Rumors in the press (CRN – Project Zephyr) have speculated that VMware is about to offer its own cloud has created an email thread among us analysts that we felt was worth sharing. The core issues discussed in the thread are 1) what is VMware going to do about the success of Amazon EC2, OpenStack and CloudStack, 2) what is the relationship between VMware’s success in the enterprise and potential success in a public cloud, and 3) what is the best way for companies to “on-ramp” into the public cloud. This lead to a discussion which started on the point of whether or not vSphere was scalable enough to be a platform for a successful public cloud computing offering.
The Current State of Public Cloud Computing
This all got started with a post by Bernd Harzog which compared the cloud strategies of Amazon, Microsoft, VMware, and Google. We all know that VMware keeps an ever watchful eye on Microsoft, and in “Microsoft’s Three-Pronged Windows Azure Strategy“, we speculated that Microsoft’s recent moves on the IaaS front with Azure might provoke a reaction from VMware. That post contained the table below. The Zephyr rumors are basically about VMware making its red “X” in the bottom row into a green “check”.
|Vendor||Amazon AWS||Microsoft Azure||VMware vSphere
|On premise private cloud|
|Cloud Partner (Service Provider) Program|
|Vendor Offered Public Cloud|
Analyst Comments (From Oldest to Newest)
I have to wonder if this is more about PaaS than IaaS? Sort of for CloudFoundry?I also have to wonder if this is really about EMC more than VMware. Remember, VMware has Mozy which I think is backed by Atmos (or could be). Both are EMC clouds. So is this EMC getting deeper into the Cloud business and not really VMware? Were they forced to do this?So CloudFoundry + Mozy + Zephyr…. It is the general direction they (EMC+VMware) have been heading…. Their pricing is THEM as a service provider given the VTAX… and Maritz did say publicly that they want to go more towards use base licensing which implies they need to own the cloud to do that effectively.
If Zephyr is more than CloudFoundry you can see from history that this has been coming for quite a while….
My thoughts are that the public cloud providers have to date largely voted with its dollars to build clouds on non-VMware software. Amazon, OpenStack, and CloudStack have essentially run away with the early market.
VMware has two huge problems. While it has a Service Provider Program (vCloud) those service providers have mostly played both sides of the street. They have put up VMware based clouds so that they can talk the talk and steal workloads from enterprise IT, but their real efforts have been on non-VMware clouds. VMware is also looking at Microsoft who is looking effective at all three rows below, and feeling exposed.
I think some combination of VMware and EMC is going to have to do something significant in the Vendor Offered IaaS space. VMware HAS to make the red X in its third row into a green check somehow.
The problem that VMware has is the vSphere is not architected to be a platform for a scale out public cloud. It is architected to provide backwards compatibility and virtualization to enterprise workloads. So VMware needs a new IaaS software platform that is not vSphere. Doing this in a way that results in both an effective competitor to Amazon, and that maintains some compatibility with vSphere is going to be a real challenge.
I have to disagree with vSphere not scalable, it is and especially in the Cloud, VMware have put a lot of work into this space, their auto-deploy although nowhere near perfect is a good example of this, Their Cloud management tools are by far better than any of the competitors. and the ease of instantiating a VMware based cloud is an order of magnitude easier than any of the competitors I have personally tried. a vCloud can be instantiated and be delivering service within a day, where for example Nimbula would take over a week. Yes I know Nimbula has a very funky self seeding initiation program to delivery a three host quorum and expand thereafter. but the process to deliver service is laborious and manual, to the point of having to manually create network connections and firewall rules, there are little preconfigured deployment examples. and when I saw the Nimbula breifing and their SE tried to build one out, even he failed two times out of three. Cloud and OpenStack have the same issues.
From my point of view the only issue VMware currently have is their pricing model. they are too expensive for a startup MSP to move to Cloud. that said I believe that this is going to change as they will be moving to a per VM pricing model with the MSPs.
Current iterations of cloud are flawed due to approaching business from the wrong angle, the only useful services are things like Dropbox, (joke) point solutions like Mail (Gmail, Outlook.com and hosted Exchange) or expensive CRM based solutions like Salesforce and HR/Payroll applications are the only places that real inroads are being made and these services have been around since the ASP days.
Currently only VMware have a mature view on Cloud from an Enterprise migration point of view. Business Transformation is a very long road. Enterprise needs to be hand held in baby steps, that is why the first stage is VDI/Daas to move the Desktop to the Cloud. once their true Transformation can happen.
The first stage of business transformation should *never* be VDI/Daas- that’s why many end up thinking “the cloud” is sh!t. Tom, just out of interest – how many how competitors have you personally tried, and how many of those do you know as well as VMware?
I don’t yet think IaaS or PaaS have much to do with Enterprise production apps, more of an Internet company or test/dev or ISV market. PaaS in particular tends to be about newer applications some (but not all) born in the cloud. And I certainly don’t see the DaaS thing as driving the IaaS or PaaS markeplaces. To be honest I’m struggling with DaaS as a long-term play at all – it’s like VDI – someone tries to reinvent Citrix with something that isn’t as good.
There are, in my view, a couple of things going on here. First, everyone is concerned about the lead that AWS has in public IaaS and nobody really know how this is going to end. Second the Platform-specific PaaS (i.e. one where you build your application to the Platform) is struggling so both Microsoft and Google are opening up the underlying IaaS layer inside their PaaS so that you can deploy non-platform-specific applications on it, thereby increasing the potential market for the product, and having a go at AWS on pricing. Platform-specific PaaS like SalesForce Force is receding into a marginal or tactical or legacy solution. Univeral Platform PaaS (like CloudFoundy and OpenShift and Salesforce Heroku) is where PaaS is going and potentially has legs, but at the moment I would say the jury is still out.
I think there is some kind of separation happening between datacenter and IaaS cloud, and a lot of this has to do with pricing, or more specifically about how the available margin is distributed amongst the system component vendors. I think VMware will need something other than vSphere to sit underneath CloudFoundry, not just for itself but also for the others who are building on CloudFoundry, at the moment everyone is using AWS or OpenStack. If PaaS happens VMware could dominate the PaaS market with CloudFoundry – it is favourite at the moment – but it needs to be able to make the right decisions for that platform and not cannibalize the vSphere cash cow.
By the way I don’t think you can can get Azure as a private cloud? I think you can get a bunch of technologies bundled together as the Microsoft Private Cloud but it isn’t Azure. I think you can get Azure as a public cloud or if you are a service provider you can get it as an appliance or (as of a couple of weeks ago) in bundled form. Also Azure is PaaS for .NET and IaaS for vanilla Windows and Linux, and the only bit the hosting providers get access to is the IaaS (and some basic stuff around provisioning websites). It is clear Microsoft has its own pricing issues around Azure.
DaaS is not a long term play, but a half way house to get people moving to the next paradigm
I think we have failed to address the most important question. That question is, if VMware chooses to build its own cloud in its own data center in Las Vegas or someplace else based upon vSphere, will VMware be able to field an offering competitive with Azure, Amazon EC2 and the forthcoming Google Compute Engine cloud? My opinion is that since VMware started with enterprise virtualization as the design point for vSphere, and the other three started with multi-tenant public clouds as the design point that the answer is no – and that VMware will have to come up with a different IaaS offering to compete in this space. There is also the possibility (also hinted at in the press) that the recent organizational moves at VMware have something to do with VMware and EMC collectively positioning themselves to compete with maximum effectiveness in both the enterprise virtualization (Software Defined Data Center) and public cloud realms.
Bernd, I understand your question, and I think unfortunately for VMware/EMC the answer lies in using commodity storage to drive down price points. But I’m not sure your question is all that important. If you are an EMC shareholder, what matters to you is not whether VMware can be #5 or #6 in a public IaaS marketplace dominated by AWS, but whether CloudFoundry can be #1 in a the PaaS market in 5 years time. Dominating the market will drive up the strategic value of CloudFoundry and this is also the reason why there there are split rumours – a different corporate structure would likely generate more shareholder value.
The fact is that vSphere itself scales quite well. You can have 2 nodes well over 1000 nodes and it just works. Clustering is a bit different as you can only have 32 nodes per cluster but you can have a very large number of clusters. The breakup is merging vCloud with vSphere. vCloud presents resource pools to tenants for their use. With only 32 nodes to a cluster, you have to create sufficient clusters that make up the various levels of vClouds. Or ~32 clusters, that are tiered over 1000 nodes. However if you buy REALLY large nodes, then the resource pools go much higher.
Perhaps they will announce larger clusters. The technical capabilities of vSphere far exceed any other hypervisor, with a price however that also matches those abilities. Pop on top of that vCloud and you have another set of issues.
The real problem we have with clouds today is NOT what they are built upon, but more how the Tenants can manage their own security, tenancy, catalogs, etc. In many ways, vSphere is unimportant in the cloud. The real item to discuss is vCloud and how it presents workloads, etc. So far it is far easier to implement that any of the other clouds I tried to implement.
But let’s look at what a VMware/EMC Cloud would bring to the table:
- Ability to move legacy Enterprise workloads to the cloud via vCloud Connector
- Ability to use Enterprise Class Hypervisors with all the bells and whistles
- Ability to scale applications to Big Data ranges fairly easily
- Cloud based Backup
- Ability to Design Applications for the Future — CloudFoundry
- Really pushes the Hybrid Cloud concept
- Integrated Desktops as needed
- Ability to Orchestrate one-off and continual workloads
- Ability to Charge back within your tenancy
- Built in Security
- Performance Management of your Tenancy
Hopefully an integrated solution from top to bottom.
All in all it is compelling if they can price it properly.
“Perhaps they will announce larger clusters. The technical capabilities of vSphere far exceed any other hypervisor, with a price however that also matches those abilities. Pop on top of that vCloud and you have another set of issues.”
That seems to make a lot of sense to me. I had this thought in my head that VMware was looking for bigger fish and had thought they might be heading towards ISP as a focus but building a public cloud would be considered along the lines of bigger fish. I have to wonder about the risk and reward. If they build it would this be a crossing boundaries with some of their partners and non-compete partner status? Clusters have had the ability to get bigger with each of the releases so I wonder how big a cluster can be with vSphere 6 when it comes out and to see how this comes along over releases to get to the ability of running very large clusters. Speaking of that, what number, how many, do we put on a larger cluster?