Yesterday I wrote about my skepticism about service brokers. Today, I see this blog post on GigaOm, "Is the cloud exchange concept ready for primetime? Deutsche Boerse thinks so."
Basically, The German company said on Tuesday that it will launch a global, vendor-neutral marketplace for compute and storage capacity in 2014
When I dig deeper, it seems that, yep they wanto be a broker of sorts. And impose fees, flatten service definitions so comparisons can be made across providers, and generally turn computing power into a commodity.
Remember Enron's bandwith exchange? Allow me to remind you. Enron set up a subsidiary to exchange bandwith between providers - which is not a bad idea as bandwith is more fungible than datacenter services and the telecom industry already had peering and interconnect practices.
But, as this Wired article details.
"We were trading bandwidth a few months ago, but the market wasn't there," said Al Butkus, vice president of energy trader Aquila (ILA). "It didn't really have to do with Enron so much as the dot-coms. There's just too much capacity out there and too little demand."
Currently, Butkus said, few ISPs and telecom firms want to use a commodity-style exchange to sell contracts. Deals tend to be negotiated over the phone, often with a broker who gets a fee for arranging the transaction.
Although Butkus believes that most bandwidth will eventually be sold through a commodity-type market, it won't happen overnight. His latest estimate is that it will take three to five years to develop a healthy and viable business for trading bandwidth contracts.
Three comments on the above quotes:
1) Compute moves at Moore's law speed. Much faster than bandwidth grows. There's no shortage of capacity.
2) As the man says, deals are negotiated. Contracts are in place.
3) The article was written in 2001 - Over 12 years ago. The prediction that it would be 3-5 years is wrong not by 12 years, but so far it's wrong by forever.
Here's a choice quote from the GigaOm article from yesterday July 2, 2013.
Like any exchange, a fee will be involved for both buyers and sellers. However, Baumann argued that this will still prove cost-effective and efficient for large buyers:
“A certain percentage goes to trade, but this is 10 times better than current procurement processes in cloud computing, which take one to one-and-a-half years. Procurement costs are enormous and time-to-market is very enormous.”
The Deutsche Boerse Cloud Exchange will begin operations in Frankfurt and New York around February or March next year, before adding Singapore for the Asian market 4-5 months later. “It will be a global play by the end of 2014,” Baumann claimed.
Clearly they haven't heard about Amazon services - it takes a credit card and 5 minutes. As for 10 times better than current procurement processes, as I said yesterday, I don't buy it.
There's a very nice academic paper from IBM Zurich (pdf) on this topic that makes some of these points and more in a more formal way. Basically, these are illiquid markets due to insufficient commoditazation that lead to wide spreads in pricing introducing high risk for the broker of services.
There may be some aspect of the offering that I'm missing and makes this scheme workable - but it hasn't become apparent.