Docker buys SocketPlane as it builds out its container-networking strategy

You can add another acquisition to Docker’s plate with the startup set to announce on Wednesday that it has bought a small networking startup called SocketPlane. The acquisition, whose financial terms were not disclosed, is just one more step in Docker’s plans to become the de-facto container-management company whose technology can play well on multiple infrastructures.

SocketPlane’s entire six-person staff is joining Docker and will be helping the container-centric startup develop a networking API that makes it possible to string together hundreds to thousands of containers together no matter if the containers “reside in different data centers,” explained Scott Johnston, Docker’s SVP of product.

The upcoming networking API can be thought of as an extension to Docker’s recently announced orchestration services. Although the new orchestration services makes it possible to spin up and manage multiple clusters of containers, the current Docker networking technology built into the platform only works well for “a handful of hosts and containers” as opposed to the thousands needed in complex environments like Spotify’s infrastructure, explained Johnston.

In order for Docker’s orchestration services to really come to fruition and deliver on the promise of spinning up and managing tons of containers, Docker requires an underlying networking API that enables all of those containers to speak to each other at a massive scale. Docker decided that it needed this technology expertise in-house, so it turned to SocketPlane, whose staff includes networking veterans from [company]Cisco[/company], OpenStack and OpenDaylight, said Johnston.

The goal is to create a networking API that can work with the gear and applications from providers like VMware/Nicira as well as Cisco and [company]Juniper[/company], explained Johnston. Theoretically, the API will make it possible for a user to build a distributed, containerized application in one data center that uses Cisco networking gear and have that application move over to an OpenStack-based cloud environment that uses Juniper gear “without breaking everything,” said Johnston.

If VMware has its own networking technology that users like, users should be able to “swap out” the Docker networking technology and use the other technology, said Johnston. Google’s own Kubernetes container orchestration system currently swaps out the existing Docker networking technology for its own networking technology, said Johnston, but once the SocketPlane team builds a workable open networking API, you can imagine a world in which users can alternate between Kubernetes or Docker’s Swarm orchestration service if they choose to.

“Say this works and we have ten or twelve implementations of the API,” said Johnston. “A network operator might want to take advantage of that.”

This makes for Docker’s third publicly-known acquisition. The startup bought devops startup Koality last October, which was preceded by the July acquisition of London-based Orchard Laboratories.

Story clarified to emphasize that the APIs will not be swapped out.

The whole world is sucking down mobile data like it’s water

We may be pushing more of our cellular activity over to Wi-Fi, but we’re still guzzling mobile data like it’s going out of style, according to the latest estimates from Cisco’s Mobile Visual Networking Index. The telecommunications equipment vendor puts out its Visual Networking Index twice a year, once for wireline and once for mobile network traffic — and estimates how much we’ll see in the coming five years.

And when it comes to cellular, the entire world is going to go from consuming about 2.5 exabytes a month in 2014 to 25 exabytes a month in 2019, with a large portion of that growth coming from new device users in developing countries in Latin America, China and the Middle East. Just for comparison’s sake, an exabyte is a billion gigabytes. My cellular plan lets me have 5GB a month.


That’s a lot of cellular activity and some of that will be spread among 2G, 3G and 4G connections according to the folks at Cisco. But what’s more notable is that the individual data usage will increase so much — from almost 2GB used per month in 2014 in North American to almost 11 GB — brought about in part by adding more devices to the network. Think about not only traditional tablets and laptops, but also cars and connected gadgets, such as backup connections for a home hub or a connected medical device.


Thomas Barnett Jr., director of Cisco’s VNI program, expects there to be changes in carrier pricing to go with this increase in data, but it will most likely be in the guise of new shared device plans, not necessarily in more generous gigabit allotments given directly. It’s worth noting here that these numbers do not include Wi-Fi offload, which worldwide takes about 44 percent of the traffic off the network, according to Barnett. In the U.S., that number is about 66 percent.

The big drivers behind this growth won’t surprise many: People coming online and more devices. In developing countries, people coming online and the rise of smartphones in the hands of those people will drive much of the traffic growth in those countries. In the Asia-Pacific region, shown in the above chart and the one below, it’s worth noting that the averages are a bit skewed because rural China behaves like a developing country dragging down the averages of cities in South Korea and Japan where connectivity is oftentimes better than anywhere else in the world.

In North America and Western Europe, the reason for traffic growth will come from more video consumption, but also more devices coming online in the form of the internet of things. The chart below offers a pretty involved take at each country’s drivers if you have the patience to study it.


So let’s break down one of the macro trends driving traffic — the internet of things. Cisco has taken a closer look at both wearables and how the rise of newer, low power wide area networks (it calls these LPWA) like the Weightless or Sigfox networks might affect traffic. It thinks that those networks have a possible advantage for certain types of traffic, especially in western Europe where they seem to be taking off, but they don’t seem to be taking a significant burden off the existing cellular networks anytime soon.

That’s because there are many costs associated with transitioning from one network to another, and it’s not a move to be made lightly. If you have a 3G module in the field, it will likely remain there until it dies on its own or your carrier kills the 3G network.

This chart represents all of the device traffic including M2M across networks.

This chart represents all of the device traffic including M2M across networks.

This chart represents only M2M traffic  across networks. You can see that even among M2M 3G is winning through 2019.

This chart represents only M2M traffic across networks. You can see that even among M2M 3G is winning through 2019.

Cisco also broke down some trends in the wearables market, and estimated that a GoPro camera connected to a cellular network running for 2 hours would generate 600 MB of data. It did a similar case study for Google Glass a year or two ago just to send shivers down the spines of its telco customers, but the point is not all that crazy. People may not shoot video on cellular networks regularly, but video downloads still constitute a huge percentage of cellular traffic and video uploads are not terribly uncommon, especially at big events.

Meanwhile Cisco estimated that currently only two million wearables were connected to the cellular network in 2014 and estimated that number would only increase to about 42 million by 2019, accounting for just a small amount of the overall number of devices and traffic on networks. But it estimated that if those devices were medical they might need some service guarantees or require dedicated networks.

Finally, we’ll end with a fun chart that shows where the internet of things might fit in with all the other devices that connect to the cellular networks. I only wish Cisco had added cars to this chart, but it’s still fun to see that my smart watch might generate more data than a flip phone.


Comcast’s next-generation Xi4 set-top box passes the FCC

It looks like Comcast is getting ready to ship its next-generation set-top box to consumers: A filing for the company’s Xi4 set-top box just popped up in the FCC’s database. The filing, which has been approved by the FCC, is heavily redacted but it does give us some clues as to what the new box is all about.

Comcast's next-generation TV sett-op box, as shown in an FCC filing.

Comcast’s next-generation TV set-top box, as shown in an FCC filing.

The new set-top box is being made by Pegatron on behalf of Cisco, which is notable, because its Xi3 predecessor has been made by Pace. The device goes by a range of model numbers, including Cisco CX041AE1 as well as Cisco ATV8100, but it looks like Comcast will be calling the box Xi4v1-C. The box includes support for 2.4Ghz Wifi and Bluetooth, which is likely being used for the remote control. A photo of the FCC label shows a square device, much like what Comcast has been showing off for some time as its next-generation X1 set-top box.

Interesting about these new boxes is that they likely won’t contain a local hard drive anymore. Instead, Comcast is now storing all of it’s customers’ TV recordings in the cloud as part of its cloud DVR.

Comcast has been gradually expanding the footprint of its cloud DVR service over the last several months, and brought it to San Francisco last summer. At the time, Comcast executives told me that the company could eventually offer unlimited storage for TV recordings in the cloud, and possibly even ditch set-top boxes altogether.

Key IoT takeaways from Cisco Live

I spent last week at Cisco’s Executive Symposium, which included just over a hundred CIOs from around the world exploring how the Internet of Everything was going to change every sector, ranging from financial services to mining to media. (IoE is Cisco’s take on the Internet of Things, and likely an attempt to capture the tech phenomenon under their own branding).

The keynote address from Cisco CEO John Chambers included some strong predictions on the future of big IT, namely that Chambers believes the consolidation of big IT will only get worse with just two of the top five IT giants being relevant in five years. This has implications for IoT because Cisco is betting hard that being able to offer end to end architecture under one brand will be key to making it easy for companies considering IoT connectivity to simply establish such connectivity. Systems integrators were definitely under fire.

IoT and the companies that play key roles in its deployment are important for cleantech. From the smart grid to the smart home, it’s now becoming clear that some of the most promising innovation opportunities will be in leveraging IoT for energy and resource efficiency. The game changing energy opportunities like solar and next generation batteries remain critical and need more R&D. But for the startup or entrepreneur the access point may be leveraging connectivity and data to encourage sharing or improve efficiency.

Some key takeaways from the conference:

1) Connectivity isn’t the end goal. Chambers echoed it when said “IoE isn’t about connectivity.” Connectivity is just the tool to capture data. And the data isn’t the end goal either. The end goal is business intelligence and you get there by applying analytics to data to solve problems. The value is in the service.

2) Start simple. Alex Hawkinson from SmartThings, which sells a hub to connect devices in the home and has rolled out an open API, noted to CIOs looking at developing IoT strategies that they should start simply. Sometimes figuring out complex projects that involve everything from sensors to security to analytics is a challenges. Do the simplest thing that adds value and use that as a springboard.

3) Consider all the available data. The Weather Company’s CIO Bryson Koehler said that when he considers what The Weather Channel can do with IoT he doesn’t just think about the data his company has direct access to. Rather what gets him excited is potentially drawing correlations between lots of different data, some of which may not necessarily be generated internally. Like mining public weather information or geological information to improve forecasting.

4) Dealing with the signal to noise ratio. One challenge that surfaced for many folks who have worked in IoT revolved around dealing with the volume of data and making sense of what’s useful. Otherwise the risk is getting bogged down with a sea of data sets and not knowing what to do with it all. In some respects this issue relates to hiring and having the right data scientist experts available to manage and glean real insight from data.

Chambers opened his keynote by citing the Cisco projection that mobile data traffic would reach 190 exabytes by 2018. I mention this figure because it relates to a theme that Google’s Director of Engineering Ray Kurzweil spoke to at the symposium: the difference between linear and exponential change.

When it comes to data growth and processing cost, we continue to see exponential movement. Which suggests that whatever we believe the tech landscape will be like in 3 years, we’re likely underestimating the magnitude of change. That’s particularly true for data volumes and IoT.

And for entrepreneurs interested in developing products and services for everything from cleantech to cloud, the opportunity remains in finding valuables services that can be built on top of all that data growth.


Cisco memo: We can’t build anything

Cisco in an internal memo outlined its plans for the changing nature of networking. It also acknowledged a $100 million investment in Insieme, a company started by three Cisco executives and that it can buy it for upto $750 million. Read the memo & what it means.