August 2011


With a string of recent distribution and collaboration announcements, it’s time to look at Cloud Foundry’s progress since a beta announcement in April 2012.

VMware harvests fruits of SpringSource Acquisition
EMC VMware acquired SpringSource a little over two years ago. At the time, I wrote that the hype surrounding the deal greatly overshadowed the real opportunities that SpringSource brought to VMware.

After re-reading my original analysis, I still stand by the post. Even today, while SpringSource technology underpins VMware products like vFabric and Cloud Foundry, neither could be viewed as helping to move VMware’s revenue needle in a noticeable fashion.

VMware CEO Paul Martiz confirmed this during VMware’s 2Q2011 earnings call:

…as well as we continue to invest in the Spring Framework and the combination of the Spring Framework with Cloud Foundry. But I think it would be fair to say we’re still plowing the ground there. And we expect those investments to pay off well over the longer term, but we’re still in the development phases of the market.

Considering the typical five year payback periods used to evaluate acquisitions, all I can think is that VMware has a busy three years ahead of itself to justify the nearly $420 million valuation VMware paid for SpringSource. That said, with SpringSource technology, VMware is in a significantly better position to become grow beyond a hypervisor vendor, into a platform vendor with the likes of IBM, Microsoft and Oracle.

Whether VMware can pull it off, is still to be seen.

VMware expands Cloud Foundry distribution channels
This week, VMware announced deals with Canonical, Dell and enStratus to significantly expand distribution channels for Cloud Foundry technology.

Of these, the Canonical deal appears to be most interesting. Cloud Foundry can benefit from Ubuntu’s leading share of Linux cloud and virtualization deployments.

In explaining the collaboration with Canonical, VMware staff wrote:

Now starting with the 11.10 release both the (Cloud Foundry) VMC Client, and VCAP server functionality will be available directly as Ubuntu packages created by Canonical. With over 20 million active desktop users and a strong IaaS server OS popularity it represents an important milestone for the open source distribution of Cloud Foundry, and is just the beginning of an ongoing collaboration with Canonical. Having the VMC client pre-installed and ready on millions of developer desktops makes a Cloud Foundry app deployment just a few commands away for anyone using Ubuntu.

Cloud Foundry interest expanding, but not yet a game changer
Against this backdrop of potential opportunity for Cloud Foundry adoption is the reality of usage and interest to date.

During VMware’s 2Q2011 earning release, VMware’s prepared comments highlighted 25,000 developers signing up for Cloud Foundry. That certainly is a respectable number of interested users in three months since the beta announcement. It will be interesting to watch this figure over time. It’s not uncommon for new products to gain interest when first announced, only to trail off in the long run.

It is interesting however that the various Cloud Foundry Git repositories on GitHub are “watched”, a proxy for interest level amongst GitHub users, by fewer than 800 users, while leading repositories count well over 5000 watchers.

Of note, interest in the Cloud Foundry project targeted at Java applications is less than 20 percent of the interest of the overall Cloud Foundry project.

Considering the revenue that Java attracts from enterprises, even in the face of languages such as Ruby, or PHP, Cloud Foundry’s growth into enterprise accounts could be less than a smooth one.

Looking at Google search trends at open source based platform as a service offerings, VMware Cloud Foundry, Red Hat OpenShift, Amazon Beanstalk, and CloudBees self titled platform, it’s clear that the market is still wide open, with each offering in the 15 to 25 percent range.

Add Google App Engine into the comparison, and interest in Google App Engine dwarfs the interest in Cloud Foundry and others by an order of magnitude.

I purposefully did not include the established platform vendors, IBM, Microsoft and Oracle in the comparison above. As much attention as Google App Engine has received, and offerings like Cloud Foundry are getting today, they’ve yet to crack the enterprise market in a meaningful way.

In conclusion, it appears that Cloud Foundry is making some good progress, but the road to enterprise acceptance, adoption and revenue is well ahead of it.

I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies, or opinions.”

Google announced Chromebooks  just three months ago to wildly positive and equally negative punditry. Evaluating recent product announcements and business growth for Chromebooks, it’s becoming increasingly clear that Google has a winner with Chromebooks. If you haven’t been following Chromebooks closely, you’d better start.

Chromebooks are a disruptive innovation
I have previously countered ZDnet’s Ed Bott’s claims that Chromebooks aren’t Windows killers. Here are of two key points I raised:

1. Google’s pricing strategy is a step toward IT as a service. By reducing the cost per notebook and business applications to approximately $35 per user per month, Google was reducing the total cost of ownership to less than 20 percent of today’s cost of acquiring, maintaining and supporting the IT infrastructure needed per knowledge worker.

2. All apps that some users need can run in a browser. Simply put, a Chromebook is not for every employee. However, a majority of knowledge workers, specialized workers or mobile users could use a Chromebook with little to no impact to their workflow.

Google continues to make Chromebooks enterprise ready
Google’s claims Chromebooks are designed to get better and faster over time through software updates.

Google recently announced the availability of new features that support their claim of Chromebooks “getting better over time”.

VPN support and Secure WiFi have been added to the latest Chrome OS release, which is the operating system software that runs inside a Chromebook.

With these two additions, businesses that protect access to their wireless network and restrict remote access to their internal network – that would be virtually every business I know of – can now consider a Chromebook in their enterprise.

It’s a little surprising that Chromebooks were targeted at businesses without support for WiFi security at a minimum. VPN support would be a close second in basic requirements for a business seeking to use Chromebooks with mobile employees.

That said, Google did close these two holes in three months since first shipping Chromebooks for businesses.

Google also announced a Tech Preview of Citrix Receiver for Chrome OS, which would address users who need to run existing applications not suited for a traditional browser. For instance, Google and Citrix show Adobe Photoshop on a Chromebook through Citrix Receiver.

I’m still convinced that this is a checkbox feature versus something Google truly expects broad adoption of. For instance, the Citrix Receiver Tech Preview currently counts 38 users on the Chrome Web Store.

Chromebook customer traction is encouraging
While the pace of feature additions to Chromebooks in encouraging, Google’s client references for Chromebooks are even more impressive.

Google groups client references into several categories that could resonate with potential business buyers.

IT departments have to contend with the challenge of supporting branch locations. As Google rightly points out, onsite IT support at some branch locations can be expensive and impractical. Google now counts the likes of AmericanAirlines, RubyTuesday and Jason’s deli as clients using Chromebooks to reduce the cost of IT at branch locations.

Another key target user group is specialized workers. Virtually every business has a set of users whose IT needs don’t expand beyond email, and access to intranet and web-based applications. These users are perfect trial groups for rolling out Chromebooks at your company.

Salesforce.com, Groupon, Logitech and InterContinental Hotels Groups are key clients using Chromebooks to meet the needs of specialized workers.

Finally, Virgin America, National Geographic, the City of Orlando, and Konica Minolta are amongst reference clients using Chromebooks for mobile employees. Again, I find it interesting that these organizations adopted Chromebooks for mobile users before VPN and secure WiFi capabilities were added to Chrome OS.

Your organization could very likely find mobile users, specialized users and branch office deployments that could benefit from Chromebook usage.

With the strong list of client references Google has already collected, what are you waiting for?

I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies, or opinions.”

VMware’s vSphere 5 new licensing model raised eyebrows and tempers across the industry just three weeks ago. Bowing to the negative response, VMware has announced three changes that should appease customers – but will still likely lead to higher virtualization costs for VMware customers.

Customer backlash with original vSphere 5 licensing model
When first announced, the new vSphere 5 licensing model was positioned as being a positive for customers. VMware didn’t however mention that each vSphere 5 CPU-based license came with a fixed amount of virtual RAM (vRAM) entitlements. If your configuration has more vRAM than is entitled for use with the CPU license of vSphere 5, you need additional licenses.

As I’d previously covered, VMware customers complained of 2x to 3x higher licensing costs for vSphere 5 versus vSphere 4.

VMware announced three changes that attempt to address concerns around vRAM as a licensing metric and the resulting increases in cost for a VMware customer. Users commenting on VMware’s forum seem to be less irate with the new changes but still question the vRAM model.

Increasing vRAM entitlements per license
First, the vRAM entitlement per vSphere edition has been increased from 24/24/24/32/48 to 32/32/32/64/96 gigabytes for vSphere Essentials, Essentials Plus, Standard, Enterprise and Enterprise Plus respectively.

Let’s evaluate the result of this modification on our previously discussed customer example. We have a customer with a two-socket processor with no more than 12 cores per socket, with 256GB of RAM.

Under the original vSphere 5 licensing model, the customer needed six CPU-based vSphere Enterprise Plus licenses — not two — to be entitled in order to use the full 256GB of RAM on the system.

Under the revised vSphere 5 licensing model, the customer would need three CPU-based vSphere Enterprise Plus licenses — not two.

The first two CPU-based licenses, sufficient for the two CPUs on the system, would have provided 96GB of vRAM entitlement each, for a total of 192GB.

However, since the scenario included 256GB of RAM, the customer would have had to buy a third vSphere CPU-based licenses in order to use the full 256GB of RAM.

Using the old vSphere 4 licensing model, the customer would only need to purchase 2 CPU-based vSphere Enterprise Plus licenses. As a result, a customer in this situation would still be paying 50 percent more than they did with vSphere 4, but at least its not 300 percent, as was the case with the original vSphere 5 licensing model.

Capping vRAM usage to encourage virtual machines with large amounts of vRAM
The second licensing change VMware announced is that the amount of vRAM counted against any one virtual machine is capped at 96GB. Using our example above, if your system has 256 GB of physical RAM and you want to allocate 128GB of vRAM to two virtual machines, VMware will only reduce your vRAM allocation from 256 to 192GB. As a result, you’d be 64GB under your 256 vRAM entitlement even while using a full 256 GB of vRAM across the two virtual machines.

VMware made reference to the fact that customers could use a 1TB vRAM virtual machine wile only reducing their vRAM allotment by 96GB from the total vRAM pool.

Customers would still need sufficient vSphere CPU-based licenses to use the 1TB, or whatever number, of physical RAM available on the system. As such, this change only applies to how much vRAM is used from within the vRAM pool. This is pertinent for additional vSphere CPU-based licenses that may be needed in following years if the customer surpasses the total pooled vRAM allotment in the previous year.

VMware aims for Tier 1 applications
With the third licensing change VMware now calculates a 12 month average of consumed vRAM rather than tracking the high water mark of vRAM used.

By limiting the vRAM counted per VM to 96GB and tracking a 12 month average of vRAM usage, it’s less likely that a customer will use more than their vRAM allotment per year.

The combination of the second an third licensing change make it much more attractive for customers to run multiple Tier 1 applications on VMware by reducing the licensing hurdles for doing so.

The risk however, especially after the current licensing fiasco, is how licensing may change in the future, after customers are heavily reliant on VMware for even their Tier 1 applications.

As we’ve discussed before, it’s good to evaluate options, especially as open source hypervisors become more mature.

As you move more business critical applications to VMware, consider moving some of your existing, less critical, VMware applications and environments to an open source hypervisor.  Using a mixture of VMware and an open source hypervisor, is likely your best long term option for balancing costs and flexibility.

I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies, or opinions.”