Netflix, Video, and Service Economics

Netflix is still smarting from its abortive attempt to raise prices and change its business model by splitting its mail-DVD and streaming activities, or so says the Street pundits.  I’m not totally convinced.  Yes, I believe that these things did in fact increase customer dissatisfaction and churn, and yes that’s responsible for their larger-than-expected loss.  That’s not the question, though.  Announcing something unpopular as an alternative to accepting death isn’t a hard choice.  I don’t think that Netflix had, or has, any option here.

We are too used to thinking that everything should be free online, or if not free at least working its way in that direction.  Naturally everyone’s upset at a price hike.  The challenge is that as a public company, Netflix is expected to create shareholder value, which means increasing profits over time.  Their licensing costs have nowhere to go but up.  Their distribution costs have nowhere to go but up.  Their total addressable market is certain to plateau and there’s increased competition from other players, from Apple and Amazon to Hulu and the networks and even studios themselves.  So faced with rising costs and shrinking margins instead of growing ones, they do what they have to do.  Better face the demons now than later, when they’re going to be even more terrible than they are today.

There is a question raised by Netflix, though, and that question is whether there can be successful “portal sites” at all.  If consumers want to pay less every year and studios or networks with content want more revenue themselves, then the guy in the middle is going to be squeezed.  Ultimately you have to shorten the distribution chain, and since most of the content ends up cached in a CDN anyway, why not distribute directly if you’re a network or studio?  After all, it’s their brand, their content.  Which is why content is king, of course.

That operators are looking for ways to create revenue with lower traffic is getting clearer every day as they move into more advanced services.  That has implications on where capex goes, shifting from network to data center.  Yesterday, Bell Canada announced the selection of Juniper’s QFabric as the connection heart of a managed service hosting data center, which means as the heart of a cloud.  I’ve always said that fabric technology in the data center was a perfect match to cloud infrastructure, which is of course essential if you’re going to host managed services or any other kind of software-feature element in network services.  Not to mention cloud services.

Data centers aren’t automatically fabric consumers.  First, you need to have a lot of scale in order to drive the number of layers in a switch hierarchy up to the point where fabric benefits would offset the cost of modernization.  Second, you need a lot of horizontal traffic, between servers or between servers and storage, or you don’t need the high-performance connectivity that fabric provides.

Cloud services in demand an efficient resource pool, which in turn demands a lot of horizontal traffic.  Cloud services are based on economy of scale, which demands a lot of scale.  Application-specific servers or even VMs that are statically assigned to servers tend to generate mostly vertical traffic, and they also don’t demand highly connective storage resources.  As a result, storage can stay on dedicated SANs and switch hierarchies don’t hurt much.  But if you add in a cloud dimension you add in a need for flexible assignment of applications and components to storage, and you generate more of both inter-process and storage traffic.  Same with big SOA-driven applications, which is why some verticals built around a behemoth mission-critical software core are fabric prospects despite their smaller data center footprint.

Some on the Street have been saying that it’s the ramp of QFabric that’s going to turn Juniper’s trajectory upward again, but the challenge for Juniper is quantity.  Carrier clouds are nice wins, but there are only so many carriers.  Juniper either has to make QFabric valuable beyond the cloud or prove that everyone will build one.  Everyone won’t build one, which means looking beyond the clouds in an era where the cloud is just such a facile justification for change that it’s hard to resist.  The BC deal is a good start to be sure, but QFabric needs to be fleshed out, made part of a broader architecture, to link it to opportunities on the largest possible scale.


Leave a Reply