VMware’s Cloud Push
Even though many think I am an open source wonk, I have always argued in this blog and elsewhere that open source is more about choice than about source code availability (an argument many others in the cloud industry seems to have picked up in the recent days). I have always argued that the users should be presented with wide variety of choices including proprietary and open source solutions. In fact, this thinking is one of the reasons I have used the heterogeneity argument in my definition of federated cloud ecosystems. I see heterogeneity in cloud infrastructure platforms as an antidote to any monopoly that could develop in the market.
Another important characteristic is the presence of heterogeneous platforms in the ecosystem. A federated cloud ecosystem can be realized just with one cloud platform as long as the platform supports federation but it will just shift the monopoly from the service providers to the platform. Monopoly at any layer is bad for the users and this characteristic ensures that we avoid monopoly at the platform layer. I don’t support monopoly even if the underlying platform is open source. The idea of heterogeneity not only helps avoid any monopoly, it is critical to handle the needs of long tail users. More importantly, as I will explain in a future post or research report, it will help cloud customers avoid the so called cascading failures (Cloud is a complex adaptive system and heterogeneity is critical to avoid cascading failures in such systems) in IT.
In this context, vCloud is an important player in the federated cloud ecosystem I aspire to see in the infrastructure market eventually.
After the perceived failure of vCloud Express, VMware switched gears and started pushing hard with vCloud Director. Initially, their push on the service providers side appeared to be anaemic but they seems to have picked up pace in the last year. According to VMware, there are now 94 clouds in 19 countries that are powered by vCloud and their service provider business grew by 200% in 2001. Though not great compared to their virtualization marketshare, these are decent numbers taking into account cloud hosting market is still at its infancy with enough room for growth. A solution like vCloud Integration Manager is critical for further push in this market.
In fact, I want to see more cloud infrastructure providers than less (in some form of consolidation). I feel that more choices will empower the customers and also cater to the diverse needs (both in the technological and regulatory sense) of users from different parts of the world. This is one of the reasons why I am advocating the idea of federated clouds in this forum and elsewhere. If you hear pundits claiming consolidation and handful of cloud players offering compute services like what is happening on the electricity side, feel free to ignore them because they are just being shortsighted. It isn’t going to happen anytime. We are going to see a surge in the cloud infrastructure providers in the coming two years and vCloud is going to be one of the serious contenders in the space.
But
A year or so back, I had a chance to talk to CEOs of two of the service providers who are transitioning from the traditional hosting to cloud hosting. At that time, they were relying on VMware’s vCloud initiative to offer their cloud offerings. When I pressed them about the economic viability of using VMware technology to compete with the commodity cloud providers like Amazon and Rackspace, they both agreed that it is going to be tough. There is no way they can compete with service providers using open source infrastructure platforms on commodity servers when they shell out huge licensing fees to VMware. Eventually, they have moved into offering tiered cloud services with open source technologies driving their commodity cloud offering and VMware technology staying beneath their enterprise offerings. Now, with OpenStack marching ahead in full force, we are going to see more and more service providers adopting OpenStack to offer cloud services at commodity prices.
Some pundits right away dismiss the idea of enterprise cloud as DOA but I disagree with them. I do agree that we will see modern applications being built for cloud architecture (with design for failure and built in performance optimizations (to perform better on commodity hardware)) in the future. However, it doesn’t mean that there are no role for enterprise clouds, at least for the next 5 years or so. First, organizations are not going to shelve their legacy applications but more and more of them are going to push these applications to run on public clouds. However, these applications are not suited for commodity clouds like Amazon EC2. Enterprise clouds will have a role in catering to the needs of such applications. Though legacy applications may become historical footnote sometime in the future, it is not going away anytime soon. Second, not every organization want or need modern applications running on commodity clouds. Some of them may stick with legacy architecture for business or legal reasons. Remember, we still have organizations running mainframe after so many decades. Enterprise grade clouds will cater to these use cases as well. Third, some organizations might build their modern applications on “premium clouds” just because they can (much like how some people want to drive Hummers when they can as well do the commute using public transport or a low end car). Enterprise clouds will come handy for these kind of use cases too. In all the above mentioned use cases, VMware will be a strong contender.
However, it is my opinion that VMware’s cloud platform will be catering to a niche user base in the long run while commodity clouds based on open source software and cheaper hardware, will be powering most of the infrastructure needs of the world. In the long term, it is going to be tough for VMware to sustain their dominance in the infrastructure market. If CloudFoundry (previous CloudAve coverage) takes off as expected, VMware can tap into it to stay relevant in the future infrastructure market. How successful they will be in taking advantage of CloudFoundry to sustain their infrastructure side of the business is difficult to predict at this stage. It is going to be an interesting five years for VMware’s infrastructure division.
Related articles
- VMware makes it easier to wholesale clouds (infoworld.com)
- VMware debuts cloud integration manager for service providers (zdnet.com)
- VMware crafts mega-controller for public clouds (go.theregister.com)
Can you elaborate on the tie-in between CloudFoundry and VMware’s future in the infrastructure market? I didn’t think CloudFoundry had any dependence on the VMware stack. Thanks.
It doesn’t have any dependence but VMware can sell end to end PaaS solution with their infrastructure underneath. If they succeed in that, it could help them sustain into the commodity cloud dominated future.
Krish,
I run the vCloud team at VMware that, among other things, prices and packages our software for service providers. We’re also responsible for vCloud Integration Manager.
A generally thoughtful post, except when it comes to your economic argument: “There is no way they can compete with service providers using open source infrastructure platforms on commodity servers when they shell out huge licensing fees to VMware”
There are two things wrong with this statement:
a) The assertion that open source is less costly because there are no license fees
b) The assertion that “huge” VMware licensing fees make or break a service providers’ business case & economic model
Here’s the deal about delivering cloud service: however you achieve it, you need to rent the services of some extremely capable software engineers — lots of computer science PhDs, in other words. There are three main ways that cloud service providers do this today:
1) Rent the services of a lot of computer science PhDs by directly employing them. This is the approach of service providers who choose to build their own cloud stack (often incorporating open source software). Hiring and retaining these people is difficult and expensive, which is why only a handful of providers have done this, notably including Amazon Web Services.
2) Get other people to pay for the computer science PhDs, and rent their work for free through an open source project. This is the OpenStack approach. There is an additional cost, though: to make the cloud stack operational and functional, they must also rent a set of smart engineers through direct employment to fix up, modify, integrate and customize the open source code. This is still difficult and expensive, though not as hard as employing everyone you need directly.
3) They rent the computer science PhDs of firms like VMware, by licensing software. Almost everyone except AWS does this, including Rackspace who licenses both XenServer from Citrix and vSphere from VMware.
The net? You need to rent the expertise, whether you pay directly in salaries or license software. Service providers choice the blend according to their business strategy. There are no short-cuts. Direct employment requires substantial investment in salaries well ahead of actual revenue, so also suffers from the cost-of-money burden in any financial calculation. Also, you need to budget for ongoing maintenance of any code or configs you create, so there’s a cost long after the code has been written. If you want support of open source code, you need to budget for that too. That’s what all the aspiring OpenStack start-ups are banking on: they they will be the ones collecting those maintenance fees. Licensing (at least, if you use VMware’s service provider program) has no up-front investment — you pay in arrears, based on actual usage. Maintenance and support is included in the price.
On to the second assertion: that VMware licensing makes or breaks the economic model for SPs with “huge” licensing fees. VMware’s SP licensing is ratable, paid in arrears based on the amount of virtual RAM used per month, and works out to be approx 5-10% of the raw VM service revenue (not including any value-add). In the economic models we have validated with our 6000+ service provider program membership, the top three costs are:
1) Hardware — servers, storage, switches etc.
2) Operational costs — administrators, engineers, management overhead
3) Sales, marketing, general & administrative. Sales forces are expensive — whether they’re yours, or belong to a channel partner.
Put another way, you could have an EC2 based on VMware for pricing that is 5-10% higher, and get all the additional benefits that VMware provides in its cloud stack — like storage and network I/O controls to stop noisy neighbors, persistent VM file systems, HA etc. etc. For comparison, EC2 customers today are happy to pay >100% more for a dedicated instance that doesn’t suffer from noisy neighbors on the same host.
Regards,
Mathew
Mathew,
Thanks for your detailed response. Here is my reply.
I never asserted that it is less expensive because open source is free. I have only argued that open source software coupled with commodity servers beats the cost of VMware license coupled with the relatively premium hardware needed to run. I have tested the economics with a few service providers who explored both the options and they were pretty clear that they cannot compete with commodity clouds operating at the scale like AWS if they relied on VMW’s pricing. If you have a cost comparison done by an independent person to share, please feel free to share. I have been involved with open source software for more than 10 years and I will never make a simplistic assertion that open source is cheap because it is free.
I do agree that people are willing to pay a premium for better I/O, HA, etc.. In fact, it is exactly the reason why I am disagreeing with fellow cloud pundits that enterprise cloud is DOA. However, you are talking about what is happening in a snapshot of time but I am taking a more long term view and speculating where things are headed. On this issue, at least, it is the case of your speculation vs my speculation. We can only settle this after seeing where the market is headed in 5 – 10 years.
Thanks for the clarification. You can run VMware on commodity hardware. We have plenty of SPs in the VMware program who do exactly that today: white box servers, commodity NFS storage, white box networking gear. I won’t get into the pros and cons of this approach, but to assert that it isn’t possible is false.
Well, I am open to talking to these service providers directly and finding out how they are doing. My opinions are formed based on what I observe in the industry. If I do see service providers using vCloud software on commodity hardware and compete with the likes of AWS, I will definitely change my opinion. If you can put me in touch with these service providers for a direct briefing, I will be more than willing to talk.
I’ve been in this industry a solid 15 years now, and I have to agree with you Krishnan. The pricing is brutal at this stage. A lot of VMware’s products are modeled around when your office suite cost ya $400 bucks, it came in a box, and you installed it on Windows only…
So needless to say with the comodity pricing of AWS, Rackspace, and others, there will have to be some rebalancing after this disruptive change. 🙂
More data points: http://it20.info/2012/03/the-cost-of-doing-public-cloud-with-vmware/