In service provider networking infrastructure there are typically at least two sets of networking gear for handling the IP packet and optical core routing and switching requirements.
Juniper Networks (NYSE:JNPR) aims to help converge those layers with a new Converged Supercore architecture. The Converged Supercore includes the new PTX Series Packet Transport switch, which can scale to 3,800 terabits of switching capacity. At the heart of the PTX is new silicon from Juniper called Junos Express, which has been built specifically for networking transport and has 3.55 billion transistors. The new Converged Supercore initiative comes as Juniper is growing the scalability of its networking efforts with QFabric for data center deployments.
Luc Ceuppens, VP of Product Marketing at Juniper explained to InternetNews.com that when you look at service provide networks there are three overlays: There is an optical transport network, a circuit switched network and then an IP network on top of that. All those networks are very difficult to provision and are often over-provisioned in order to deal with potential bursts in demand.
“We created a new ASIC called Junos Express that is designed for packet transport, it’s a huge MPLS switching fabric that scales to big numbers and allows service providers to optimize the core of their network for packet transport,” Ceuppens said.
The ASIC is combined into the PTX switch that includes an optical transport capability that enables service providers to merge their optical and circuit networks into a more cost effective network.
“We’re introducing an MPLS switch and we’re announcing that we’re going into the optical transport space 100 percent,” Ceuppens said. “We’re no longer just doing optical interfaces; we’re now doing full end-to-end networks and network management with all the provisioning that it takes.”
The idea of trying to converge the optical and IP networks is one that Juniper first began talking about in 2010. Rival networking vendor Alcatel-Lucent has also been talking about initiatives to converged networking transport as well.
The other key item for Juniper is the fact that new Converged Supercore can deliver what Ceuppens referred to as massively scalable architecture. The total system scale of a PTX switch is 3,800 terabits per second
The 3,8000 terabits per second (Tbps) of capacity don’t come in a single switch chassis, but rather by way of clustering PTX switches together. Digging into the performance numbers, the PTX will start at 480 Gigabits per second of performance per slot and will expand up to 2 Tbps per slot.
Ceuppens explained that there are two initial versions of the PTX chassis. One of them is an 8-slot version, which at 480 Gbps of slot capacity will deliver 8 Tbps of capacity. There is also a 16-slot version which, at 480 Gbps per slot, will deliver up to 16 Tbps of capacity in a single chassis. When the per-slot capacity is raised to 2 Terabits a slot the overall capacity of each chassis goes up by 4x.
“If you want to expand it more you have to go to a multi-chassis configuration,” Ceuppens said.
Juniper already has a multi-chassis configuration called the TX Matrix for its T-series core routers. The TX Matrix enables multiple Juniper T1600 core routers and could be combined to deliver up to 25 Tbps. Ceuppen snoted that Juniper will be introducing a similar Matrix-type cluster for the PTX at a later date.
With the new Converged Supercore, Juniper will also help to enable their QFabric infrastructure for data center scalability. QFabric was announced in February as a way of collapsing layers within enterprise networks.
“QFabric brings massive scalability with low-latency into the data center,” Ceuppens Where the PTX comes in, is the interconnection of data centers over a wider area and creating big pipes between data centers.”