Bringing Cloud to Enterprise IT
Posted by Dan Chu
Vice President, Emerging Products and Markets
All clouds are not created equal. Google recently posted a blog entitled “What we talk about when we talk about cloud computing” which outlines what Google perceives as advantages to its approach to cloud. While VMware agrees that a major part of future enterprise architectures will reside in the cloud, we differ greatly on our approach.
To save
everyone the time and energy, the summary of their post is essentially that
Google uses cheap hardware that they expect to fail and smart software to build
the equivalent of a giant computer, that Google AppEngine can deliver the cloud
for traditional IT, and that the Google model can produce the fastest
innovation for end customers.
Google has a
valid and interesting model but we are finding that it simply doesn’t work for the
vast of majority of business IT. Let’s
address the issues individually:
1. Building the Giant Computer
Google says
that they take “a large set of low cost commodity systems and [tie] them
together into one large supercomputer” with their software. Fundamentally, we agree with this approach as
we have done exactly this for thousands of customers over the last 10
years. Using virtualization, VMware’s
solutions have become the de-facto standard for customers looking to improve
the efficiency of their datacenters while saving costs.
Over
130,000 customers run
While we
are aligned on the overall direction, the Google blog claims that their scale and
approach of managing servers lends a key advantage. If companies had unlimited resources and were
able to build massive datacenters with all of these commodity servers, the
Google model may be the way to go. However,
this isn’t the picture of most datacenters today. What virtualization is able to provide is
improved performance of applications, improved utilization of existing
resources and nearly unlimited scalability. And
The
2. Leveraging an Enterprise-class Cloud
Google
follows up by promoting its AppEngine stack as the way to deliver capacity and
scaling for applications and databases, “to deliver the set of scalable services
that customers would otherwise have to maintain themselves in a virtualization
model.”
This sounds
good, until you run into the issue of trying to run your core applications on
AppEngine. Customers are looking to
match their IT platform to their business needs, not the inverse. The Google approach calls for a least common
denominator set of non integrated cloud services that everyone squeezes into. Customers
want the flexibility and breadth of solutions that exist today along with the
efficiency of the cloud. Customers are
not about to re-write or modify their applications so that they can run in a
specific cloud. In particular, given
current macro-economic circumstances, customers have a high priority for a
cloud platform that can take their existing apps, and enable them to take
advantage of the cloud.
Finally,
customers want advanced business continuity, availability, and management
capabilities for production, enabled by such technologies as VMotion, High Availability, Fault Tolerance, Storage VMotion,
and SRM. Today customers are broadly using these
capabilities and
The Google
blog further suggests that “there is limited value to running an Exchange
Server in a virtual machine” and that customers should just use Gmail. Enterprises aren’t going to move off an
enterprise class mail platform for a personal-use platform. To take Exchange as an example, it represents
the kind of business-critical core IT application that is what customers want
running on
3. Innovating with Cloud
The Google
blog closes by asserting that “IT systems are typically slow to evolve,” and
that Google is much faster to innovate.
This supposition is mostly focused on Google Apps and its pace for new
feature rollout. This is fine for
customers who are looking for exactly the features that Google happens to be
working on, but for any other IT needs that a customer might have, the Google
stack is a black box to the customer without the component architecture that
lets many different types of partners integrate and contribute their new
technologies. In contrast,
Summary
For
customers looking to maintain the flexibility to move back and forth between
the external cloud and internal IT, Google’s proprietary platform is like the
“Hotel California or the roach motel” where your apps go in, but they never
come out.
VMware provides choice—allowing customers to put new and existing
applications where they want, when they want.
So in a nutshell, if you’re a business of pretty much any size and want
to consume cloud services, there are options that allow you more flexibility,
reliability, compatibility and mobility.
And if
you’re already a
Dan Chu is vice president of Emerging Products and Markets for
This is a great post and brings some sanity back from Google's claims, but there are some points missing.
First I have to agree, for companies that have already made a capital investment in infrastructure, using VMWare to create a private, internal cloud is the way to go to maximize flexibility, elasticity and utilization. But there is one thing the public cloud providers have that ultimately will win the race, multi-tenancy. The public cloud providers have the ability to house multiple companies/accounts on the same hardware to maximize utilization. This ultimately will lend to higher utilization and a lower cost/CPU cycle. It's very difficult for a company to achieve the same levels of utilization/efficiency independently, particularly if IT is not their core business.
There are privacy and security issues that accompany multi-tenancy but ultimately they will be worked out. There are bright minds working on them and at the end of the day, it just makes sense. This is why power plants exist and every building/house doesn't run it's own generator anymore.
On the point of flexibility, Google is a bit behind but Amazon's EC2 supports a myriad of platforms, including Windows, Solaris and Linux. To remain competitive, I imagine, ultimately, Google will need to as well.
In the near term, private and hybrid clouds will take hold. But on the 5-10 year horizon, it's seems that the promise of the public cloud's economies of scale will ultimately triumph (for all but the most private applications).
Posted by: J Schwan | May 13, 2009 at 06:22 PM
The Google approach (and to some extent Amazon's) assumes that the main design goal for the application is scalability.
Obviously, this does not makes sense, as scalability is the #1 problem for only very few organizations like Google, Microsoft etc.
For most vendors and IT departments the big obstacle is time to market and development power.
While Google relies on using Stanford PHD's to write JavaScript, most normal organizations can't rely on that.
I makes much more sense to take developers who know SQL and use tools that allow fast development rather than moving to completely new platform that can scale to 100 million users, which might not be your target.
If we look historically VMWARE has been very successful because it did not ask people to rewrite code, although it would probably be better from theoretical computer science point of view. Doing it without change , giving 95% of the value makes much more sense, IMO.
Posted by: Ophir Kra-Oz | May 23, 2009 at 11:33 AM