The VMware HoL is one of the most strenuous IT environments I can think of. It’s not really “prod” like many. It doesn’t have huge revenue impact for downtime like others. And like some in some vertical markets, it doesn’t have life/death consequences.
But in the challenges of standing up an environment that can support thousands and thousands of VMs, hundreds of simultaneous users, and most strangely – be creating and destroying things (including huge spike workloads of pre-populating labs) that look like the most intense service providers cannot be overestimated.
It’s also insanely visible workload, at VMware’s biggest event.
For the past few years – I’ve done a “peek” behind the curtains of ONE part (it’s a huge effort across VMware and the community of partners that support them) of what goes into a HoL. Read about past years - 2011 here, and 2010 here.
In 2013 – the HoL was supported by N-1 VNX platforms (for NAS and some other stuff), but ran almost exclusively on 4 (yup, count ‘em 4) XtremIO X-bricks. It was a fraction (1/20th) the size, power, cooling – and ran without a hitch at less than 1ms latency the whole time. It also had super-freak levels of TP reclaim and VAAI performance.
It highlights the power of the “clean sheet” design model of XtremIO when the workload is specific. These huge virtual server and virtual desktop use cases are a huge sweet spot for XtremIO.
It’s just as notable that the vast majority of enterprises don’t have ONE workload, but many – and at certain scales, the benefit of a “do everything well” + “use some flash, some magnetic media” VNX-style thing is awesome. Add it to the pile of “why one way all the time is the wrong way” examples.
The fact that XtremIO ran under these crazy circumstances highlights how much we’re putting it through the wringer. In it’s directed availability period (which has been running for months), hundreds of customers have used them, burned them in, and bought them. Anyone who says XtremIO isn’t for real is – well – wrong.
BTW – in past years, use of vCenter Operations at the HoL also accelerated development of official vCenter Operations adapters – even officially becoming available as a simple and easy way to get vCenter Operations with a storage focus from EMC when you buy an array (EMC Storage Analytics is an OEM of vCenter Operations Enterprise). EMC Storage Analytics 2.0 adapter is now out with support for more on VNX and now also VMAX. based on this year, expect it for XtremIO shortly :-)
Itzik Reich (thanks for the video, Itzik!) did a great blog post on the details at the HoL here.
Josh Goldstein, a good friend on the EMC XtremIO team did a post (more pics and videos) here.