RAID options, speed, recommendations

[cs86]

Hello community,

I'm a little new with servers, FreeNAS, and networking so please bear with me.

I have some questions on raid options, and backup along with the speed that goes with it. I'm hoping I'm posting this is the right location. I've tried to do some searching, and haven't come up with the answers that I'm looking for. I have a general idea of how the different raid options work and would like some advise for what I'd like to do if I get a FreeNAS system set up.

A little background on what I'm saving. I intend on using this for small business. I use 3D modeling software to draft projects. The size of the projects can range from 50mb to 1gb including lots of PDF files at times. Through the year I'm guessing I may save up to a total of 10GB. I'd like to have remote access to these files periodically on a laptop. So I think the speed of the network may be of some importance.

Goals of storage and backup: I'd like to have about 4TB of usable data. I think this will be large enough for years until I need to upgrade. I would like to have a type of backup that can backup daily or weekly and maybe back up the server to an external hard drive once a month. Luckily I never have issues with viruses, malware, or ransomware. I've seen other that have experienced ransom ware and want to somehow protect and be able to restore files if necessary. I'm more worried about this if I add users to the network that may stumble onto the wrong site or open the wrong email.

My questions are.
Are snapshots in FreeNAS only available with ZFS type storage?
Are snapshots a good way of backup for hourly, daily, weekly backups?
What type of raid would be advisable for a the type of backup that I've mentioned and which gives decent I/O speeds?
- Is it possible to Raid 2 drives, mirror the 1st set to the 2nd set in 1 week as a backup, and another mirror to a 3rd set in 1 months time?

Since I'm new to ways of backup and servers I'm not sure what my options are and how to go about it.

Thanks for any guidance.

 

[danb35]

Are snapshots in freenas only available with ZFS type storage?

Yes, but ZFS has been the only filesystem available in FreeNAS for the last 3+ years anyway.

Are snapshots a good way of backup for hourly, daily, weekly backups?

Yes and no. They'll protect against inadvertent (or malicious) deletion or corruption of the files, but not against catastrophic loss (motherboard fries killing all your drives, building floods or burns down, etc.). You want backups to be on a separate device, and really at a separate location.

What type of raid would be advisable for a the type of backup that I've mentioned and which gives decent I/O speeds?

I/O speeds really don't sound like a major concern. The simplest is probably two 4 TB (or maybe 6 TB) disks mirrored.

 

[Nick2253]

Through the year I'm guessing I may save up to a total of 10GB.

Did you mean 10TB/year here? 10GB/year is pretty small compared to a typical NAS workload.

I'd like to have remote access to these files periodically on a laptop. So I think the speed of the network may be of some importance.

It's pretty easy to saturate a 1Gbps ethernet link with a typical setup, so your NAS shouldn't be your bottleneck. Your bigger concern is your ISP internet speed, especially since you'll be limited by your upload speed for this application (many cable internet providers have asymmetric speed, with downloads as the top-line big number, and a relatively tiny upload).

In any case, I'd strongly recommend using something like OpenVPN to handle remote access. You do not want to put your FreeNAS system directly on the internet.

I'd like to have about 4TB of usable data.

With only 4TB of data needed, a FreeNAS setup will be much more expensive per TB than something like an off-the-shelf NAS. In particular, the overhead from ZFS forces you into a class of hardware that typically supports 6+ HDDs. Something like a TS-251 with dual 6TB drives would give you 6TB of storage (well over 4TB depending on your tolerably utilization percentage).

What type of raid would be advisable for a the type of backup that I've mentioned and which gives decent I/O speeds?
- Is it possible to Raid 2 drives, mirror the 1st set to the 2nd set in 1 week as a backup, and another mirror to a 3rd set in 1 months time?

With only 4TB needed, I'd probably recommend a single mirrored pair of 6TB+ drives. The other direction your might go is 6x 2TB drives in RAIDZ2, especially if you can find them at a good price. With this setup, you're only "burning" 4TB for redundancy, so even if the 2TB drives cost slightly more per TB, you're looking a better price/usable TB ratio.

If you do decide to go down the ZFS route, I would not recommend doing backups to the same box. ZFS snapshots basically accomplish the same thing. Instead, you want some kind of remote/cloud backup. The only extra thing you're protecting yourself from with same-server backup is total pool loss without total system loss. However, you're still susceptible to total system loss, including any kind of local disaster, and the cost of the extra drives will pay for multiple years of online backup.

 

[cs86]

Did you mean 10TB/year here? 10GB/year is pretty small compared to a typical NAS workload.

I did only mean 10GB/yr. This could be more but I was trying to roughly guess. If I was storing more videos, picture, using it as an email server, etc I'd probably increase the amount of data going into it. In the future if I add employees my rate will go up.

I just found this and was reading on it:
https://constantin.glez.de/2010/06/04/a-closer-look-zfs-vdevs-and-performance/
Awhile back I thought I read the ZFS primer, but I'm wondering if I am misunderstanding how freenas stores data. I'm still trying to grasp the vdevs, and their purpose.

From what I'm gathering is it better to use more drives as 1 large pool (6 2TB drives), use Raidz2 with different levels of snapshots, and then have the server backed up to cloud based storage (carbonite, google, etc. storage). This allows raidz to stripe across 6 drives, give good redundancy, faster read and write speeds, and all the backup I need. Am I making sense of this?

Thanks for the posts.

 

[LIGISTX]

I have to agree with @Nick2253 about buying a pre-made solution as far as cost goes, although if there are certain features of freenas you wish to utilize, obviously this changes the equation. You can get a rather “cheap” little synology or qnap solution that would do most of this (I believe these days they have built in cloud backup options to amazon, google, ect, and decent webui management) with little headache and not a lot of up front investment.

If you do go the freenas route, it is rather robust and allows for great expandability and customization cuz plugins and jails. Also, snapshots do protect against ransomwear as previous posts alluded to. Snapshots only take up space via the delta of the file size, so if no data changes the snapshot will be basically 0 in size, where as if you delete every file, the snapshot will be the exact size of all files deleted. But the beauty is, if you get ransomweared, you can restore to a snapshot prior to the infection and it will delete the encrypted contents and leave your original data.

It is robust, powerful, and a fun learning experience for the tech minded person, but it isn’t “cheap” and requires a good thought process in terms of picking parts and some up front cost in said parts.


Sent from my iPhone using Tapatalk

 

[Nick2253]

I just found this and was reading on it:
https://constantin.glez.de/2010/06/04/a-closer-look-zfs-vdevs-and-performance/
Awhile back I thought I read the ZFS primer, but I'm wondering if I am misunderstanding how freenas stores data. I'm still trying to grasp the vdevs, and their purpose.

Here's the hierarchy for ZFS: device -> vdev -> pool -> dataset. Devices are combined together to create vdevs. Vdevs are also where any kind of redundancy live. For example, if you have 6x 2TB drives, you would put them in one RAIDZ2 vdev.

Multiple vdevs are combined together to create a pool. And then, you can create datasets on your pool. Datasets allow you to control different features of ZFS, like compression or snapshots. The data itself lives in a dataset.

From what I'm gathering is it better to use more drives as 1 large pool (6 2TB drives), use Raidz2 with different levels of snapshots, and then have the server backed up to cloud based storage (carbonite, google, etc. storage). This allows raidz to stripe across 6 drives, give good redundancy, faster read and write speeds, and all the backup I need. Am I making sense of this?

RAIDZ2 is a pretty common setup here because it provides a good balance of redundancy, utilization, and performance, especially for file serving applications. If you need more performance (I/O), then the next step would be striped mirrors.

If all you are doing is file serving with your NAS, then RAIDZ2 should provide more than enough performance. Even a single drive should come close to saturating 1Gbps ethernet.

For cloud backups, a common setup is to create a jail or VM, share the datasets to the jail/VM, and run the backup software there. However, I believe that pretty much any QNAP NAS would be capable of also providing this feature (you may however be limited in supported services).