Nutanix looking for a way to burst VMware's bubble Rumours are circulating about a hypervisor built by Nutanix, and Nutanix open-sourcing its software; two intriguing moves. Why would Nutanix make these moves as VMware ratchets up its attack on hyper-converged vendors with EVO:RAIL? Start-up Nutanix is a hyperconverged infrastructure appliance (HCIA) supplier, widely-regarded …
One thing worth noting is that other hyeprconverged players are already working on Docker support. Maxta, for example, supports VMware, KVM, KVM/Openstack (Via Mirantis) and Docker. And that's just one of the many, many players in the hyperconverged market.
Nutanix isn't the only one talking about open source either, and there are at least two "open hyperconverged" projected that have started in stealth. Hyperconvergence is going to be a commodity just like the hypervisor. The money is in the management tools.
Make it easy to use and cheap to deploy with licensing that's simple and comprehensible...or get out of the game!
Is a one trick pony, great if your infrastructure is purely Linux and you need to quickly scale out specific functionality - So great for cloud, but actually relatively useless in your normal enterprise business environment.
I really hope your correct Trevor, perhaps finally there will be a decent alternative to vsphere, and something that will give VMWare a reality check. Think I'm going to hold off upgrading my VMWare licensing for a few months..
@K - Yes Docker tries to do one thing well, but that's its strength. It's not meant to solve every problem, but it solves a lot of common ones. Any complex IT system would consist of a mix of containers and traditional VMs, and indeed containers inside of traditional VMs.
As for what Nutanix is going to do, I agree it's more likely they are adding their own management tools on top of KVM than actually rolling their own hypervisor. The same may be true for Docker containers.
<selfpromotingrant>
Not sure if you were referring to us but we are, with NodeWeaver, one of those Open HCI coming out of stealth.
We haven't done much media noise yet but the presentation of the product at a recent "Cloud" event in the UK generated a lot of interest (and sales).
Nutanix, Scale Computing and Co are incredibly good products which are aiming to the Enterprise market where hundreds of nodes are the norm but what about the 95% of the European market which is made of SMEs?
Does every company in the UK need to spend £100K+ to setup an efficient hyper-converged/hybrid cloud infrastructure? Maybe not. Maybe a lot of businesses would like to start with a couple of nodes investing very little and then scale-out by adding appliances or converting existing VMWare servers into HC nodes as it takes 5 minutes.
Then does the hypervisor really matter? We've chosen KVM as we think is a lot more efficient than others but that doesn't stop you migrating existing VMWare, Xen, Hyper-V workloads with no or little modifications. If the "Cloud style" management of the infrastructure is so easy and you can easily integrate to other management suites through API why the customer should pretend to use VMWare as hypervisor?
Sorry about the self promoting rant but we think that there is a market out there that needs to know they can afford Enterprise class products that provide the same features found in Nutanix or EVO:RAIL and that those products can be in most cases more cost effective than any Cloud service.
</selfpromotingrant>
Disclamer: if you haven't noticed I'm personally involved in the promotion of NodeWeaver and many other Open Source & Linux based product so the views expressed are certainly conditioned by it.
Yes, PVecchi NodeWeaver is one of the projects I was talking about. There are others. I actually just learned of a new one a few hours ago. Don't ask me how I ended up being "hyperconverged guy", but here I am. :/
NodeWeaver is interesting, and I can't wait to see the code when it's ready for a more open beta. Keep an eye out for some of my articles this month, I've got entire series on hyperconvergence due.
The only thing that is still in beta of Nodeweaver is marketing but we are working on it. NodeWeaver, after a few years of development, testing and having been in production also in mission critical environments is a product that's ready for prime time. So prepare 2 or 3 servers in your home lab to start testing it ;-)
Well, that's a proper home lab :-)
Anyway you probably will be partially impressed if you test NodeWeaver as you are already using the excellent Scale Computing HC1000.
There are a just a few noticeable difference: NodeWeaver NW-110, which is aligned with the HC1000, includes 512Gb of SSD cache, starts doing its job nicely also with 2 nodes and 3 nodes cost just a bit more than a single HC1000. The rest of the features seem to be roughly the same. We are working mainly on the EU market so we haven't seen Scale Computing much yet but we are winning a lot against Nutanix and naturally VMWare.
"We are working mainly on the EU market so we haven't seen Scale Computing much yet"
Sorry to have to be the one to inform you, but you will be seeing a lot more of Scale in the EU market very soon. Hans and I will be seeing to that. Also expect Maxta, Yottabyte, Gridstore and many, many others.
Also: for the record, I have three HC1000s, so I do get to test what moving things around the cluster is like. I will be interested to give NodeWeaver a proper look see, just like I have for nearly every other hyperconvergence solution out there.
If you work with one hyperconvergence company you are biased towards that company. If you work with all the companies (or nearly all) you're developing expertise. I prefer the latter to the former, and I haven't' been disappointed. Each company has their charms, and each solution is special in it's own way, with it's own use cases.
So I'll find time later this month, and we'll see about getting Nodeweaver spun up!
"I'm not sure you ever heard a ScaleComputing pitch"
Uh, Hans, my company - eGeek Consulting Ltd - runs it's production infrastructure on Scale Computing's appliances. Jason sent over 3x HC1000 nodes and they've proven quite reliable and usable thus far. At some point The Other Scott Lowe and I are going to set up a multi-site replication setup for testing, and I think Justin Warren is on the list too.
Actually, you know I know about Scale because we had a whole discussion on Twitter about Scale. Anyways, Jason's really reaching out to the community at large and working with many of us to make sure we know what they have to provide. For the most part it's good stuff. (I should have a review up some time next week, as a matter of fact.)
Anywho, thar she be...
Looking forward to some more details on NodeWeaver.
Got to say, we looked at Scale computing enviroment to offer us a converged solution, the problem they have though is the Appliances they sell were just not up for the job. Even entry level was a QC CPU with 32GB RAM per node, and they could not be upgraded, which is such a waste and the only reason we went the other way (We did look at Nutanix, but they were simply too expensive).
On the NodeWeaver, please don't make this same mistake. Remember customers are buying these to last 3-5 years, hence they need to be able to add additional resources, without needed to constantly fill the rack with more nodes.
Entry level of scale is indeed a single socket Quad Core Xeon with 32GB of RAM that can't be upgraded...because Intel won't allow you to use more than 32GB of RAM with a single socket Xeon. Period.
But had you looked at one model up from the HC1000, you'd have found Scale Computing offered the dual socket models in a plethora of configurations that could be upgraded at will. You are complaining that the lowest possible tier model can't be upgraded because Intel designed their lineup that way. Sorry mate, there's nothing to be done about that. If you want the flexibility you have to pay the minimum cost for a two-socket server, even if it has one CPU. That ain't Scale's fault, and it would be the same if you rolled your own.
Trevor, I'm not complaining, just stating the fact.
Scale do have control over their pricing. We were offered this as a high performance, low-foot print yet cheaper alternative to a typical virtualised stack - Where for just a little more than the entry level Scale (using UK prices), we managed to secure 4 Dell R620's, each with Dual 6 Core Xeon's 64GB RAM, and a EMC VNX 5300 Hybrid SAN with FC/iSCSI. By the time you added resilient switches onto the Scale solution there was no saving and the performance from the VNX just leaves the Scale lack-lustre.
I understand Scales position, they are a smallish company, competing with global vendors, they don't have the margins to go toe-2-toe with Dell, HP and EMC etc on pricing. But from a business perspective, I have to deliver the biggest-bang-for-buck possible.
The other issue though was Scale had moved from IBM's GFS to DRBD (I think it was this), which whilst a capable system, is well known for being troublesome, getting it working reliably is voodoo, and since it was Scales first generation of HC3 we simple would not risk it.
I am not remotely sure that I understand. Scale is very easily better bang-for-buck than Dell, HP and EMC. Especially when you factor in the cost of the sysadmin time to set up and baby those devices. But then you switch from talking about TCO and start saying Scale is voodoo. So which was it that drove you away?
It sounds to me a bit like you took a cursory look, but it sounds to me like you either took a cursory look whilst hoping for a reson to discredit, or you were hoping for cheaper hardware than Dell/HP/EMC and were trying to compare hardware to hardware without factoring in the cost of any licensing on the Dell/HP/EMC side.
A 3 node HC1000 should be between 20k and 30k. That's between $6.7k and $10k per node for storage, compute, hypervisor, management tools and integration services. That's not the best price I've seen for 32GB/node, but going up a level to the dual CPU nodes that come with more RAM and can be upgraded doesn't bump the cost/node by much. In both cases they end up far cheaper than storage + compute + hypervisor + management tools + integration services from pretty much any place I've seen that demanded more than Synology for the back end storage.
Trevor, no need to get defensive :) your free to disagree. I never said Scale was Voodoo, please re-read what I wrote, you'll see I wrote reliably configuring DRBD is voodoo.. massive difference.
I do stress when we looked at Scale HC3, they had only just announced it, so configuration options were limited to the single CPU + 32GB RAM. But Scale UK prices would appear to be significantly more expensive, especially if you want 24/7/4hr response, this almost double the price of each node -
3xHC3 + Switching + 24/7/4hr support = £30k
4xDell R620 (£15k) + EMC VNX SSD/HDD (£20k) + 2x VMWare Essential Plus (£5k) = £40k
Note 4 Dell R620's..
So HC3 costs £10k per node, whilst doing it yourself costs £10k and you get double the performance - I'll put my hands up here and say the Dell and EMC kits is heavily discounted.
Can't speak to the UK. You guys get shafted on almost all tech. 1$ = 1£ is nonsense, but seems to be prevalent.
And while Scale may not have made huge sense when they first came out, I think revisiting - especially if you aren't in cuckoo 1$ = 1£ land - is worth consideration. But so are organizations like XByte and Servermonkey, especially if you are state-side. It's never simple; there are so many good combinations.
Scale's usefullness is that they are "easy button" simple for SMBs. And SMBs don't always have top teir sysadmins to help them do things. If you already have a top teir sysadmin, and his time is worth nothing, then rolling your own makes sense.
Up past the midmarket, does Scale make sense? I don't know. They don't have too many really high end configurations. They have some decent mid-range stuff now, but no high end stuff.
I live and breathe DIY. Despite this, I am really starting to come around to the whole idea of appliances. Companies don't make money resizing LUNs. So I am really falling in love with organisations that provide infrastructure appliances, so that I can stop fretting about keeping the lights on and start using the technology to hand.
But that's probably just because I'm old and cranky and no longer have the enthusiasm for this sort of thing. When you've fixed a thousand printers and debugged a debunked the iSCSI microbursting myth a few hundred times it all seems so...boring...to roll your own. And I question constantly if my time isn't better spend doing practically anything else.
Regarding "but they can't match or outflank the juggernaught that is VMware and VMware's partners' channel."
I reckon the VMware channel is ripe to support something else. VMware is doing a great job of screwing their channel partners. Margins are slim to zero and VMware will push their own Services in competition to the partner. I don't know a single VMware parter who is happy.
I'm just waiting for the music to stop so VMware can see there is no channel to support them.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
