Defining a storage solution for my college project

Hey there,

I am in the middle of designing a (theoretical) server solution based on Univention Corporate Server. One part of it is configuring a file server for students to use which for it I am going to make a hardware recommendation. Since I am not as firm with enterprise-grade gear I would be happy if you make some suggestions.

Based on my estimations I want to serve a total of 500 accounts with a quota of 15GB each, so about 7.5TB of effective storage (most likely 10-12TB to account for parity). I consider that more than enough for saving documents, however the option to increase the quota is given.

  • I would like to know which RAID-controller I should use for this project. Would SATA3 be enough for this or should I recommend a SAS-based solution and elaborate its benefits (price also matters)?

  • Which brand of drives are usually good? I heard Toshiba is a reputable brand.

  • Which RAID configuration should I shoot for and which tools can I use under UCS (Debian fork)? Personally I have only tried ZFS so far and none of the traditional RAIDs but I figure I will likely go for the latter in this case.

  • (Optional) What would be a good rack-mounted enclosure? Something that supports hot-swap of HDDs would be ideal.

Feel free to pick any of the above. I am thankful for any input! :slight_smile:

As it is theoretical and you request an enterprise level response I have the following suggestions.

Storage is to be provided on the existing enterprise grade SAN device, which may or may not need budget allocated to expand the existing storage.

Is your hardware suggestion going to include running the software in a VM on the existing enterprise grade virtual host? That is what would be expected by IT professionals. I suggest that if you provide a hardware example, you should also specify that the software can and should be virtualised as best practice.

Don’t forget the backup solution, which again should leverage existing backup methods/solutions. You mention students using this. Being able to easily restore the system state quickly may be critical. Virtual machine snapshots etc.

Sorry if this is seen as an easy answer. There are many advantages to this approach, such as the existing IT staff already being familiar with the systems, and that the existing systems meet the required level of availability and performance.

I hope this is of some help.

1 Like

Who cares. It’s just a college project. Just spitball some specs and write it up. Nobody in academia will have a clue as to if what you come up with will work, and that’s not the point of college anyway.

1 Like

45drives have cases that support 60 drives.
WD gold or seagate enterprise He drives should work.
Large vdevs using raidz2 or raidz3 might be ideal.

1 Like

Although @dent_nz is technically correct @NetBandit is actually correct.

3 Likes

I should have thought of that earlier. Thank you for that very useful tip.

Totally. Only problem is the college IT-staff does not want to mention what’s being used in the datacenter so we agreed on improvising here. I will talk about what a sufficient backup solution should provide.

That’s right on point I think. However, the professors grading my paper are having a background in the field so I better not half-ass it too much. :smiley:

Thanks for the feedback. Time to write something down.

1 Like

Go get em! Let us know how it goes !!!

1 Like

I would go with a SAS-based solution, as longevity should also be taken into consideration.

HGST is fancy, but Seagate SAS drives are also good as long as it’s not the 3TB drives.

I would use ZFS, as hardware RAID configurations are becoming a thing of the past with the larger data sets in use.

SuperMicro has some good equipment. But you should also look at recommending an HP or Dell rack solution. As you said price matters, and having a service contract for your hardware is also a good idea.

That being said, is this JUST a storage solution?

Also, is UCS required? or could you propose FreeNAS, XPEnology, or another storage server based alternative?

1 Like

The real answer to this is that you would never want to build your own storage server in a corporate/enterprise or even SMB environment. Mostly due to the fact that you need that warranty. Dell can provide a 4 hour part replacement service, you don’t want to have to compete with that yourself.

It really depends on the number of users and the applications being used. Anywhere from a high end NAS for a very small business with no money and just using it for file or incorporating it into their DC all the way to a Nimble or EqualLogic / Dell EMC SAN. Having you build your own really isn’t going to help you out in the real world where you need to evaluate a business’s needs and find the right product for them.

1 Like

Thanks for the rich feedback!

I also took that into consideration but i didn’t want to mess with ZFS on Linux on that system. I am using it for my own purposes at home though. For actual usage I would have had to deploy it on BSD or Illumos and attach it via iSCSI but i have no experience with both OSes at this point.

Say, I heard the FreeNAS community recommending against virtualizing ZFS. Is it really a bad idea even if you pass a storage controller via VT-d and give the VM a good amount of exclusive RAM (like static hugepages)?

UCS is a requirement as I am primarily using it to deploy LDAP and Active Directory without being tied to Microsoft solutions. It’s the main premise of this project.

Although a software engineer by training, part of my job is to pick ideas, figure out which ones look “interesting” distill them into plans and present them to extremely technically clued in management, so they could come up with staffing and purchasing decisions (it’s just one way of getting things done).

Generally, if you’re looking for budget to sell something to management it actually helps to have several options (3 or 4 alternatives considered, it’s just good form).

In addition to just hardware, each of the alternatives should include estimates of effort required to get the system running, maintenance, and best guess of projected cost of getting off the system once it’s reached the end of its usable lifespan. (basically capEx and opEx)

A thing to factor into cost of hardware is also the cost of recycling and expected value you can get for hardware on second hand market once it’s past its lifespan, or if you can recycle internally - even better.

Also, mention some risks during deployment, and risks during exploitation, for each of the alternatives considered.

If you plan on periodic disaster recovery testing, and have plans to test restoring from backups, mention those as well. (Or mention there’s no plans, or plans are TBD)

Mention any service / support plans / warranties vendors might have.

For the technical nitty gritty,

Backups.

Isolation and abuse prevention (what’s the expectation there and what’s the plan, how much does it cost?)

Given how it’s a storage system, you need to mention if/when you’re going to implement and respond to policies, for example, what’s your legal hold strategy. (If you have employees or students storing data and named in a lawsuit, typically you’re required to not delete data, or allow employees to do so). Inversely, if a data is deleted how long before it’s unrecoverable.

If someone walks out with some hard drives, is the data encrypted? is there any physical security looking after the data?

These are just some of the things that once you’re working in a mature organization become part of your regular internal processes. When coming up with a pitch on why you think something’s worth spending money on, or why it’s worth hiring more people, you generally don’t provide all these details, but you’ll be expected to be able to answer precisely and concisely and provide references to research and other documentation to support your recommendation.

You do your week or two of research, you show up, you go over it again at a high level offering for an opportunity to ask any last minute questions you get your 1 extra staff and 10million in hardware or whatever you asked for, and they’re happy cause you solved their problem and they can focus on something else next, and you’re happy cause you get to go back to your day to day.