Shared publicly  - 
My own letter to OpenStack management

+Randy Bias of +Cloudscaling started quite some stir in the cloud community when he wrote a note to +OpenStack's management saying it should work on building in API compatibility with Amazon's Web Service stack (OpenStack is an open source foundation and code base that powers tons of cloud infrastructure used by CERN, Comcast, Rackspace, and many others).

You can read Randy's note, and the discussion it caused over on Techcrunch:

Lately I've been meeting with a raft of Amazon customers, from Flipboard to SmugMug to Box to others. Heck, a good percentage of the hundreds of interviews I've done over the past four years are with Amazon customers. If you don't know, I work for +Rackspace in a unique role: I build relationships with startups and get to know why they make the technology choices they do (I bring people onto my show that are often hosting on our competitors, whether Amazon, Google, Microsoft because that's the way Rackspace will keep up its competitive edge and not get complacent about the innovation happening elsewhere). 

Not a single startup has told me that they won't go with OpenStack because of API compatibility issues. Rather they say they don't see an innovation alternative yet to Amazon (although increasingly they are paying attention to OpenStack BECAUSE OpenStack has gotten momentum as an innovation competitor). 

Which is why I am writing this letter to OpenStack's management.

If you think Cloud innovation is finished, or that only Amazon can innovate (IE, do new things for new markets) then by all means you should drop everything and make OpenStack 100% compatible with Amazon's APIs. 

But, if you believe, like I do, that we're entering into a new age that demands new technologies (thanks to wearable computing, sensors, and new business demands to personalize service and build new collaborative enterprises -- thanks +Jeremiah Owyang for doing such good work detailing why that's important) then you must dismiss Randy Bias' advice and get back to work on building the future.

This all got very clear to me when I was talking with +MindTouch's cofounder and CEO +Aaron Fulkerson. I've copied that interview here. He's built a system that lets companies radically improve their customer service. It's built on top of Amazon. It's funny to recommend you use a service that is built on our competition, but that's how I roll -- I want you to use the best service, even if it doesn't serve me or my employer well.

After the cameras were off we talked about the infrastructure choices that Mindtouch has made. He told me that nearly none of the technologies underneath Mindtouch existed five years ago. Things like +MongoDB

He told me he doesn't see innovation slowing down at all and he's always looking for a better way to do things. If a new cloud technology came along that let him analyze database entries 10x faster, for instance, he'd immediately consider it. +Don MacAskill, just last week, told me the same thing and told me he wishes that OpenStack would do things Amazon isn't able to do. Later today SmugMug is turning on a two-year redesign and he's bet his business 100% on the cloud and is looking to set a course for SmugMug's next few years and is evaluating what cloud providers are the most innovative. For him it isn't about API compatibility. That won't get him to switch his business over to OpenStack. He would consider moving, though, if OpenStack brought something to the table that Amazon couldn't match (he's very happy to talk with strategists at OpenStack to explain what his business needs in the next year or two).

For both Aaron and Don, moving their entire businesses from one cloud provider to another wouldn't be made just to save a few pennies per gigabyte or because there was API compatibility or even to get better customer support, like what Rackspace offers. The reason they'd move is if there was a 10x improvement to their own business.

That's the challenge to OpenStack at this point.

There are hundreds of companies working together on OpenStack. Including my employer, who has put tens of millions of dollars of investment (since we gifted a large part of the first code base to this open source effort and have since dedicated many employee hours to this). But even with hundreds of companies working together and this investment in code, R&D, and money OpenStack has limited resources.

It's clear we have two philosophies that are conflicting here. One wants those limited resources to be spent on making APIs compatible with Amazon. One wants those limited resources to be spent on making new cloud systems that will have a 10x return for people like Aaron and Don.

If you believe cloud innovation is slowing down, you should listen to Randy Bias because there will be huge value in providing an API alternative, not an innovation alternative, if that is the case.

If you believe, like I do, that we are going to see more change in cloud infrastructure in the next five years than in the past 10, then keep investing in real innovation and keep pushing to bring the contextual age to the world (which will need 10x or even 100x the capabilities of today's clouds -- think about what Toyota will need if self-driving cars become commonplace. The Google self-driving car generates 700 megabytes PER SECOND of data. No cloud today could deal with that level of data throughput, even if you filter it first in the cloud. Heck, lets not get that esoteric. What happens if Nike FuelBand adds three more sensors and sells 100 million units?).

All of the businesses I've talked to want to see an innovation alternative to Amazon. OpenStack is set up to be that innovation alternative, but spending your limited resources on just copying Amazon is NOT what they are asking for and I believe would derail OpenStack's momentum (which is very real, just look at the last design summit which had thousands of attendees, or look at the wins, whether at CERN, Comcast, or other places).

But, maybe I'm wrong. Maybe the next five years won't see a surge in new sensors. Maybe the next five years won't see new wearable computing devices take off. Maybe the next five years won't see demands for new contextual systems. Maybe the next five years won't see demands on businesses to get more data on their customers. 

I'm pretty sure I'm not wrong, because I've met with enough startups to see a new wave of innovation is demanded.

Thanks for listening.

To everyone else, do you have advice for OpenStack's management?

Do you agree or disagree with my conclusions? 

UPDATE: Randy claims I misrepresented him by making him sound anti innovation. I'm sorry about that but you are asking OpenStack to do work that takes engineers away from innovation to do reverse engineering and API compatibility development and testing work.
Jeremiah Owyang's profile photoMark Collier's profile photoRandy Bias's profile photoLloyd Dewolf's profile photo
Thanks, +Robert Scoble - good info.   By the way, are those your "Glass" glasses perched on your head?  You mean you don't wear them ALL the time? 
I have to say, I'm going to OpenStack (Rackspace) first and then Amazon has to do the work to convert me if they want. Rackspace's APIs have no issues and they are easy to use. No issues at all.
Thank you for the post +Robert Scoble, it's for posts like this one that i "suffer" through your animated autoawesome photos :) ;)
Good stuff, Robert. As you know, there will always be people looking to the past for direction. I am excited that RackSpace and others are looking to the future. Thanks for the insightful letter.
It's more a matter of lack of vendor competition I think, regardless of OpenStack. The ease with which Amazon combines IAAS and services into something which can be PaaS, but doesn't have to be; makes it a difficult target for competition.
PaaS offerings aren't quite the same, not even OpenShift over OpenStack, because they are too restrictive they don't allow the flexibility (ironically) that having your own iron provides.  EC2 and AWS with any combination required for the idea of a Platform for a company is still the nearest thing to godliness available.

But, in Europe especially, we need vendor competition in this market segment and it needs to be in every EEA country to satisfy the varied DPA requirements.

This is somewhat orthogonal to your open letter I'm afraid but I think its the real problem and OpenStack on its own isn't and shouldn't compete directly but providing the IaaS upon which simple services can be easily and transparently hosted is where it needs to be.  

(Still waiting for the thousand service flowers to bloom on OpenStack)...
We've got stuff I want to move to openstack or AWS. At the moment it's about data sovereignty. Especially in light of the NSA revelations we won't be moving to anything owned by a US org. So that leaves us with openstack but no vendor. I believe there's still no good vendor catalog that shows which parts of openstack are available at which vendors, so it's just too low a priority right now to spend excessive effort to find a vendor. 
I dont see amazon inovating in the cloud space..i tried but never use it really in a day by day basis.i done see Amazon as inovator in the cloud...i may be wrong.
Amen!  While conformity with AWS may be appealing, the real jewel is to get Amazon to be compliant with an open standard.  To do this, the open platform needs to have all of the important features that the Amazon stack has to offer.  And then customers need to demand adherence to open standards as a key decision point for opening up their wallets.  Until and unless the customers demand adherence (and they should), there will be no economic incentive for Amazon to conform with.  

BTW, the other alternative is for Amazon customers to demand that Amazon completely open their platform.  Then, it will be a decision based upon features/capabilities and market momentum.  Do I expect that to happen?  No!  So I'm placing my bet on the existing open standard.  
+Robert Scoble The one thing missing from Randy's letter was the one thing you touched on - what do customers want? And you are right - cloud customers want flexibility and broad support - two things you can't get from AWS. I have argued that the OpenStack ecosystem wouldn't even exist if it was an AWS clone.

Providing an AWS API would not add flexibility to your cloud ecosystem, it would actually handcuff it in constant reverse engineering every time AWS changed it's API's. OpenStack is open... if AWS wants to provide API's (and maintain them) , they are as free as anyone to contribute code to make it happen.
I think you're wrong, but for different reasons. Please excuse the long note, but Rackspace is following a similar pattern to what happened at Digital. I feel you've missed one very important angle.

One of the folks you interviewed recently was Gordon Bell. One of his assets was the ability, in one slide, to summarise upcoming technology choices and to convey where the industry was going. After he left DEC, the company got stuck on a 4 year quest for the next CPU technology (VAX follow on). Look at Gordon's logarithmic CPU speed, cost timeline, at it was obvious to go 64-bit CMOS. He'd tell you the company was mad to follow high cost, Trilogy ECL designs for IBM mainframe performance when NVAX was on schedule to run just as fast on $300 silicon in the labs. The company wasted 4 years (and hundreds of million dollars) on the wrong technology choices, blinded by the light of overtaking IBM in size, and flip flopping between 32-bits and 64-bit address space designs to frustrate all progress in engineering. The end result was little progress, so Exec management culled the very projects that were DECs future. 5 years later, they'd lost Dave Cutler, but got his Prism/Mica design to market - adding a few instructions and calling it Alpha AXP.

Alpha, albeit 5 years late, nonetheless outperformed every CPU in the market by 2x for the next 10 years. However, the delays were fatal, and after 2 attempts to become relevant again, the company got bought by Compaq.

So, what is Rackspace missing? Data point one is at: For four specific weeks between May and June 2013, Digital Ocean transferred in nearly 1,500 Rackspace customers, and over 600 from AWS. Besides those, look to see how many first timers, I would guess mainly Startups, also appeared. Then ask yourself, what is happening here?

The one thing to discount is how many of these were from bit coin related traffic, taking advantage of special low cost hosting offers and which may invalidate the impact of the perceived growth.

On the other hand, Mike Prettejohn at Netcraft mentioned to me that the number of host machines at Digital Ocean seemed (by mid July) to be circa 10,000 and growing by 30% per month.

Hence, at face value, it's like a classic yacht race where two competing vessels are nip and tuck to lead the race, when something unexpected flies past them both.

The fundamental question is, of course, what will the computing landscape look like in 5 years time. Is it best served by a complex beast that will be costly to service providers and undifferentiated, or is it likely to be utility provided compute, storage and network capacity that have taken proprietary tech vendors to price points their current business models can't reach?

Another data point is how software stacks are simplifying out (think node.js, meteor.js, Ruby on Rails). These also appear to feature very strongly in Digital Oceans base.

So, are Rackspace blind to what's nibbling them underneath? It's that rather than AWS vs OpenStack that would be of more concern to me - if I were a Rackspace employee - at the moment. Your mileage may vary.
+Robert Scoble absolutely. Matching AWS APIs is race to the bottom. "innovation alternative" is the key. I was primary dev of two public AWS services (RDS and Beanstalk) and now work for Azure so I say I know about this area a bit. From my personal POV, there is absolute sea of opportunities, including things that AWS does not get, does not know how to do or has not considered. The problem with OpenStack is design by committee. The best thing that can happen for them if someone breaks away with wave of fresh innovative services, not necessarily backed by OpenStack official APIs.
What is innovative about OpenStack today? How is OpenStack set up to be that innovation alternative?
That is another opportunity (what +Robert Scoble just mentioned) - cater to extend-your-private cloud community. I personally believe that to be a stopgap on the way to true utility computing (all public clouds via standard APIs), but today this is huge market.
+Robert Scoble understand, but you've missed my point. Rackspace may have gained 4 Enterprise, at scale customers, but at the same time it lost 1,475 customers in one month to one named competitor. That would worry me (though I completely understand if Rackspace Exec Mgmt were aware, and had changed the future focus to service intensive Enterprise customers only).
Open source will win in the end , good arguments +Robert Scoble.  Just because it is open source vs corporate allows Openstack to be more versatile and reactive to needs of its customers.  Flexibility and being able to meet the needs of your customer base is the only way to go forward these days.
FWIW, it's $10 per month for a single core, 1GB memory, 30GB storage and 2TB network capacity instance parked out on the Internet with it's own IP address. With CentOS, Fedora or Ubuntu preloaded, so add Node.js, MongoDB, Meteor and CoffeeScript, and away. Not fussed whether it was SSD or not, and certainly for much more than DB work :-)

Simplicity Sells.
Too often, we seem to select technology in the same fashion that victors were chosen in Highlander - via a winner-takes-all death match.  After all, there can be only one, right?  Wrong.  Both can survive - though both have their specific use cases.  

Dan Kuisnetzky proposes an analogy: grocery stores and restaurants (see  He asserts that OpenStack is like a grocery format - where some assembly (i.e., food preparation) is needed by the consumer.  He further asserts that Amazon is much more akin to the restaurant where you order a meal that is prepared for you.   [Note: Take the analogy for what it's worth.]   

This analogy does have merit.  But as consumers, sometimes we want both options.  Or maybe we want both at different times.  So how do we, as customers, direct both of our suppliers?  How do we tell them which services we want to build ourselves and which ones we are willing to purchase?  Maybe the answer lies in SaaS vendors.  When are companies willing to choose a big SaaS vendor?  When they've failed doing it by themselves.  And when will companies choose to build it for themselves?  When they've failed to get the value from a SaaS vendor.

Of course, that quandary leads me to one inescapable conclusion: there will always be the need tor good consultants who can work both sides of these equations.
Here’s the problem as I see it. We’re three years in to OpenStack, and we still haven’t got one of the basic IaaS services: elastic load balancing. Why? In large measure, it’s because we’ve spent too much time on accommodating diversity — different hypervisors, different storage services, plugins and metaplugins for networking — and not enough time on creating a crisp set of requirements and executing against them. Too much time accommodating vendors (and I work for one of them, though I’m speaking for myself), too little time on delivering a consistent vision for the customer. We can do better.

In engineering, constraints are useful. They help you stay focussed, they make it clear what’s core and what is not. (See Rob Hirschfeld’s recent posts on the subject.) Linux succeeded because of POSIX. MySQL succeeded because of SQL. And the ironic thing is that the current OpenStack APIs don’t provide an alternative set of constraints, because there are so many extensions and options. (Hence Josh’s “RefStack” proposal.)

Innovation doesn’t have to be expressed at the service/API level. Adopting the S3 API in full wouldn’t limit the competition between Swift, Ceph, Riak and others to come up with an object storage system that offered kick-ass operational and economic benefits. And if ELB compatibility had been included in the original set of goals, I expect that LBaaS would be much further along than it is today
Did you? I thought you were accurate. Maybe he could highlight what he feels was misrepresented?
There is also this little company called google that has a cloud service. It might do some damage in the marketplace.
+Robert Scoble none of what you listed in response to me is technological innovation. It's very much the Linux model of assimilation. What are the technology innovations that would motivate the companies you listed above?

OpenStack's biggest challenge today is the gaps in functionality and it's the competitive APIs that make this glaring!
+Lloyd Dewolf, I think that you may be confusing invention with innovation.  But your point is valid.  Technology that brings new features and functions to the market has immense merit.  

But as a general rule open source  tools/technologies are not usually about bringing inventive new features.  They are about using open processes to create economic advantages - not just feature equivalence.  Openstack is about leveling the playing field for customers.  It is about providing commodity services at commodity prices.  

There will always be reasons for costly (and closed) products and services.  But at some point, legacy/commodity goods and services demand openness and true competition.  After all, wasn't that the point that the US Constitutional framers made when they gave time-bounded monopolies for invention?  And once invention slid inexorably towards commoditization, the IP would also go into the public trust.  
Interesting thread +Robert Scoble - the key question here isn't what startups want but rather what is the compelling proposition for real enterprises. Those are (generally) where the real cloud opportunity lies and the prospects that most in the OpenStack community want to be thinking about. these are also the customers that +Randy Bias sells to at CloudScaling. 

My 2 cents on the firestorm here-
+Geoff Arnold & +Ben Kepes  well put!

+Lorin Olsen I quite agree. I think there is innovation in the way OpenStack is being developed and positioned, but that isn't +Robert Scoble  argument. Robert doesn't back up the infrastructure innovation that he says will move the people that he has talked to. His response to me is ask someone else!

Of course, not a single startup has told +Robert Scoble that they won't go with OpenStack because of API compatibility issues. That would be out of context. I'm 100% certain that this issue has come up numerous times when potential customers have gotten to the technical details and real costs with Rackspace engineers.

Oh and the limited resources argument is grating considering there are significant fundamental differences -- read hypervisors -- between Rackspace private edition and what they use to power their own "open cloud" [sic]. All argument about limited resources will fall on deaf ears when discussing a vendors driven project with naturally divergent and conflicting goals.

On the other hand, +Randy Bias  constructs an argument backing it up about as well as anyone can, reflecting customers requirements.
I've tried posting this comment twice but it never shows up in the stream.  Third time...

Robert, I think you are spot on with the notion that innovation is just getting started in cloud computing. 

Forrester recently estimated that IT spending will be over $2 trillion dollars in 2013.  That's $2,100,000,000,000.00  Even if Amazon's cloud computing revenue were $5B this year (a generous estimate), that's 1/5 of 1 %.  Certainly those numbers aren't directly comparable per se, but it goes give you an idea of the order of magnitude we are talking. It seems foolish to suggest the game is over. 

Also, one of the reasons OpenStack has attracted developers is the opportunity to break new ground.  I doubt we would have seen the developer community grow from 20 to over 1000 in just 3 years if the mission (to build the ubiquitous cloud computing platform) had ruled out doing new things from the start.  

APIs really are a red herring IMHO.  OpenStack actually includes an AWS compatible API, but only 30% of users are choosing to use it according to the recent user survey: Certainly that number could go up or down, but I don't think it should be a distraction from the goal of moving quickly to deliver the best cloud computing platform on the planet in response to real user needs.  

The users I talk to are focused on empowering their developers to deliver new features or products quickly.  OpenStack is an attractive platform to target, because it's 1) free and open source 2) innovating quickly with the help of >1000 developers with domain expertise is compute, storage, and networking delivering new releases every 6 months 3) there are now OpenStack public clouds available in more cites than Amazon has regions, and 4) Users can become an active part of the community with a say in the future direction of the platform they are betting on, without fear of vendor lock in.
+Robert Scoble has any of the other major backing companies come out strongly against the value of Amazon EC2 and S3 support? I don't know of any, but I know of a number of companies that have vocally expressed the opposite view and put effort behind projects to address it.

You should really have the conversations with direct +Randy Bias and other executives at OpenStack companies focused on the private cloud including my employer Piston Cloud and our customers. Without doubt API compatibility is often selection criteria. If only that 30% that +Mark Collier  referenced in that self-selected survey will help wake up Rackspace!

I generally agree with you +Mark Collier and agree that almost all focus should be on innovation ahead, the problem with +Robert Scoble article is that it's even a more polarizing argument than +Randy Bias. It's only a conflicting view today, after , if you are Rackspace -- I've never heard an answer that operations people agree with.

+Mark Collier, I try to be careful not to claim "OpenStack actually includes an AWS compatible API", as it is but a tip of the AWS API -- or do you have data that demonstrates otherwise?
Good luck with that. At 10,000 feet it sounds ground, but infrastructure is all in details. I'd love to see this broken out into the real opportunities and challenges in the various areas.

At the same time, the pragmatic will put the polarized views aside and get specific about the strengths and weaknesses, enable operators and developers to be flexible, and focus on the limited areas 3x-10x improvements on the horizon.

I do hope you continue to take the opportunity to speak to those focused on OpenStack for private cloud. It was inspiring to see Rackspace Senior Vice President of Private Cloud Jim Curry invite HubSpot CIO Jim O’Neill to the stage to share how HubSpot integrate OpenStack for private cloud with their significant AWS utilization.

I'm extremely excited to continue to see innovation in OpenStack!
+Robert Scoble 

I'm glad someone has posted a highly visible contra to my open letter. I just wish it were more useful than this one. 

Here's the problem with your article, Robert:

- I never said OpenStack shouldn't innovate
- I never said OpenStack should ONLY be AWS compatible

Whether this misrepresentation is deliberate or not is really of no consequence to anyone other than me. What is of consequence to everyone who read your article is the fact that you are treating the embrace of AWS APIs as a zero sum game. The logical extension of your argument is that OpenStack loses if there is better AWS compatibility. 

I don't see how that can be.

The flat networking in OpenStack is like EC2 Classic networking. We're supposed to stop working on this because it's how AWS does things? Work on the AWS APIs helps people move from AWS to OpenStack and we're supposed to stop working on them? The formation of the Cinder project and its scheduler are extremely similar to AWS EBS and we're supposed to stop working on them?

* You are actually advocating limiting innovation in OpenStack. *  Compatibility with AWS is a legitimate area of innovation in OpenStack.

It is misleading to suggest that working on AWS compatibility detracts from OpenStack. It does not.

Robert, I respect you. You are entitled to voice an opinion, and that opinion carries weight. But your opinion is not informed by the same real-world experience in OpenStack that others have brought to this conversation.

But what disappoints me most deeply is that your entire premise is based on a straw man that does not represent my views. Maybe that's based on a misreading of my open letter (and the many and varied public comments since then), or maybe it's intentional. Either way, it's quite unbelievable that someone of your reputation wrote and published something this far off base.
+Robert Scoble First off, let me thank you for driving so much interest and traffic to me, the Cloudscaling website, and my point-of-view. It's a rare employee that will violate long-standing public relations best practices and drag the large company into a big public debate with the small company.  The exposure is fantastic. Thanks again.

Second, your assertion about development being a zero sum game is obviously inaccurate, particularly when it comes to open source communities.  It should be obvious, but I'll spell it out for you: the idea that the number of developer hours is fixed is untrue.  The number of developer hours is highly variate, particularly because the developer count changes constantly, largely based on individual interest.  OpenStack is winning because its been an inclusive environment.  You are advocating being an exclusive environment, which might actually reduce developer hours.  It also runs completely counter to the OpenStack culture. The most ardent of those who disagree with me do NOT advocate ignoring or removing the AWS APIs.  You, and by extension, Rackspace, are a lone voice in the wild there.

I spent time with +Mark Collier  at OSCON last week.  I read his article.  There was nothing inaccurate in his article and nothing I disagreed with.  Mark had a very well reasoned and balanced point of view as he should in such an article.  His initial response says it all: "There's not a binary 'yes' or 'no' answer to this." At most he disagreed with both of us.  He certainly wasn't agreeing with you on your viewpoint of ignoring the AWS APIs and architecture.  Not sure why you are even bringing this up.

With regards to your misrepresentations ...

You said I stated that OpenStack:

"should work on building in API compatibility with Amazon's Web Service stack"

What I actually said was:

"Now, the issue has become urgent, and I hope to convince you to join me in advocating that OpenStack immediately and deliberately embrace the APIs and features of established public clouds."

The AWS API was the original native API in OpenStack and I'm calling for it to continue to be updated and expanded upon.  Not to be built.  It doesn't need to be built as it's already there.

You also said:

"It's clear we have two philosophies that are conflicting here. One wants those limited resources to be spent on making APIs compatible with Amazon."

But that is not my philosophy.  The APIs in OpenStack are already compatible with AWS.  They had the AWS APIs before the Rackspace APIs were added.

Also, it's obvious that the resources are fungible and that it's not a zero sum game and positioning it as a zero sum game, when it's clearly not, is just a way for you to try to continue to prop up a poor argument by asserting a reality that does not exist.

You also said:

"but spending your limited resources on just copying Amazon is NOT what they are asking for and I believe would derail OpenStack's momentum"

Clearly this implies that I asked for "just" slavishly copying the Amazon APIs, which is not what I said.

I said:

"In short, the community controls the direction of the project, and it’s time we advocate a public cloud compatibility strategy that is in all our best interests ..."

This clearly calls for compatibility with major public clouds, including Amazon, Google, Azure, and even Rackspace.  In other parts of the letter I even advocated adding the vCloud Director APIs.

Clearly, this is a straw man argument on your part, claiming that I said X, when I actually said Y, and then attempting (poorly) to knockdown the X argument.


+Mark Collier It's 33.5% and that's a very significant constituency as I am sure you would agree.  Thanks. 
+Robert Scoble What's lamer is trying to make me look bad by calling me out on some error in protocol on my part for Google+, a tool I don't use. I'm a newbie and will make mistakes.

Calling out newbies making honest mistakes just makes you look petty.
Wow +Robert Scoble! +Randy Bias provides thorough, on-topic responses, and you bully him about not addressing every single angle and about Google+ etiquette.

+Robert Scoble I am excited about you getting more involved in OpenStack!
Add a comment...