Qualcomm StreamBoost: A Solution in Search of a Problem?
In January, Qualcomm introduced StreamBoost, a software that can be licensed to be installed on domestic broadband gateways by their OEMs. The software takes tight control of the routing with the purpose of providing traffic shaping at the entry point to the home. The outcome is dynamic, intelligent Quality of Service (QoS) management.
I spoke recently with Mike Cubbage, the director of business development for Qualcomm Atheros (the network technology arm of the company, about StreamBoost, and found it to be an interesting offering with questionable application.
StreamBoost differs from standard routing QoS because prioritisation is given on a per-application basis (rather than say on a "per port" or "per IP" basis): Many thousands of applications' "bit-stream profiles" (for want of a better description) are identified by StreamBoost's central services, and those profiles are shared "via the cloud" with all the StreamBoost clients. This means that StreamBoost can actively manage many types of data using the same transport layer protocols (such as HTTP traffic), merrily distinguishing between HD YouTube videos and video being downloaded from Dropbox for example, and prioritising accordingly.
This type of central-intelligence combined with dynamic/active QoS is something that I have seen to some extent in CDNs. I asked Cubbage about this type of intelligence being of value to ISPs, and he commented that ISPs are generally cautious of traffic shaping intelligence, given the political minefield of network neutrality.
There was one aspect of this that I was a little uneasy about: I appreciate that at times households use their internet connections to capacity, and at those times StreamBoost would seem to be a logical way to at least optimise their experience at those times. I am just not convinced that domestic broadband services suffer that much from contention at the subscriber's gateway; I think far more contention happens at the DSLAM or at the BRAS further back in the ISP model, so while StreamBoost will be still optimising the use of the pipe, the local pipe is not the point where the pipe is saturated, and this will typically still cause problems with variation of the available bandwidth to the house no matter how organised that overall house-per-house aggregate traffic is.
I don't want to do the technology down: It is the first time i have seen a system like this with central intelligence, and I think as technology goes it is progressive. I do, however, get a sense that Qualcomm may have developed this technology to sell to ISPs and CDNs, and hit the net neutrality issue by forcing them to look for other ways to recoup their development investment. StreamBoost is certainly a logical approach to recoup that investment, because some edge networks certainly are highly contended and could really do with more intelligent, stream-aware WAN gateway optimisation.
I just happen to think that the contention is predominantly "in the network" and not at the edge, and so i am not sure that this isn't a great solution to a problem that not many people really have. That said, if you do have the problem this is an excellent solution.