A popular growth area in the datacenter over the past decade, has been the concept of a hyper-hyper-converged infrastructure, and everyone wants a piece of the pie including HP, IBM, and Dell EMC. All of these manufacturers will tell you that hyper-hyper-converged infrastructure is the greatest thing to happen to the X86 space in the datacenter since, well, we started building datacenters. However, like any technology, this is not a one size fits all solution. While there are many great advantages to the hyper-converged infrastructure concept, there can also be drawbacks. In this post, I want to cut through the fat and discuss the pros/cons of hyper-converged infrastructure and hopefully generate some conversation in the community around where this concept fits and where it may not.
To set the stage, let’s first define what a hyper-converged infrastructure is. At its most basic level it is a validated configuration which combines physical compute, networking, and storage resources to form a single “out-of-the-box” solution. Layer on unified or single pane of glass management, validated patch sets across the entire infrastructure, and single throat to choke support and there you have it, hyper-converged infrastructure. Obviously, that is a simplified definition but should serve the purpose for our conversation today. Now that we’re on the same page as far as the definition, let’s take a look at some pros and cons.
- Validated Configurations—How many times have you had to Google yourself to death to figure out what HBA works with what FC switch and is supported by the OS, not to mention researching which firmware/code levels don’t play nice together across the stack. Ok, so you could validate all of that information up front before you do the initial build of the gear, but what happens a few months down the line when patch X comes out for the OS and patch Y comes out for the HBA? Back to square one, then rinse/repeat several times per year. Yuck. hyper-converged Infrastructure solves this problem because the manufacturer becomes directly responsible for all testing of this stuff before the patches ever see the light of day. When the patch “set” shows up at your doorstep, all of that research and testing has been done. Sweet.
- Single Pane of Glass Management—So you have to go into vCenter to spin up a new virtual machine, access the web page of your iLO/DRAC/CMC/whatever to check server health, launch your storage management console to configure storage, telnet/SSH to some random IP address to manage FC/LAN switches, and now you find yourself sitting in mission control with 4 monitors just to do day-to-day management tasks. hyper-converged Infrastructure can solve this by integrating tools into the solution which abstract these various management layers into a single management appliance/utility ultimately saving time and lowering complexity.
- Shortened Deployment Time—Fed up with getting 10 boxes shipped to you to build a single server? There’s a hyper-converged infrastructure for that. Back up the truck to the datacenter door, dump off a few racks, plug in power, uplink network, done. The whole thing shows up at your doorstep fully configured, no assembly required.
- Single Throat to Choke Support—We’ve all lived this one, whether on the playground in elementary school or while working a big datacenter problem. I have problem X, so I call up my server vendor who tells me it’s not a server problem but a storage problem. Ok, call storage vendor for support and get told it can’t possibly be a problem with the array; it has to be something with the storage network. Very well, call FC switch vendor support and what do they say? Obviously it’s not the switch but must be an HBA issue; please call your HBA/server vendor. Wow, round and round we go. I’m now probably 6-12 hours into troubleshooting the problem and made absolutely zero progress. Will somebody please just OWN the issue and help? Hyper-converged Infrastructure will: one number to call, one support organization to talk to, no finger pointing. Nice.
- Vendor Lock In—If you’re used to buying the flavor of the day, that game is up with hyper-converged infrastructure. The only reason the pros listed above are possible is due to the fact that there are no unknowns in the configuration. If you need to add a server, some spindles, some switch ports, etc. you have zero choice from a vendor aspect and must buy from the same manufacturer in order to keep the pros above in play.
- Rigid Configurations—The land of hyper-converged Infrastructure is much more a dictatorship than a democracy. You will have some choices from a configuration aspect, but nothing close to the choices you would have going with a piece/part solution. Again, to keep the pros in play, you have to color inside the lines and only with an approved No. 2 pencil.
- No Tinkering—Do you build your own PC and overclock it? Do you use the latest beta code on your production systems for fun? If so, hyper-converged infrastructure is probably not for you as we’re living under martial law here. You will patch/update/upgrade only when big brother says you can, not when the latest feature is released day 1. By the same token, you can’t simply sit on the original configuration forever. At some point throughout the lifecycle, you will be forced to patch/update/upgrade to remain at a supportable level.
- Price—Make no mistake, you will pay more for hyper-converged infrastructure vs. buying the exact same configuration with the pieces and parts; somewhere between 20-50% more in most cases.
Is hyper-converged infrastructure right for you and your organization? Like I said, there’s not a one-size-fits-all answer to that question. It very much depends on your specific requirements along with a look at corporate culture to make the final call. Hyper-converged Infrastructure has had the most success to date in very large service provider and enterprise datacenters. Why? Because these organizations find a huge amount of value in the pros listed above, regardless of the price delta. They have so much going on in their environments that they simply want the infrastructure to “just work” and don’t really want to have to think about it. They’re too busy trying to figure out much higher level challenges such as automation and thinking of the next big thing that will give them an advantage over the competition. I tend to agree that for these organizations the pros outweigh the cons by a huge margin.
The lines begin to grey a bit in the smaller enterprise and midmarket commercial space where that 20-50% price delta comes much more into the picture. In my mind, the key thing to ask yourself in this space is, will you take advantage of everything that hyper-converged infrastructure brings to the table? Do you want/need a single pane of glass management interface? How dynamic is your infrastructure? How often do you run into support issues? Then we get into the SMB space. You may be surprised that I’m even bring up SMB in this conversation, but several manufacturers have, or are very close to, releasing hyper-converged infrastructures on a much smaller scale to fit the SMB market. I actually see a big benefit to this technology in the SMB space due to the very limited IT resources that you find in most SMB organizations. The “I just want the damn thing to work” concept makes a ton of sense in that space.
So, I’m interested in your comments/ideas. If you think hyper-converged infrastructure is right for you, what are the biggest pros you see in the concept? Likewise, if you don’t think it’s a fit, why not? Let the debate begin!