That should be 99.9997, Chris, not 99.997 although its actually higher than that.
Nimble shows that its all-flashers start small – and grow bigger
Nimble Storage has a new generation of hybrid Adaptive Flash arrays – the CS Thousand Series – plus a new all-flash array entry-point, the AF1000. The all-flash AF series was announced earlier this year, and Nimble has now added a cheaper AF1000 entry-level system. If you buy this, you can scale up and also out through the …
-
Friday 12th August 2016 09:39 GMT Anonymous Coward
Not to be nit-picky, but it is a HUGE difference...
...The availability note reads: "99.997 per cent availability."
This is inaccurate, the appropriate up time percentage is: 99.9997
Already achieving 5 - 9's of uptime is an awesome feat, but approaching 6 - 9's is danged impressive!
Oh, and that uptime is actual measurements via InfoSight, not some arbitrary made up number that someone guessed and published. It can (and is regularly) be proven!
Get some!
-
-
Friday 12th August 2016 18:59 GMT Simon 61
All employees respond - quick...
Quick all Nimble employees respond to Chris has made a typo...are you guys that desperate that you think the storage buying world believes everything from the fingers of Mr Mellor?
Also, more seriously, I'd like to know the block size you use to get your IOP numbers as your datasheet doesn't even state that...naughty!!!
Whilst I'm in that sort of mood, how are you measuring your 99.99(9)7 uptime? Does that include planned outages or not? Do you have the arrays dialing home and sending telemetry so you can accurately measure this? Are you willing to make this data public?
Oh yeah and 5 nine's is pretty poor these days...
-
Monday 15th August 2016 15:26 GMT Nick Dyer
Re: All employees respond - quick...
I'll take the bait...
99.99 vs 99.999 is a pretty serious error on a publication seen, read and digested so broadly in IT as El Reg - whether right or wrong.
Availability and reliability are probably (and should be) the two main criteria for any enterprise class storage solution - so yes it's highly important. You can have as many bells and whistles in a system as you like, but if the thing keeps falling over or corrupting data then it's not worth the fag packet the design was done on.
Fun fact: Nimble systems are actually running at OVER 99.999% availability. This is measured through 5 minute heartbeats of every single array deployed in the field, which currently stands at somewhere over 16,000 systems. We will happily provide the data for any customer that's interested. It's called Infosight Labs - and it's very cool.
By the way, there aren't that many other vendors that shows true >5x9 availability through actually monitored data from real-world systems - rather than 'built for' or 'lab tested'. Most just throw an abundance of overspec'd hardware at the problem to ensure that it will get >5x9 in liu of not being able to monitor and proactively fix issues, such as 3PAR, Infinidat and others.
Performance figures are driven by 4K blocks for IO, 256K blocks for GB/sec. However because NimbleOS is defined by application rather than LUNs the array is able to drive variable block alignment, compression, dedupe and performance for each app. Here's a good read as to why we do that: https://www.nimblestorage.com/blog/storage-performance-benchmarks-are-useful-if-you-read-them-carefully/
-
Tuesday 16th August 2016 00:06 GMT Androgynous Cow Herd
Re: All employees respond - quick...
Nick missed one thing from the pointed questions Simon flung. The 99.9997% uptime is inclusive of array software and firmware updates. Most upgrades to either performance and capacity of the array (two separate things once your storage is no longer spindle bound) are non disruptive to production as well. You can scale that itty bitty AF1000 up to over a million (4K) I/Ops and Petabytes of capacity - non disruptively.
It's pretty cool stuff, and that's why the employees get excited to talk about it.
-
-
Biting the hand that feeds IT
