached to a SAN gets block-level access to the storage devices. The server then applies the filesystem of it's choosing on top.

NAS = Networked data storage that gives clients file-level access with most commonly NFS and/or CIFS. The NAS is a storage device combined with an operating system that is solely dedicated to providing network access to file-level network storage. The operating system needs to be able to put a filesystem on the storage device and manage the shareing of that storage on the network.

As you can see, in essence there is no point to talk about SAN vs. NAS in your situation. Your PE2950 serves the clients file-level access. And the SAN serves your PE2950 with block-level access to more storage. If you put your PE2950 and the SAN in a bubble you could easily call that a NAS. Just because a NAS usually has a dedicated file serving role and a stripped OS to make that possible and just that, it doesn't mean that any general server could be used just like a NAS (aka fileserver). So if you only use that PE2950 to serve those 15-20 users with file-level access you should be able to put a stop to any further discussion about this by simply stating that you have a NAS with SAN-attached storage since that is what you in reality have.

This image represents my argument:
http://upload.wikimedia.org/wikipedia/en/thumb/0/05/SANvsNAS.svg/750px-SANvsNAS.svg.png

Good luck!

Regards,
Andreas Erson

AlexAnastasia1 · Answer

Hi Andrea, and thanks for the reply. I believe that we are saying the same thing here which is that a fileserver providing file level access to a client is essentially a NAS. The primary reason we have the SAN on the backend is because in our environment we receive between 5-10GB of data per day from clients and post production work will usually results in total data at 30-40GB per day. We are required to keep data on production for a period of no less than 90 days, so you can see why a SAN on the backend was purchased.

What I am trying to understand is the last paragraph of my original posting which is stating that a NAS device on the network will out perform a PE2950 running Server 2003 connected to a FibreChannel SAN via 4GB HBA. Is there something on the client side that would allow any NAS device to out perform our setup?

Andreas Erson · Answer

Yes, I am in full support of your current solution. Since you have very powerful hardware in your PE2950 it will for sure outperform pretty much any NAS on the market as long as you don't use it for any other taxing workload. And a 4GB FC SAN should most definately perform better than most local storage than a dedicated NAS would have. Dedicated NAS-boxes often has just your run-of-the-mill lowend server parts that runs it's stripped down OS.

And since your solution can act as a NAS in pretty much every way there isn't any advantage in a dedicated NAS except perhaps price, rack space and most likely setup time. But since you already have your setup up and running all of those factors don't really matter do they.

Since the file-level access to a dedicated NAS or your current "NAS"-solution would most likely be the same I don't see that anything on the client could make any difference. And I don't know of any technology that would improve a NAS-solution that reside on the client side either.

There could actually even be an advantage to have your solution instead of a dedicated NAS if you use Win2008 and Vista on your clients since you could use SMB 2.0 which would probably improve performance significantly. Add jumbo frames capability to the mix (every part of the network between your clients and your PE2950 must support it) and you would have pretty much the best possible performance possible in a NAS scenario. Most dedicated NAS-boxes don't support SMB 2.0 since they are not based on Win2008. That will change when dedicated NAS-boxes are released with the new Windows Storage Server 2008.

Take a look at Dells NAS solutions and you'll se that the most powerful ones are somewhat similiar to your solution.

Should also mention the latest trend with a hybrid NAS/SAN. A networked storage unit that can give both file- and block-level access.

FluidFS

SAN v s NAS

Was this post helpful?