I am following a discussion on an investing forum where people are talking about the relative virtues of Nutanix and VMWare as alternate solutions with odd mentions of Simplicity. It seems to me that I see fairly frequent references to VMWare here, but don't know that I have ever seen a reference to Nutanix, even though they are supposed to have similar market penetration. Anyone got any on the ground insight to offer?
We have run into them. But much less often than VMware. The sites where I have seen them suffer from horrific performance problems that are not due to anything Progress related...
Is the performance problem DB related or just general? Any idea whether that is just not being set up right?
Next year, a client of mine is moving their back end from bare-metal HP (which performs quite well) to Nutanix. It will be interesting to see how that goes. I haven't seen any specs from them yet.
Interesting too since HP is heavily pushing Simplicity over Nutanix,as I understand.
As usual the DB is what gets blamed. But so far as I can see the db is not actually "the problem".
Something in the layers beneath the db is messed up. There are various candidates but IO performance stands out as being especially putrid. I don't find that very surprising - it's basically yet another supposedly magical storage layer that nobody in their right mind would put a performance critical database on. IMHO.
"On the ground" all of these fancy infrastructure things suffer from a common problem. They get implemented universally within organizations but in a generic manner by people without adequate understanding of their strengths and weaknesses and the particular needs of the applications running on them. A mission critical ERP application is not the same as a conference room scheduling app. Or a web server. Etc. The end result is all too predictable.
And, in particular, many of them deal with small files and low transaction rates and work just fine on this sort of thing, but a large DB with high transaction rates puts a whole different load on the system. Yes?
Correct.
How much of that is lack of necessary education (or effort) and how much is inherent to the product? I.e., is it correctable by proper setup that recognizes the difference or is it simply the case that high transaction DBs just belong on their own plain disk?
It is probably possible to get "ok" performance *if* you can get the admins to actually pay attention and get beyond the perfunctory "everything is fine on our end" auto-responder. (Just as it is possible to get vmware to perform ok if someone pays serious attention to the configuration.)
No external or shared storage device is *ever* going to be as effective as internal storage. 1 nanosecond per foot is the law of the land and always will be. Furthermore every layer of logic that you insert along that path is additional microseconds (or worse) of latency. If you really need high performance you are fundamentally on the wrong path if you implement on an external device. And you double your trouble if it is shared.
Can it be "good enough" for most people/applications? Possibly.
But it rarely comes that way out of the box and if you really need the absolute best performance then, no, this kind of solution is not appropriate.
Thanks for your insight. The stuff I was hearing sounded much too glowing ...
Our customer is using it. After solving some serious coding issues (it's always the coders) in the core applicatione that application is running fairly well. Fortunately it does not do a lot of writing. Other applications still suffer like office.
Replacing rotating rust with solid state and implementing Nutanix was THE silver bullet. Unfortunately this investment did not bring them what they hoped for.
Frankly I don't really understand what Nutanix is and to me a SAN is just an external drive that is accessed through the network. So I did the Furgal test and presented the results and telling them they have a serious problem. This may not have any relation to Nutanix but to me the matter of simplicity relates to incompetence at this customer so be careful.
The Furgal test took 18 seconds a year ago, last week it took 20 seconds. Number of users dropped with a third.
virtualizationreview.com/.../nutanix-and-the-problem-of-performance-testing.aspx
regards,
gus
“less is my favorite editor. too bad it can’t actually edit files.”
Chris Lesniewski-Laas
Interesting, Gus! Thanks!
curious, what is the Furgal test?