The definition of cloud broker attracts some controversy. Too bad, because I am trying to teach about cloud computing. Recently I wrote “there are no cloudbrokers“. My argument is basically that all known instances of cloud brokers actually provide a cloud service themselves.
Michiel Steltman takes issue with this in “the definition of broker needs to change” and tries to refine the definition of ‘broker’ as a party that is not in the ‘direct chain of delivery’. There is some value in this refinement, though it invalidates most examples of a cloud broker that NIST mentions.
Ben Kepes, in “cloud broker is an invalid use of language” takes a different route by analysing the original meaning of the word ‘broker’, and then comes to the conclusion that most of NIST’s categories are not examples of brokers.
Lets dive a little deeper. NIST lists three types of cloud broker functions: service intermediation, service aggregation and service arbitrage.
Service intermediation sits between a consumer and a provider, while adding services such as identity and security management. There is likely to be an SLA (service level agreement) around this. To me this makes service intermediaries responsible for an online service, and thus acts like a cloud service provider. It is not relevant here that this intermediary is using a provider on the back end. Most of them do.
Service aggregation sits between a consumer and multiple providers, combining these services for example by integrating data. This makes it even more obvious to me that these have their own responsibility in delivering a service and therefore are a provider.
Service arbitrage is least well defined (in fact the example NIST gives appears to be ripped from a non cloud text). In its vague form it might correspond to what is commonly known as a ‘broker’. But does it run an online service, i.e. is there ‘provisioning’ of ‘configurable computing resources’? If so, it is a provider to me. If not, you might call it a broker, or a reseller, or a directory service. But this is a very limited subset of the examples NIST gives.
2 Comments on “Why I think most cloud brokers are actually cloud service providers”
Gilbert Pilz26 March 2012 at 15:40
I think we need to move past the discussion of “what is and isn’t a broker” and talk about the factors that influence the nature of multi-party scenarios (i.e. scenarios in which the cloud consumer directly or indirectly interacts with multiple cloud providers). I agree with you that, in scenarios in which the consumer interacts with a single provider and has no knowledge of or relationship with any of the downstream providers – there is no effective difference.
I wrote an article about this here: http://recursivedigressions.blogspot.com/2012/03/cloud-broker-overload.html
admin27 March 2012 at 15:11
I recommend this article. It makes a good distinction between a number of relevant concepts.