It’s an old saying; “A picture is worth a thousand words.”
I believe this still holds true, but the problem is it really all depends on who is the viewing audience of your picture and if they truly understand the deeper meaning of it.
A couple of days ago I posted a picture of a Nutanix installation in progress, after spotting it on Twitter, posted by one of our GSO guys.
It’s a pretty impressive sight, 100 nodes, spread over 7 cabinets with plenty of room to spare, 1300 disks spread over mixed model types, and delivering a total of 2 PB of storage. It includes all the switching and compute power to run the workloads on it as well.
The reason I posted the picture was to counter some of the FUD that got thrown our way recently from a legacy 3-tier company that said Nutanix couldn’t scale or be used for anything other than a small VDI deployment and that the economics didn’t work out.
Josh Odgers has made a couple of posts on it as well, which you can read here.
Loads of people liked it, and many others (coworkers, partners, community people) also tweeted or posted the picture, appreciating the innovation going on right there.
And then there came the sceptics.
“It doesn’t prove anything!”
“It says nothing about scale!”
“How’s the performance?”
“I can also show stuff in a rack!”
“Show the white-papers!”
Of course not all comments where made in bad faith or came from a nay-sayer perspective, but some of them did miss the point of the picture.
To answer the first 3 objections:
It proves very much the initial point I was trying to make regarding debunking FUD. It shows large environments can be based on Nutanix. It’s also clear this isn’t your typical SMB-style 250 VDI desktop deployment, and if economics didn’t work out, how did the customer buy it in the first place?
To even come to projects like this, an extensive PoC, in depth ROI/TCO calculations and a solid business case are required, combined with top to bottom buy-in across all levels of the organisation. Also you would have gone through a rigorous selection process, being in competition with and having to prove your worth over numerous other vendors over a period of several fiscal quarters before a customer would even consider signing a PO of this magnitude.
The fourth one, “I also can show stuff in a rack” is also missing the point. Yes, you can show stuff in a rack too I’m sure, but I very much doubt you could show the compute power of more than 200 CPU sockets and storage of over 1300 disks in the same footprint that you would need with a 3-tier architecture.
The picture shows you 7 42RU racks, filled with all the switches needed, and room to spare for future upgrades. If you would just stack for density, you could easily remove 3 racks if you wanted. The amount of power and cooling required compared to a 3-tier architecture of the same capacity is also drastically lowered. If you want some examples with math included, please go read my blog about footprints here.
The last point, “Show us the white-papers“, has been answered many times already. Just look at the Nutanix website to read about customer deployments. Read what they had to say about the deployment and what they gained from it in business value.
But I know what you are aiming at with the white-paper request. You want the dirty details. The number of IOPS, the amount of VM’s per, the VDI boot-times etc.
Don’t worry, it’s a fair request if you are into that kind of stuff and we’ll turn this into a readable document soon. Just be reasonable and don’t expect us to post a complete documentation of every environment online. We have some huge customers in U.S. Federal that would blow your mind in terms of scale and architecture, but their policy prohibits publication. However, here’s a sneak peak of a recent Nutanix win together with our partner Dell, where an environment of 600 Dell XC appliances, will be deployed in the first phase of a massive VDI deployment at the FBI.
But then I am going to ask what do you want to read in such a white-paper? If you’re interested in IOPS, you are better of reading benchmarked Reference Architecture documents instead of the design document of this environment. This design is fully catered to the customer, making sure it can deliver the performance and scale that is requested. I’m pretty sure they aren’t going to run IOmeter tests 24/7, and they won’t hit any IOPS limit anytime soon.
If you expect IOPS to be a main design choice here, you are probably gonna be very disappointed. IOPS is not a design element anymore since in a Nutanix environment there is so much IO performance released by removing the constraints of 3-tier, asking about IOPS is almost the same as wanting to know the response time of a LCD screen you’re gonna use for playing Call of Duty, or if your switch has cross-over ports. It’s solved, done, no need to worry about it.
The real and most important value is going to come from faster job completions, easier management, flexible scale, reduced planned and unplanned downtime, TCO reductions etc.
The writeup of the environment in the picture will arrive in due time since installation has only just finished, but this hasn’t been designed on the spot. VCDX certified experts have designed and are building this.
In the mean time we already have reference architectures for environments bigger than this posted on our website, and also loads of articles including all the details on different workloads and real world performance on the blogs of Josh Odgers, Michael Webster, Dwayne Lessner and Lukas Lundell.
But since I am still a man of pictures and this post is exactly 1000 words, I’ll just leave this here: