Trust Silicon Valley to try to make OSS exciting
Alexander Harrowell, Senior Analyst, SoHo & SME Services
A major technology trend at MWC is the increasingly rich ecosystem of open source projects that promise to automate the day-to-day business of providing telecoms services. Telcos, for their part, have discovered the DevOps concept, and not before time. A crucial lesson about DevOps (one which the IT majors learned more than a decade ago) is that the "Ops" part is as important as the "Dev" part, and the software and processes it uses are critical to its success. An enormous amount of effort has gone into creating software that supports the software development and deployment cycle, which then supports the applications. This lesson is just beginning to be hoisted aboard in telecoms.
Consider a typical multisite, medium-sized business, a customer of Telco X's business division. This company has somewhere between 100–500 employees and between three and five locations, quite possibly in at least two countries. Companies like this make up the core of the global Mittelstand, typically being export-oriented manufacturers. The classic telecoms product for these companies at present is probably a multipoint Ethernet VPN with direct peering to the major clouds (something like Comcast Business's EVPL service). In the bright, bright NFV future (call it "SD-WAN" if you like), each of the five locations needs a branch-office device with at least one virtual machine configured. Furthermore, there is a requirement for two VMs (two, for redundancy) reasonably local to each site to terminate the connection, plus two at the E-NNI interface with the cloud provider, another pair at the internet gateway, and any additional machines supporting applications. We should assume at least five to 10 virtualized software loads somewhere in the network per SME, which unavoidably means that Telco X's business is going to be one of the larger OpenStack deployments in the world, as it has several thousand SME and a few dozen large enterprise customers, and that its operational challenges will be rather complicated.
This is something our conversations with major IT vendors repeatedly brought up. A common source of project failures is that the operations challenge gets rapidly worse after a threshold level is crossed. As a result, it is tempting to duck it and stick with old software releases. This, inevitably, leads to an accumulation of technical debt and, down the track, an expensive and painful forklift-upgrade to catch up, which probably results in problems integrating legacy applications or recondite hardware. Alternatively, new customers get the new stuff and the old ones get the old stuff, which both annoys the old customers and leads to a rapid increase in the complexity of the system. One of the advantages both cable MSOs and disruptive operators like Free have is a relatively homogenous fleet of CPE, network equipment, and software. This supports a faster deployment cycle, less time-to-revenue, and faster recovery when things break, as they inevitably will.
But if you are talking about a software deployment cycle at all, you are ahead of many carriers. This is why we were so cheered to hear that Ngena is committed to three-monthly releases, the same as the OpenStack drumbeat, which is probably no accident. In a software-defined future, operators will increasingly compete on the basis of their development and deployment practices. They will also need to cooperate, notably to achieve a more homogenous technology base. This explains another MWC17 trend – even presentations on 5G standardization are starting to mention open source.