Hyper-converged infrastructure? Pop open some 'Azure in a can' HPE and Microsoft have teamed up to give everyone else a right good kicking. The product of the now is the "HPE Hyper Converged 250 for Microsoft CPS Standard”. This is a very long name and so I propose instead that we shall think of it as "Azure in a can 250". Sticking to the Azure-in-a-can theme, HPE's new gizmo does exactly …
I can buy a Dell 4 Nodes in a Single 2U Chassis, 48 Cores, 128Gb Ram, 14TB Storage (24 x 600Gb 2.5 SAS HDD) for six grand and put a lot of software, MS or other, on it for another £49 grand, if the $:£ were at exchange rate, but I guess the HPE rate will be $80K = £80K+VAT.
That MS software is a pig to set up, configure and maintain. If you can do that, great! You can save a pile of money. This is provided all pre-canned, tested and supported by the vendor. Zero knowledge or expertise required. Unpack, plug in, start using.
There are a lot of shops that just don't have the skills to futz with Microsoft's byzantine bullshit. There are plenty of shops that might have the skills but just do not have the time. Azure in a can is fantastic for both types of shops and lets the technical types focus on things that actually make the company money instead of fussing about making infrastructure go.
I've said for ages that "businesses don't make money resizing LUNs". Well the same principle applies to infrastructure as a whole. I can build an entire datacenter out of Quanta computers and KVM/Openstack that I compiled, imaged, configured, manage and maintain myself. I'd be a complete idiot to do so unless I was planning to play the game at Facebook scale.
And no, this isn't some richy-rich enterprise sysadmin from the land of muchos budgetos saying this. It's me: SMB sysadmin from the land of "outrageously poor" holding court. Individuals with technical skills good enough to fight through Microsoft's software stack and set up an hyperconverged hybrid Azure cluster are misspent actually doing that. Their skills are rare enough that they should be doing things that make a hell of a lot more money than you'll save pissing around with infrastructure.
Maybe you pay 3 months or even 6 months worth of one technician's salary by using Azure in a Can. That's about the time it would take to fully test and certify for deployment your home grown version. But if your tech is good enough to do so, he or she should be spending those 3-6 months automating things, working on information security and migrating legacy workloads to failure-aware models. That will save your company way - way - more money than the 3-6 months salary you pay out "extra" to just not have to worry about setting up a private cloud.
Trevor:
While I share your optimism regarding the Azure-in-a-can from HP, the devil is in the details, and being that HP is involved, there are a couple of details that I would be really worried about:
(1) If this is anything like their VMware + StoreVirtual hyperconverged solution (which I think is the same box, but different software), it will not scale. We demoed the VMware + VSA version last year and from what the installation technician said, HP was working on being able to add another 4-node box to an existing cluster. If this were focused only on the S end of the SMB market, then that may be an acceptable tradeoff, but I could see an M (especially a higher-end M) running into problems.
(2) IIRC the 4-node sled is a part of their SL series servers, but it is a "weird" hardware configuration. From what I've seen of their "weird" hardware configurations (*cough* E5000 *cough*), the engineering is pretty crap, and hardware problems are to be expected. We made the mistake of buying one of these "engineered solutions" (*cough* E5000 *cough*) and have had TONS of problems, and even their sales people have admitted that the product was crap.
(3) Related to (2) above, often there isn't alot of knowledge in HP support about the "weird" configurations, and often even the simplest questions and fixes need to be escalated to L2 support. And even when you've been escalated (which takes time, and the response can be lacking at times), there is alot of conflicting information because of the lack of internal knowledge about the product. Compounding this is the complete and utter lack of information on the Internet about problems with this particular solution, which makes it so you MUST call HP support for even the most minor questions.
(2) and (3) above create a perfect storm of shit for an SMB admin. A critical component breaks, you're worried that the same component breaks in another blade, so you call HP support, get tossed around for a while and a week or two later it gets fixed, just in time for another critical component to break. You spend a ton of time just maintaining this weird configuration until the warranty runs out and you get the OK to replace it, and this displaces time to improve things in the business. This was my experience with the E5000, where I spent almost a year opening cases with HP every month to fix ANOTHER component.
I understand why they did the 4-node 2U sled, but IMO it would have been a safer bet to do this with 4x DL360 Gen9 servers that are heavily customized (similar to the StoreVirtual physical appliances).
Scale have been established for at least 6 or 7 years, therefore they are not a startup, but an established business! I first had dealing with them in the UK in 2009/10 and they had already been going in the US for several years before that.
Sorry if thats being pedantic.
Trevor, what's your opinion on the equivalent Dell offering? Dell got a lot of noise out of their announce since it was a few months before HPEs, but I have no idea on whether it's good or not...
http://i.dell.com/sites/doccontent/shared-content/data-sheets/en/Documents/dell-hybrid-cloud-system-for-ms-solution-brief.pdf
Nutanix IMHO priced themselves out of this market. The story was great the product was too, but at over $100k to start with, they basically killed their product as an SME offering. We wanted to take it to market as the "on prem" component of a hybrid cloud solution, but the numbers simply didn't stack up. Our customers simply failed to see how a single SuperMIcro box with some software on it was worth that much. The Dell equivalents were simply much cheaper.
All-flash hyperconverged system, exceptional Microsoft integration. Can run Windows Azure Pack (WAP) on it, no problem. More scalable, greater performance and, from what I've thus far, less cost than the HPE version. You should have a look Trevor. Read this.... http://www.gridstore.com/press-releases/2015-03-03-2/
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
