The Amazon of IoT
Kenta Yasukawa Co-Founder and CTO, SORACOM, INC.
“When Amazon web services came onto the market, it stimulated cloud innovations in many industries. Today, AWS powers many major brands including Netflix, Spotify, Airbnb and Lyft. We’d like to become the Amazon equivalent for IoT.”
Interview by Abraham Joseph, Editorial Director, IOT Insights – 1 August 2017
What does SORACOM do?
We connect things to the cloud. Some people think IoT is about connecting things to the Internet, but essentially it is about connecting things to cloud services and making use of the intelligence implemented in the cloud. We provide the services to help developers do that.
What IoT developers want to do is to create networks of their servers and their devices and implement intelligence based on the information collected from their devices.
Who are the other co-founders of SORACOM?
Our CEO Ken Tamagawa and our COO Daichi Funato. Ken and I worked together previously. He ran the technology side of Amazon’s business in Japan, and I was on his team.
What was your role before co-founding SORACOM?
I was an Amazon Web Services (AWS) solutions architect. My role was to provide technical help for customers that were using AWS to develop their infrastructure.
I worked with a wide variety of clients including technology startups, gaming companies and media companies. That experience gave me the ability to think about cloud-native architectures for different systems and applications.
What was the initial spark that led to SORACOM?
Before joining Amazon, I worked at Ericsson. In 2010 to 2011 cloud computing was just emerging and I thought that telecom infrastructure could be implemented in the cloud.
After I tried AWS, I continued to believe that it should be possible. However, I realised that I needed to extend my knowledge of how to use the cloud, so I decided to join AWS as a solutions architect.
One night I was having a beer with Ken, and I mentioned the idea of implementing a mobile core network on AWS. The following morning, Ken came up with the idea of connecting things and people to improve lives and make the world a better place to live. That was when we decided: let’s do this. Shortly after, we teamed up with Dan. Dan previously worked for NTT DOCOMO and had a strong telecom background. Together, we had the fundamental elements for the foundation of the company.
The time you mentioned, 2010 to 2011 was that not the time of network architecture evolution to IP and IP Multimedia Subsystems (IMS)?
Yes. That was exactly the time we were discussing IMS. At Ericsson, we talked a lot about IMS, and there were lots of research projects about IMS. But to me, it looked like IMS was more about replacing old switches with IP-based switches rather than about replacing switches with web technologies.
Implementing a telecom stack on AWS sounds hard. How was all that heavy lifting done?
We started as an MVNO of NTT DOCOMO. When we launched our first service, we needed a GGSN or PDN Gateway to terminate the cellular network at the edge. Then, I started to look at the protocol stack to figure out how it could be implemented in the cloud. Ken, Dan and I spent some time discussing this, and together we designed the architecture. Subsequently, we implemented the first proof of concept.
How long did that take?
It took about 6 to 8 months to develop a solution that accepted ‘live’ customer traffic. At that time, we were operating in stealth mode. We offered alpha and beta versions of the service to some friendly customers, got feedback, improved the offering and prepared for a public launch. Thanks to the capabilities that AWS offered, it was not as hard as it might sound.
From my ancient telecom background, I imagine reams of specifications, years of development and millions of lines of code. How easy did you find it?
It is true that it was not simple. We needed to deal with telecom protocols, the Internet world and web technologies. Thanks to our backgrounds, we had insights into each aspect. We combined forces to make it possible. Additionally, we leveraged AWS managed services such as Amazon DynamoDB, Amazon S3 and Amazon SQS that gave us tremendous capabilities from the onset. The ‘super power’ given by these AWS services enabled us to focus our resources on the core parts of our platform.
It sounds like you did a lot of the coding yourself. Was that the case?
Yes, I wrote quite a lot of code. Although I have the title CTO, I introduce myself as a developer and CTO. I’m still a part of our development team, and I still do a lot of hands-on coding. I find it the best way to engage with the team.
How handy was your telecom vendor background?
Sometimes I get misunderstood by people. They think that since I worked for Ericsson, I had knowledge of the internal aspects of the switches and equipment. This is not the case. I was a researcher at Ericsson, working on connected home and application area research. To develop our platform, I had to dig into and learn many technical aspects.
Fortunately, at SORACOM we had a dream team for developing systems based on AWS cloud. This enabled me to focus solely on developing telecom interfaces.
The development of our billing system that enables detailed billing for pay-as-you-go use is an example of this good fortune. Fairly early in the development of the platform, I said to a colleague, Akio Katayama, now VP of Engineering, “Periodically, I’ll send an archive containing the amount of data each SIM has consumed to an S3 bucket. Can you build a billing system?” The outcome is the real-time billing system that is now at the heart of the SORACOM platform.
What was the most challenging part of the technical effort?
We spent a lot of time figuring out how to develop a scalable and reliable infrastructure. We needed to implement the GGSN and PDN Gateway so that they scale as we increase the number of servers. Also, we wanted to have no single point of failure in the system.
In the web world, there are lots of existing frameworks and programming languages we can use for building common web servers and reliable, scalable systems. For telecoms, there are not as many interesting tools and insights. We needed to design new architectures and invent cloud-native telecom gateways. That was a challenge, but it is a good challenge because it enabled us to develop and patent important intellectual property.
Unlike the traditional ways of implementing GGSNs and PDN Gateways using ultra-reliable hardware, our way is the cloud way, using EC2 instances running on commodity hardware.
Any server in our system can fail at any time without affecting the service, and when an instance fails, the system automatically heals itself by replacing the node.
How have you tackled security?
Security has been the most important consideration in our architecture. The fact that we have our GGSN and PDN Gateway running on AWS not only contributes to infrastructure security but also it enables our value-added services to improve security for customers. For example, if the customer’s back end is also on AWS, we can peer the customer’s virtual private cloud with our core network and the customer can implement an end-to-end virtual private network by taking advantage of our connectivity solution. Also, we have application layer services that provide additional security for device communications. For example, we have a service called SORACOM Beam that can receive data from a customer by opening an endpoint at which we apply Internet-grade encryption to the data. If we receive a single UDP packet from the device, we extract the payload and forward it with HTTPS to the cloud side. This approach has added the benefit of improving communication efficiency by reducing overheads on cellular links.
We hear a lot about devices being hacked. How does your solution solve this problem?
Typically, manufacturers ship devices that connect to cloud back ends through the public Internet. In many cases, the devices use home Wi-Fi to connect to servers via the public Internet, and there are several points where an attacker may steal or alter information.
When a customer uses SORACOM Air and services built on it, their devices are connected to the cellular network and data is transferred through the local operator to SORACOM’s back end through the inter-operator network. Also, as I mentioned earlier, if the customer is also running on AWS infrastructure, we can pair with the customer’s back end through Amazon VPC peering, a private fibre or a virtual private network and data does not go through the Internet.
Does SORACOM offer a solution for managing devices?
We provide management at several levels. Customers have always had the ability to view and manage their SIMs through our console. Often the SIM card is tied to a device ID and in the past customers managed devices through their SIMs via the console.
Recently we launched a new service, SORACOM Inventory, to allow customers to manage their devices as well as the communications service. SORACOM Inventory makes it possible to change settings on, restart, issue commands to and monitor various metrics on an individual device or group of devices remotely and securely. It leverages SIM identity to ease the provisioning of credentials to each device and enables device management with low protocol overheads.
Have you seen any changes in the types of devices people are trying to manage?
Yes. Recently, we have been seeing requests for the management of less capable devices like tiny controllers and sensors that are not running a fully-fledged OS. When we designed SORACOM Inventory, we considered those restricted devices.
What other features are we likely to see from SORACOM in the coming months?
We are always in listening mode and continuously collecting requirements from customers. When these lead to a common feature we can offer, we consider implementing it. Our roadmap is determined by what we learn from customers.
You wish to democratise IoT by liberating developers. How much experience does a developer need to have to use SORACOM’s services?
Our services aim to get rid of the barriers that developers usually face when implementing communications between devices and servers. For example, once a developer has a SORACOM SIM and a 3G/LTE module connected to their device, the knowledge they need is a basic use of familiar protocols such as HTTP or MQTT. Our application layer services provide additional security enhancements and help integrate with cloud services so that devices can use simple, low-overhead protocols. Developers can send data to our HTTP endpoint, and we can apply SSL and the additional authentication. Also, if they use SORACOM Air and set the private networking features properly, they don’t have to be afraid of going through the Internet. This means that developers do not have to waste time on the typical heavy lifting associated with using most other IoT connectivity solutions.
What type of developers are you targeting?
Any developer is welcome. We have a wide spectrum of developers, including those working with startups, SMEs and large enterprises like Canon, Komatsu and Toyota.
We designed our platform so that any size of customer can leverage our services. Our self-service model is liked by startups and SMEs. At the same time, the features we offer to build secure networks are of great importance to large enterprises.
What are the most important things you’d like people to know about SORACOM’s solutions?
There are three things. The first is that our services are meant for IoT innovators – people who want to leverage the key technologies available today, like cloud, secure connectivity and new devices, and focus on innovation rather than the complexities of managing communication protocols and implementing security layers. We handle all the heavy lifting, leaving them to focus on their devices and their applications.
The second is that we offer a self-service model. Customers can order a SIM card online and get started at any time. They do not need to call us or fill in an application form.
The third is that our pricing is based on a simple pay-as-you-go model. Customers can implement quick prototypes and grow them by adding features.
When Amazon web services came onto the market, it stimulated cloud innovations in many industries. Today, AWS powers many major brands including Netflix, Spotify, Airbnb and Lyft. We’d like to become the Amazon equivalent for IoT.
Doesn’t that mean having to fight off many of the big players like Amazon and Microsoft?
We have good relationships with all the mega-cloud vendors. Although I said we would like to be the Amazon for IoT, it doesn’t mean we plan to compete with the mega-cloud vendors. Rather, we would like to extend their service capabilities to IoT devices. For example, we have an integration with Amazon’s AWS IoT that enables devices to exploit AWS’s capabilities simply by configuring some parameters on the platform rather than having to write code. Similarly, we integrate with Microsoft Azure event hubs so that developers can use machine learning or the analytics services offered by Microsoft Azure. Recently, we announced integration with Google IoT and Google Cloud Pub/Sub.
What are your proudest customer case studies?
The first would be Komatsu. They have huge construction machines in the field and have been players in M2M for a long time. They have had cellular modules in their machines and have been monitoring them remotely. They had a different provider but switched to using our service because we offer cloud integration as well as secure, bidirectional communication between their machines and back end services.
The second example I’m especially proud of is Tokachi Bus, a local bus company in Hokkaido Island, Japan. They started by putting a GPS locator on each bus and tracking bus locations in real time. Then they offered a smartphone application to bus users. The reason I chose this example is that they found us shortly after we launched our service, ordered the SIMs online, built the system and started offering services to their customers. We didn’t know about them until they launched the service. They are a typical self-service customer, and their case shows how easy it is to use the service and how easy it is to expand a solution built on our service to support additional customers.
What are your greatest hopes for the development of IoT?
My greatest hope relates to the vision Ken expressed in our initial discussion that led to the founding of the SORACOM. That IoT will be a major force for good in the world, that it will help to improve lives and help us build better, safer communities. Also, since there is always the opportunity for technology to be misused, I hope that we will be wise enough to develop and deploy IoT in ways that minimise potential abuses.