A brand new way of building democratized networks
This article was published in CIOL on April 30, 2020. You can read the original article here
Pravin S Bhandarkar, Founder and CEO of RtBrick, talks about bringing their software solution to the legacy Indian telecom networks to cope with the ever increasing appetite of data which services like OTT have generated.
We have developed a new way of building networks, using cloud and Internet native principles in networking. The networking industry is going through an evolution in which there is desegregation: Hardware can be bought from one location and software from another, much like the mobile Android ecosystem where you can buy hardware and software separately. We focus on access, meaning the network piece that is used to connect from the Internet to home and small office users.
There’s a lot of traffic generated by OTTs and something like AWS for business applications. The networks are engineered in a specific way that they can scale up horizontally, as the content increases. On the other side are service providers, which are becoming big pipes. They don’t have their top line growing, but then they have to invest for the customer experience in growing their networks. This growth is asymmetric, they use incumbent or integrated systems. We say that is instead of using integrated systems, let us build a small grid for you, at every location. This is like taking cloud in infrastructure down to central offices, and you can build as much as you want.
So you buy one rack unit, you put our software on it. Let’s say you’re servicing 300 customers, our price point presents a viable business. You go to 5000 or 10,000, just keep adding units. Our software takes care of managing all these units, connecting it, giving you redundancy, all the capabilities that an integrated services unit. Your price point tremendously goes down from a capex perspective. Then because of software automation, we use tools like DevOps. We leverage the Linux tool chain. What we are bringing is Internet native technology to the telcos.
We are calling it distributed SDN. SDN still has a centralised controller. We firmly believe that networks have to be distributed, because data networks were built to overcome nuclear attacks. It’s a packet based network. So we think it’s going to be hybrid. At certain points you want distribution elements like redundancy, resiliency and grid-like behaviour. But when you want to deliver services, you can do it from a centralised location.
Sitting on the access layer
A network has the core, where traffic is carried from point A to point B. Then there’s the Edge, where the services are deployed. Finally there’s the access, where they connect to users. So we are in the access network and you can connect to public, private or hybrid clouds. We reduce costs for the providers of this access services right. One of the drivers is that networks are moving from older copper based systems to optic fibre, which is coming to the home. We simply build software on top of the new systems.
The biggest advantage is you can turn on services fast. If they don’t work turn it off. We want to democratise the network: Give you control of what you’re doing and you should know how to use it best. We give you the tools that will enable you.
Moving disruption to the core
On the Edge there are pressures of delivery of services and enhanced customer user experience. So that and access is where we start to see disruption. When we started off, we just went with the place where we were going to earn revenue. We said if we are doing something of value someone is paying us means we are doing it right, rather than look at what is the coolest problem to solve. We think it will start from access/Edge and eventually move to the core. That’s how the network will change. We are probably the slowest 4G network in the world. One of the reasons for that is that our ability to manage the congestion as well as user experience is a challenge. Our solution from day one has built in Hierarchical Quality of Service (HQOS) which manages to prioritise.
Optimised composable software
The integrated systems built monolithic software that worked from 1999 to 2010.After that we have the cloud and Software Defined Networking from a networking standpoint. A lot of components are now available in the market either in open source or through GitHub communities. But there is a limited readiness, at least in the networking world, to adopt and that is because there is some cost associated with maintaining businesses. So we built from ground up. We factored in a million transactions per second. We have composable software. If a have a particular protocol, you can just run that one protocol. It is just plug and play. You can use two protocols or stick with a keepalive protocol. That reduces your memory footprint, that reduces your execution footprint and that reduces the size of the package also. So it is really optimised.
On the Internet of Things
One, there is an explosion of IoT devices. So how do we address those devices? Two, there are a lot of proprietary technologies with which this has started off. Right now worldwide it is all getting unified on IP, we think it will be IPv6-enabled, because it allows you addressability. IPv6 has a huge space so they can address trillions of devices. It depends on what kind of services you want. Let’s say you have a windmill farm with IoT devices to measure certain parameters. They will generate a lot of data, which will eventually become Big Data. Some pre-compute is needed at an intermediate point, which is the IoT-Edge compute. Then you go back to back end to process this data and use business analytics.