"Built-for-Enterprise" vs "Fully-Custom" Rack Server

[malkortechie]

I'm quite experienced in putting together workstations and laptops from salvaged parts: my current TrueNAS server is a salvaged Dell workstation with the HDD's actually sitting on _top_ of it. This is fully function, but I've got my Christmas money and I'm ready to get a bit more professional. Also, I have LOTS of experience with consumer-side equipment but very little with proper back-end/server kit.
I want to move to a rack server but I've only enough money to do this once: and I'd like some experienced advice as to which path to choose. A "Built-for-Enterprise" server - such as a Dell Poweredge R710 - would let me not only learn administrating server software but also proper server hardware: like I might encounter if I move from the Help Desk to the Systems Admin area. A "Fully-Custom" Rack Server will give me flexibility of design: I can short _installed_ memory now for a board which will be upgradable for longer. Going fully-custom, however, will prevent me from getting experience with enterprise-grade equipment.
I'm nearly 50 years old but I'm not planning to retire anytime soon: so I need to learn as much from this as I can, either way.
I would really appreciate your informed comments: to help guide my decision. Thanks for your help.

 

[ChrisRJ]

Can you be more specific (e.g. an example) what you mean with "fully-custom rack server"?

If it means for you to buy a rack-mount case from an unknown company, do the same for SATA enclosures, etc., I would recommend against it. The relative ease you have with hardware today, compared to the 1980s/early 1990s, comes from standardization. Or in other words: not leaving the trodden path. If you buy a server from a reputable company, you have a combination of components that exists thousands of times. And it has been used millions of hours in demanding settings. Going for something custom, means that you loose those advantages.

What you should do in my view, however, is allow for quite some time to learn, plan, and think, before starting to implement this. As an example, when I got myself a new NAS in summer of 2020, it was a 3-month process just to come to the point that I wanted neither a Synology, nor a used rackmount server, but build something (see signature). In total I spent about 5 months, incl. reading a lot here on the forums. You will learn an awful lot from others that way.

Also, being clear about your use-case is critical. What I read very often is something "I want to do light video editing". This is rather vague and can mean very different things in terms of required hardware. First: what is "light"? It is not quantified and will mean different things to different people. Second: What does "video editing" mean? You could do this directly off of the NAS, or copy from the NAS to you editing machine, perform the work, and transfer the result back. There are orders of magnitude between those two scenarios in terms of hardware required. This is just an example, but it hopefully illustrates what I mean.

Taking the core of the paragraph above to a more technical level, think about what you need in terms of disk performance. Many people reduce the latter to sequential read speed for large files, which is a best-case scenario. But you want to plan for the worst case, that applies to you. If you want to use the NAS as a storage backend for VMs, your worst-case would be synchronous random write operations. So my NAS, which is pretty much used as an archive system, would be totally unsuitable for even a single VM, unless speed is absolutely not an issue. But for my use-case it is perfect. Again, just one example and there are other factors as well.

In total, when looking at myself, reading a lot on forums has always been how I learned such things. In the mid 1990s it was Novell NetWare so that first I would run my own server, and two years later offer consulting services around it. Personally I think one cannot outsource that kind if learning, if you want to be sure to end up with a result that makes you happy.

Hope that helps!

 

[jgreco]

A "Built-for-Enterprise" server - such as a Dell Poweredge R710 - would let me not only learn administrating server software but also proper server hardware

It will also present some potentially annoying and significant challenges for use with FreeNAS or TrueNAS, because the Dell servers tend to be highly integrated, and sold in configurations that contain useless crap like RAID controllers that are a bit difficult to remove. Basically all servers built by the common vendors (HP, Dell, IBM, etc) for large numbers of drives will integrate a RAID controller, but this is wrong for FreeNAS and TrueNAS.

Going fully-custom, however, will prevent me from getting experience with enterprise-grade equipment.

It isn't really clear what your concept of "enterprise-grade" equipment is.

If you mean something made by companies who then back it with a 4 hour onsite warranty, that limits you somewhat.

However, lots of companies buy their servers from Supermicro or a small list of alternative vendors. Supermicro sells preconfigured chassis and also some prebuilt systems, and there's a large number of systems integrators who sell specialized systems. That includes both iXsystems and one of my companies. Supermicro has the upside of building standardized components which you can then use in ways that perhaps few others have considered. This is optimal for FreeNAS and TrueNAS, because you can get rid of the frickin' RAID controller and just BUILD a system to have an HBA as part of the design.