1

Since our school got some new hardware recently, i got the opportunity to renew our file-sharing Infrastructure.

All students and teachers have their own home-drive. Teachers and students also have additional drives for different uses (classes, resouces, etc.). Login-Scrips connect the drives.

Until now, we had a very old filserver (windows 2003) for the home-shares and an old readynas for the network drives (via smb). Speed is an issue, because sometimes, big files need to be read simultanously (eg. video-files).

All other servers (ADCs, Print, WSUS, …) are vitualised on a VMware-Host (Proliant 380 G9).

We now got a readynas 3138. (Raid5). I thought about sharing NAS-Folders (inluding home folders) directly and use teaming to speed up a little. Another option would be to create a VM for the FS with an iscsi-drive on the NAS. I dont have lots of experience but is the NAS-Option better when favoring speed?

4 Spice ups
2

Welcome to Spiceworks.

Would go for some NAS or better a Windows Server (VM) with a DFS in front.

This way, you can easily scale and redirect single shares to new servers using DFS without anyone noticing.

1 Spice up
3

I agree with going DFS on a windows fileserver vm with your readynas. Does it have a fiber port?

4

I agree with both of the above. I only use iSCSI for multipathing to hypervisors or dedicated server storage. iSCSI doesn’t scale well with multiple users - you’d have to create a separate target for every user, and map it to their system.

3 Spice ups
5

Oh, I didn’t think about DFS. Thanks for the replies. i’m glad i asked here!

1 Spice up
6

I think I’m understanding what you’re saying, if not, you must be mis-informed. iSCSI is a SAN technology, much like Fiber Channel, so it scales extremely well with demand if you setup your LUNs correctly. Are you talking about setting an iSCSI on each workstation, then mapping the LUN to the workstation? No, nobody would ever do that. You’d set all that up on the file server (let’s say File01), then go to your domain controller (let’s say DC01) and set your Active Directory home folder for users to that target, and have the home folder mount as whatever drive letter you want when the user logs in to a workstation You can do that, or you can do folder redirect which is basically like doing NFS shares on Linux. Either way is acceptable, I typically setup redirect though I find home folders to be much easier to troubleshoot but a little more cumbersome for users.

The only real question would have been “NAS or SAN?” not “NAS or iSCSI?”

Back to the original question, it really depends on how many users you have, and if you even have a SAN currently. If you have less than 3000 users, and this is just for simple file access like docs/etc, then you could probably get away with a NAS just fine. If you were asking about putting VMs/high transactional databases on the storage, we’d start talking about block level storage (SAN) instead of file level storage (NAS) as there are other benefits (iSCSI/FC) that come into play with high throughput needs.

2 Spice ups
7

Also, you’re thinking of teaming your NAS NICs for speed. It’s not just used for speed, it’s also used for failover should a NIC fail. Might want to consider that when you’re setting them up. I typically see a 2 out, 1 in, 1 disaster failover (on a quad NIC NAS). The reasoning being:

*Most of the data on the NAS is going to be read, not written, so you want more available NIC traffic flowing from the NAS.

*At least 1 NIC is used for writing to the NAS.

*1 NIC for failover. You don’t usually want more than 1 NIC as it is strictly for disasters. So slow traffic is okay while you troubleshoot, while more traffic is available during X% of the other up time.

You can also consider (on a 4 NIC NAS), 2 in and 2 out. I’ve saw both setups in many environments. Both have benefits. I’ve traditionally setup a 2x1x1 but have been switching to a 2x2 mostly because it gives the most throughput that’s balanced, and there’s no worry of failover because if 1 fails on in or out, there’s still 1 other NIC available. Doing it that way doesn’t take away 1 NIC just to be used as a standby.

2 Spice ups
8

I understand perfectly - I think we have a misunderstanding. I was talking about scale in the frame of what the OP was trying to accomplish - like you said, no one would ever do that, because it’s not a scalable solution and not what iSCSI and other SAN technologies was created for. Managing a separate LUN and target for every user in an organization would be a nightmare. That’s why I said I use it for hypervisors and server storage.

In the respect of scaling for capacity and size, yes, SAN does this well. That’s what it was designed for.

1 Spice up
9

You’re loading many big files simultaneously. Is your file server the bottleneck? Can your LAN manage the transfer?

I’m just wondering if a faster file server is going to address the file transfer issue - which I assume is what you’re unhappy with.

10

DFS-R or a clustered SMB3 file server like on this scenario:

(both are free options either way so think about manageability and other requirements like application support here)

2 Spice ups
11

Nice. Just did a quick scan over your article.

Out of curiosity: What about DFS-R on top of clustered SMB3? I like DFS most for as kind of a directory for shares. Great if something is changing sometime in the future. There are other functions, sure, but this is my primary use-case.

1 Spice up
12

What for?!? SMB3 is already clustered so why do you want an extra copy of your data done THAt way?

1 Spice up
13

Just a theoretical question. Was just thinking about a cluster per physical location getting replicated using DFS-R.

But there are multiple possible approaches for this :slight_smile:

1 Spice up