On Cloud Brokers and Whether they have a Future (Pun Intended)

While in San Francisco recently I chaired a debate to look at whether there is in fact a future in cloud futures. The debate was an event from the Cloud Market Forum, an initiative that I’m on the steering committee of that is “…an executive consortium for the world’s leading cloudonomics, IT, and
financial experts to discuss, debate, and ultimately shape, the future of cloud from an economic perspective.” Full disclosure needs to be given that the forum was set up by Strategic Blue, a company I mentored during TechStars Cloud and an organization that firmly believes that there is a future in cloud services arbitrage.

Anyway – the debate came out of Cloud2020, an event that Krishnan Subramanian (formerly independent analyst, now Director at RedHat) put on earlier in the year during which one heated session looked at the futures market for cloud computing capacity. It was only a short session but one that created a storm of controversy and encouraged some of the more thoughtful members of the community to put pen to paper and opine on the topic. in particular it is worth reading the following posts:

Given the caliber of these writers, and the very fact that so many people were motivated to write in-depth posts about the topic, it was obvious that a fuller discussion need to be had, that was the genesis of the debate in San Francisco during which Randy Bias from Cloudscaling and Mark Thiele from Switch argued against the validity of cloud marketplaces and John Cowan and James Mitchell argued the opposing view.

The really interesting thing for me at the debate was seeing the two sides argue across completely different topics. Bias and Thiele did an excellent job of discussing the technical factors that make cloud brokering difficult (lack of standards, the limited fungibility that exists with cloud resources etc). Bias in fact spoke of this factor when he said that:

I believe there will be a market. It’s just a question of when. The biggest impediments to these markets developing are 1) measurability and 2) it’s fundamentally a different kind of market unlike anything we bought, traded and exchanged in the past. There’s all kinds of things there that we don’t understand yet and we won’t understand until we start to experiment, so thank you for starting the experimentation

He also spoke to the complexity around what is trying to be traded here. He (rightly, given his position) discussed the huge differences between a barrel of oil (largely an undifferentiated product) and the complexities of IT resource, to his point:

The value measurement is something we’re not paying enough attention to. A GB is easy. A VM is a real freakin’ problem. If I go out to market and I want to get X thousands of TPS from a relational database with a million rows the problem is bounded down to where you can actually measure the value and you can actually measure what you get from the provider, but if it’s just VM’s it’s highly problematic…in order to resolve it, we’ve got to figure out what are the workloads people are using, can they be chunked into groups and classifications, such that we can actually do value measurement on the buyer and supplier side

On the other hand Cowan and Mitchell argued that this is very much a business opportunity and as such proxies will be found to make up for the technical limitations currently extant. And this is what really struck me about the debate – no one really argues against the notion of a cloud marketplace, they all just worry about what will be used as measures and how to solve the technical problems to standing up a marketplace. And this is the real barrier to the success of a cloud marketplace. Get technologists in the room  and discussions ensure about the (fully justified) technical burdens around cloud exchange – as James Urquhart rightly pointed out from a technical perspective:

What is the commodity? Quite frankly we are talking about doing commodity trading on a very complex system. It’s not one thing – it’s a bunch of things that are interrelated with each other that don’t operate unless all the other agents are in place and communicating to each other the way they need to in order to execute

But having that discussion in a marketplace debate is like having electrical engineers discuss switching technologies during a debate around electricity marketplaces. We know electricity marketplaces exist, we know distinct forms of electricity are exchanged and we intuitvely know that there is lots of “plumbing” that needs to occur for this to happen seamlessly. but do any of those people running energy marketplaces spend their time worrying about hi current switching bridges? Or ramp up time for coal fired plants? Of course not, they’ve abstracted the technology away from the transaction.

And this is the key requirement that will allow a cloud marketplace to really work. Those providing the marketplace need to give up on having their conversations with CIOs and CTOs and focus instead on having a financial conversation. I t seems to me that the failings of previous cloud marketplaces (SpotCloud among them) had as much to do with the wrong message aimed at the wrong audience as it did with timing and market maturity. Final words go to Strategic Blue CEO James Mitchell who, I believe, should now stop talking to technologists and head back to his suited and booted former life talking with purely financial types. Mitchell summed it up nicely when he said that:

The majority of this has absolutely nothing to do with technology. It’s totally irrelevant. (From the financial market perspective) I just want someone to tell me how much of something, what it is so I know its the same thing being delivered, when they want it, and what price will work. That’s all I care about

3 Comments
  • Very interesting discussion. And I agree it’s probably at the purely financial level that this trading question should be dealt with.

    But in James’ quote, where he says “*what it is* so I know it’s the same thing being delivered”, the “what it is” is exactly the question we don’t have an answer for (at least for now).

    What makes the answer so complex for VMs is that, even if we could compare VM one-to-one, the “total value” of the VM also comes from what other services are around it (do software running on this VM have access to DynamoDB like systems, BigQuery like systems, S3 like systems …).

    Maybe such comparisons will be possible in “subsets of the cloud” (ie limited range of services). But from what I can see, the more it goes, the more we see high level services pop up here and there (AWS Redshift being the best example …) so the VM-to-VM comparison might not even be relevant in a few years …

  • Augusto Piazza |

    Before creating a futures market of cloud products, you need to create a marketplace for those products; check which are the one most demanded and traded; only then the question on futures will be most relevant.

    Before that, you may, as a futures trader, buy the wrong product and quickly burst with your unsold inventory which can’t be stored. A very risky business.

    There are several companies trying to build these marketplaces.

    ComputeNext for example has been addressing all the issues you are mentioning in the article for the past 3 years and come up with some answers.

    It has now more than 20 providers that have joined their marketplace.

    Not enough? Indeed, it still a small market, but behind these providers there are more than 700 standardized computing offers and 40 storage offers.

    It’s therfore the beginning of a supermarket for cloud as you start having through a single cashier a real comparable choice in price, location, quality and quantity of public cloud services, which is what a cloud marketplace is all about.

  • Pingback: On Standards in the Cloud

    • Augusto Piazza |

      The meaning of each word is important.

      By “standardized computing offers”, I meant “normalized computing offers”, i.e. offers that are comparable in terms of quantity, quality and price, across 15 different “standards” or platforms, all managed by a single RESTful API.

      The accepted platforms include such well-known names as Openstack, CloudStack, vSphere, EC2, KVM, etc.

      The belief behind this approach is not to choose any standard, but to accept them all by using an asynchronous external API that abstracts the complexity in discovering and utilizing heterogeneous cloud infrastructures.

      Thus making it a very relaxed world for users, as all offers are live and actionable i.e. their workload can start running within minutes after been reserved and purchased.

  • Pingback: On Cloud Brokers and Whether they have a Future (Pun Intended) - Jan Brass

  • Pingback: On Standards in the Cloud | The Diversity Blog - SaaS, Cloud & Business Strategy

  • Pingback: On Standards in the Cloud | My Blog

  • Pingback: Cloud Brokers, Fungibility and the Amazon-Shaped Elephant in the Room | The Diversity Blog - SaaS, Cloud & Business Strategy

Leave a Reply