The consumption models, elasticity, ease of use and agility that business stakeholders want and need does not have to be limited to public cloud companies. A similar experience can be delivered with an on premises private cloud or enterprise cloud infrastructure, provided you deploy the right architecture in a similar way as a cloud provider would. However this has been cost prohibitive for many until recently and there hasn’t been a similar architecture available that realistically allowed consumers of cloud the same benefits as the public cloud, but in their own environments. This is all changing.
Some tough questions need to be asked. Why can’t you deploy racks of capacity within a few hours, on demand, like the public cloud companies can? Why can’t you run business critical applications with the predictability and availability that you need, but with the agility to get more releases to production with less defects and higher quality testing and in less time. All without worrying about if your availability zone is going down, or your system will be deleted as the single server it runs on is being shut down and removed for maintenance (actually happens in some public clouds). Why does the infrastructure have to be so complex in the first place and why is enterprise IT not as simple to use as the public cloud, with API’s that can drive it, so developers are productive and can deliver competitive advantages to the business?
These are some of the questions that Nutanix Enterprise Cloud Platform is trying to answer. Not just about the infrastructure, but services that sit on top of the infrastructure and for both elastic and predictable workloads, with consumption models to fit customers demands, and rapid provisioning just like a public cloud, but at any scale. Including starting large and going larger, or starting small and growing from there. Nutanix President Sudheesh Nair tweeted out some thoughts about this very topic that explain where Nutanix is heading and how you can embrace and benefit form it, or be run over by the tidal wave / Tsunami.
https://twitter.com/sudheenair/status/855563291728519168
https://twitter.com/sudheenair/status/855563652409241600
https://twitter.com/sudheenair/status/855563955464540160
https://twitter.com/sudheenair/status/855564131285577729
https://twitter.com/sudheenair/status/855564265486471170
https://twitter.com/sudheenair/status/855564406448685057
https://twitter.com/sudheenair/status/855564535972995072
https://twitter.com/sudheenair/status/855564685017583617
https://twitter.com/sudheenair/status/855564824151023617
https://twitter.com/sudheenair/status/855564972432252928
https://twitter.com/sudheenair/status/855565098584297472
https://twitter.com/sudheenair/status/855565234525929473
https://twitter.com/sudheenair/status/855565361030340608
https://twitter.com/sudheenair/status/855565483990560768
https://twitter.com/sudheenair/status/855565613258784768
https://twitter.com/sudheenair/status/855565744658055168
https://twitter.com/sudheenair/status/855565923574566912
Final Word
A lot of problems in enterprise IT have been caused by infrastructure teams not talking to and understanding the requirements of applications teams and the business units they support. Many problems could be solved by an open dialog and understanding. The reason shadow IT is even a thing is because public clouds are filling the void of capability that app and business teams require and that enterprise IT infrastructure teams are not meeting. There is no reason this can’t be reversed and we can greatly improve all aspects of how enterprise IT services, applications etc are delivered from the ground up, while leveraging public cloud resources where it makes sense. It isn’t an all or nothing decision and there is no reason you can’t bring the public cloud experience to your own cloud.
This post first appeared on the Long White Virtual Clouds blog at longwhiteclouds.com. By Michael Webster +. Copyright © 2012 – 2017 – IT Solutions 2000 Ltd and Michael Webster +. All rights reserved. Not to be reproduced for commercial purposes without written permission.
Just my two cents, but my biggest issue with HCI is what some people see as a big win, that is computer and storage scaling together. I hate that, like REALLY hate that as an architect. I’ve been holding out hoping that some HCI vendor would wake up and kind of realize that scaling compute and storage together has limited use cases, and creating a solution that’s more independent is what customers really need.
If we’re being honest, cloud isn’t just about decreasing time to deploy, and being able to scale quick, it’s also about saving money. Public cloud has the win it does by being able to scale up AND down. On prem typically doesn’t have the scale down ability, which is a big part of the cost saving in cloud. So if I need to purchase capacity up front, I can also build a solution that’s just as scalable as something like Nutanix using VMware + SAN/NAS. I mean SDRS + a bunch of SANS basically does 95% of the work that an admin would need to do, if you just want an easy to manage hands off solution. Carving out volumes really isn’t that onerous, and setting up new school SANs is a piece of cake. VMware’s compute with DRS already manages itself well, and i’ve been using auto SDRS for years with no issues. In fact, I would contend that SAN provides more scale and more capability than even Nutanix latest offerings. I’m sure there are customers that need 64+ node clusters, but they’re the 20% of the 80/20 rule.
So what I would love to see out of Nutanix is a solution where we get all the great simplicity of adding nodes and scaling up / down, but I’d really like to see the storage be something that’s totally external from compute. I’d like to see storage pools for all flash and hybrid tiers, and I’d like multiple (discrete) compute clusters to be able to access them. In essence, I would much rather see Nutanix release a traditional 3 tier architecture, but take all the work out of managing it. It’s not reinventing the wheel, it’s putting a better tire on it.
As a potential customer, I really do like a lot of what Nutanix is offering in its solution, but I can’t stand knowing that if my VM moves to a separate node that performance is going to tank for a bit, or that I need to plan how much storage capacity to include with my compute node. That actually makes my life harder not easier. It should be “oh, i’m running low on IOPS, I should add another storage controller” or “oh, i’m running low on capacity, I should add more storage”, or “oh, I’m running low on ram i should add another compute node”. Why mix something together and force it to scale together, when it ultimately scales better when its separate? i know one argument is put the storage closer to the CPU for latency reduction, but honestly that’s a problem for the 20% not the 80%. And with NVMe over ethernet or even the possibility of using something like infiniband for storage, there are many ways to reduce latency. Heck, even using something like the Pernix acquisition with caching would likely be good enough for 80% of that 20%.
Anyway, big fan of your blog for a while, just wanted to throw that out there. I’m not a VAR or a vendor FYI, just a guy stuck managing infrastructure and always looking for a better way.
Thanks for your comments Eric, it is greatly appreciated. There are quite a few changes to the Nutanix solution offerings that may be new since the last time your reviewed it. I apologise if I did not cover these all on this blog or that it wasn't obvious.
I think one of the biggest mistakes companies make going into cloud is that it is going to save money, the reality is it usually doesn't. Especially when they take the full impact into account. We have beaten public cloud on a economic basis several times due to many different factors that are customer specific. But you can get the economic model you want regardless of where you choose to run your infrastructure, be it public or private cloud. There are a number of vendors and partners allowing different economic models for on premises infrastructure that includes scaling up and down, within certain ranges, but you have to be dealing with the right partners and vendors for that (yes Nutanix has partners that do this with our solutions).
Now for some good news. Nutanix allows you to scale storage and compute independently and in different amounts. So you can get what you need when you need it, in the chunks you need it in. This isn't something new, it's been a capability for more than a year, possibly two years. My colleague Josh Odgers has written about the benefits of storage only nodes in his blog here – http://www.joshodgers.com/2016/07/14/scale-out-pe… which includes more than just capacity, but also performance as you get more controllers.
In an ESXi environment, scaling storage doesn't require any additional vSphere licenses either. You can mix and match the type of systems that meet the specific needs of your environment. There are still no luns to configure and no storage to manage really, just stick with your favourite way of managing your environment, hypervisor management console, prism (allows simple CRUD operations via vCenter), or api's, and away you go. Want to expand a type of resource, go ahead, it's a one click operation or just adding a storage only node or compute node of your choice. In the next software release that we've announced you can mix both hybrid and all flash systems in the same cluster as well to fit the different workloads and requirements you have.
You can also hook up external systems to Nutanix using iSCSI with the Nutanix acropolis block services, and this can be used from a wide variety of guest OS's, including AIX, Solaris, Windows and Linux. You can also use this for very high performance VM's that need to scale performance across many controllers, this is a 20% problem for sure, but the option is there. It also supports clustering solutions, such as Windows Failover Cluster. You need more controllers, you can just add more nodes. But you don't have to have a large cluster to benefit from this, it's available on clusters of any size through the Nutanix software. I mentioned the use of block services in this article here – http://longwhiteclouds.com/2016/12/31/nutanix-per…. But there is also a tech note on the Nutanix web site from our .Next conference last year showing two external Oracle RAC servers connecting into a cluster that had a mix of VDI and SQL databases on it. You can see all of this in action in the video we recorded last year –
. If you want to watch the keynote that explains it you can see this article, http://longwhiteclouds.com/2016/07/05/nutanix-nex… and on the video skip through to the 1 hour mark. The demo was done live and the desktop being used to do the demo was running on the same cluster.
I would encourage you to reach out to your local Nutanix team and if you don't know who that is I'd be happy to connect you. A lot has changed and I think we can solve all of the problems that you've raised and without having to have a traditional three tier deployment.