Go hyperconverged. Or cloud. Or whatever. You won't save on hardware

Whatever choice of hardware you make – on-premises, cloud, roll-your-own, converged or hyperconverged – don't expect to save any money. So says Keith Townsend, a consultant, speaker and writer who El Reg saw in action in Sydney, Australia, at the local VMware user group conference. At the event Townsend gave a talk titled “ …

[FlBe]

He is right ...

With HyperConerged Appliances you always overspent in the nodes because you can't scale resources (compute/storage) independently. You pay for resources that you are not using and that rises with every appliance you add.

I'm also wondering why nearly nobody is talking about security concerns, because you can't realize a defense-in-depth concept with control and data plane sharing everything. But HC looks that easy on powerpoint and youtube ... maybe because there are also some redundancies missing that are best practice in traditional IT environments.

[sempertyrannis]

Re: He is right ...

Not quite - most HCI vendors offer storage only nodes, compute nodes, and a mix of the two allowing you to scale cpu & ram and storage independently based on specific need at the time. The notion that you cant is a FUD point from 2012 and has long since been addressed by both Scale Computing and Nutanix.

[Pascal Monett]

So you don't save money

I think we all knew that, but it's nice to have an official confirmation. Someone needs to kick this info upstairs, because there's a lot of upper management types who are going to be quite disappointed for their bonuses.

[big_D]

Re: So you don't save money

We looked at a cloud telephone system to replace our aging Siemens HiPath. We also looked at a locally hosted VOIP system. Given that we had spare capacity on our VM servers, the cloud version would have cost more for the initial set-up that we paid for the local system (which includes perpetual licences and 3 years support and updates) and the monthly costs were around three times the cost of staying with Telekom.

[Anonymous Coward]

Re: So you don't save money

"We looked at a cloud telephone system to replace our aging Siemens HiPath. We also looked at a locally hosted VOIP system."

Assuming you use Outlook/Exchange and Skype for Business, MS has made full VoIP part of the Skype bundle. It's in there.

[allthecoolshortnamesweretaken]

Re: So you don't save money

Depends. Your company won't, all things considered and truthfully added up - but your department/division/whatever might. In large organisations, shifting cost (and blame) so it becomes SEP is always a goal.

[Anonymous Coward]

Re: So you don't save money

Well according to this article, he doesn't really address the elephant in the room. No one, outside of VMware/EMC/VCE HQ, thinks that "cloud" means VBlock or Rail vs Nutanix or anything of the like.... Cloud, in the real world = AWS and Azure.

I would love to see VMware and their EMC overlords try to come up with a picture where VBlock is less expensive than the cloud offerings. No chance. I would also take them up on the idea that you cannot roll your own for less than VBlock.... ANYTHING you do will be less costly than VBlock.

[ntevanza]

i find your lack of faith disturbing

Now you have one throat to choke. And when the admiral is dead, what then?

[Nate Amsden]

one throat to choke

You don't need hyper converged to get that obviously, you don't even need converged to get that. I have mostly one throat to choke(HP) with servers(rack, not blade), storage(3PAR), fibre channel storage networking(Qlogic & Brocade HP OEM), and I get VMware support through HP. Only thing that is not HP is ethernet networking(switches/load balancers/firewalls all from different vendors, none of them being Cisco or Juniper if you were curious).

And in case it wasn't obvious, none of this is "converged".

It's pretty rare that I need support that spans more than one subsystem though.

[Down not across]

Re: one throat to choke

Only thing that is not HP is ethernet networking(switches/load balancers/firewalls all from different vendors, none of them being Cisco or Juniper if you were curious).

Only curious that with all that HP (or HP supported) kit, you don't have any of their switches as I've found HP's ProCurve to be quite good.

[Nate Amsden]

Re: one throat to choke

I have been an Extreme Networks customer for about 15 years now, I like their gear, specifically the ease of use of the CLI (it is not Cisco like, you can google "techopsguys simple network management" for a blog post of mine from years ago), also ESRP for layer 2 loop prevention and layer 3 redundancy in the switching layer (again google search "techopsguys esrp" if you want details behind that).

[Anonymous Coward]

Re: one throat to choke

Right, this whole VBlock or converged story is just a re-hashing of the IBM mainframe value proposition. Not only is this not a "new" story, it is the original IT story. Only with IBM mainframe is really was a converged system, meaning that the same company and engineering teams built it from the processors on up... so not as good as mainframe with the cobble together approach of various OEMs in VCE or the like. The problem is the same though... cost. It is ironic that companies like EMC and Oracle who originally came to prominence arguing against the IBM model and for a distributed "open" model are now arguing that you should put humpty dumpty back together again into a make-shift mainframe. Turns out that managing a massively distributed environment from a bunch of vendors can be costly and laborious to manage... if only we could have a single frame which is our main IT system. What should we call it?

[pichardo0]

I disagree...

There are ways to save money in a HC solution. Dell/HP/Nutanix tend to markup the cost of DRAM/Storage, which consists up to 75% of the cost of a server. The other cost is the software and support.

You might not have the "one throat to choke" but when does DRAM/SSD's typically fail?

[Androgynous Cow Herd]

when does DRAM/SSD's typically fail?

1:00 AM Sunday morning when I really don't want to go to the datacenter.

[sempertyrannis]

Office of the CTO

You forgot to include Scale's HC3 in your calculations. Our average SMB to Midmarket customer saves ~40% vs the legacy servers+switches+SAN+hypervisor they moved away from when they implemented their HC3 cluster.

[Anonymous Coward]

Where is the cloud discussion... or is this the VMware definition of the cloud (i.e. buy more VMware licenses)? I would like to see how the VMware vision of the world comes out against AWS or Azure. Probably not a pretty picture for VMware and their "consultants".... Lets just try to pretend that those are not options.

On the HANA cost point, no s***, but overpaying (massively) in one category like DB doesn't justify overpaying massively, just not quite as massively, in another category, like VBlock. They are both plenty stupid. Some of the prices I have seen people pay for VBlocks for HANA just beggar belief. You wouldn't have an ROI on the project over a 100 year timeframe and you could get 95% of the benefits which meets all the requirements of the use cases at 10% the cost.