Which Private Cloud is Best and How to Select One

Miha AhronovitzApril 25th, 2012Last Updated: October 21st, 2012

1 30 4 minutes read

This litmus test is proposed to compare private clouds

How long does it take to place in production an application delivered as service in your private cloud? (comparing apples to apples)? Less than 1 hour? Less than 1 day? Less than 1 week? More than 1 week?
What is the skill level required for (1)? . Rate 1 any user, 2 any sysadmin, no training, 3 only trained computer science sysadmins
Does it have a ready to use billing system to be used internally and externally? Most reply ” it has “hooks” to external “unnamed billing systems”. The reply is either Yes or Not.
How the server scalability works? Manual or Automatic? Where the additional servers are located? (a) More servers on site or other sites inside the same organization are added function of aggregated demand? Or (b) servers are added from public sites for additional costswheneverthey are needed? If (b) , how outside bills are allocated to internal and external users?

Now read on to see why.

OpenStack vs Eucalyptus vs OpenNebula is an animated discussion on linked in. Here is my take.

Don’t compete on features

This discussion assumes that the winner will determined from technical features. This is wrong of course. Experience shows, the executives who back the product, – Eucalyptus has Marten Mickos – and who know all the right people – will win.

OpenStack is yet to produce a startup backed by some big names well connected. If you think this is not important, read the blog from Andreessen Horowitz http://bit.ly/ww37ZZ. You will see how Opsware was transformed from one product with a single customer, and full of holes bugs, into something that HP bought for $1.6B

Flaunting product features to win the war with competitors, is a mistake, because no one knows the winning features anyway. Martin Mickos tweeted a quote; ” Remember that not getting what you want is sometimes a wonderful stroke of luck.”
I had a look at Andrew Chen blog Don’t compete on features http://bit.ly/xu9iZn He says:
There are three key ramifications for teams building the first version of a product.

Don’t compete on features. Find an interesting way to position yourself differently – not better, just differently – than your competitors and build a small featureset that addresses that use case well.
If your product initially doesn’t find a fit in the market (as is common), don’t react by adding additional new features to “fix” the problem. That rarely works.
Make sure your product reflects the market positioning. If your product is called the Ultimate Driving Machine,…, bring that positioning into the core of your product so that it’s immediately obvious to anyone using it…. Your product will be fundamentally differentiated from the start.

I was the product manager of Sun Grid Engine for a decade and the most frequent request I had was to produce comparison with competing products LSF, PBSpro, and so on. Each time such a document was produced, it was leaked to competitors, they immediately added (or claimed they added) the features we claimed as exclusive. Some ofthefeatures were so esoteric (see A glimpse into the new features of Sun Grid Engine 6.2 Update 5, due in December 2009) that you can count the users who demanded them on your fingers. The vast majority of users did not need them

Private Clouds versus wishful thinking Private Clouds

Tom Morse has an excellent web site where he lists most private cloud offerings which are claimed to be products. http://www.cloudcomputeinfo.com/private-clouds It is a very nice work. Here are the companies he lists :

However this list completeness pays as price the inclusion of wishful thinking companies who they believe they are a private cloud, like IMO Cisco. Cisco under December 2011 claimed they integrated 3rd parties cloud software in their solutions, creating complicatedlabyrinthine implementations. On February 1, in Padmasree Warrior, Cisco’s CTO claimed in Cisco Live Europe event that

…Cisco also has plans to build out its cloud offerings, with a four-pillar strategy to help customers build private, public and hybrid clouds on its Unified Computing System (UCS)

This statement – a surprise for many engineers at execution level in Cisco who are reading on Internet what their company is up to – contradicts teh claim that Cisco has a Private Cloud Solution now.

The litmus test to identify the real Private Cloud

Can someone do one table comparison of all the private clouds offering on Paul Morse’s web site? I do not mean comparing features, just a few categories:

How long does it take make an application delivered as service (comparing apples to apples)? ( less than 1 hour? Less than 1 day? Less than 1 week? More than 1 week?
What is the skill level required for (1)? . Rate 1 any user, 2 any sysadmin, no training, 3 only trained computer science sysadmins
Does it have a ready to use billing system to be used internally and externally?(Most reply” it has “hooks” to external “unnamed billing systems). The reply is either Yes or Not.
How the scalability works? (a) More servers on site or other sites inside the same org. are added automatically function of aggregated demand? Or (b) servers are added from public sites for additional costs? If (4) , how outside bills are allocated to internal and external users?

I don’t think there is even one person among the people I know – and I know some very competent people – who is able to answer these questions for each product from Paul Morse rather complete private clouds list.

IMHO, if the resulting data center can not provide satisfactory replies to (1) through (4) questions without exception, no matter what product is used, we do not have a cloud, but another, slightly less cumbersome to run data center

Note none of the litmus test questions include virtualization. Virtualization is just one tool, not an end by itself