LINX recently introduced the option of 400GE ports for members in partnership with its vendor partner, Nokia. In this blog LINX Enterprise Architect, Mariano Juliá, looks back over the development process to get to where we are today.
Research and Key Objectives
In March 2020, we set out to research the market options to offer 400GE to our members connected to LON1. The key objectives were to find a long-term high-density 400GE solution that could integrate with both our (then) VPLS network and future EVPN topology, plus our automation and management tools.
Pandemic Delays the Testing Process
The situation outside of LINX at this time was uncertain due to the Coronavirus pandemic. From the middle of March LINX staff began working from home, but just two weeks later, everyone was in lockdown.
The restrictions meant we had already postponed a proof-of-concept test to verify a hitless migration to EVPN on the Juniper estate. Indeed, it wasn’t evident when we could come out to visit vendor facilities, or our own, for a full-blown proof of concept. The world slowed down for a while.
After two months of video calls and remote presentations, we had, at least on paper, a clearer picture of the landscape.
Decision Time
Juniper did not, and would not, have a high-density 400GE card for our main MX960 platform. The alternatives were to use the PTX or QFX switches. However, LINX was in the finishing stages of replacing the PTXs with the MX10k, and the QFX was not seen as a long-term strategic solution. The high-density 400GE card (12 or 24 ports) for the MX10k was still two years away.
Our LON2 vendors, Edgecore and IP Infusion, weren’t an alternative for LON1 either. First, it would have meant breaking the long-established design rule of having redundant vendors in the two London LANs. In any case, Edgecore did not have a high-density 400GE platform that supported MPLS natively in the chipset at the same time.
Adding to that, IP Infusion’s OcNOS hadn’t been ported to any of the 400GE platforms, and their software integration cycles are at least six months for existing platforms and 12 to 18 months for the new ones.
New Vendor Options
In looking at other vendors that LINX didn’t work with at the time, Arista offered the highest port density and the best automation and telemetry capabilities. However, they didn’t support Ethernet OAM or VPLS, and LON1 required both features.
Nokia, meanwhile, supported all the required features at the outset, had good port density, and was the most competitive. It is also in use at FranceIX, which had recently integrated Nokia in a mixed VPLS environment with Juniper EX, and DE-CIX, which run a large 7750 SR deployment.
Next Steps: Testing and Re-Evaluation
As the summer of 2020 began, it was decided to put our 400GE general deployment plans on hold until the start of 2021. At the same time, we committed to test the capabilities of the Nokia 7750 SR-2s that autumn, pre-empting any early orders.
After completing the SR-2s testing ourselves and verifying the hitless migration from VPLS to EVPN with Juniper, we started the first quarter of 2021 migrating LON1 to EVPN.
We then re-evaluated our options for 400GE. One of the major challenges of 2021 was supply chain issues and how difficult it was to acquire equipment due to the shortages in the silicon market.
This led to testing of the Juniper MPC-10 cards for the MX960 for low-density sites, as they feature a combination of two or three 400GE ports, and a combination of up to ten or fifteen 100GE ports, that could be repurposed.
By now we were confident that the Nokia 7750 SR-7s could provide better-long term scalability for our use case and decided to do a full integration PoC centred around EVPN.
The Closing Stages and Roll Out
In August, we received the kit from Nokia in our lab, and after some weeks of configuring and building the topology, we started and completed the test in the month of September. This PoC was followed by a week of testing the MPC10 in Juniper labs and a month and a half of deployment of the SR-7s in LON1 to fulfil the first orders.
As a result, 400GE is available for ordering in Telehouse North, from where we can serve the entire Telehouse campus, and available in Equinix Harbour Exchange.
Other sites will be considered depending on member demand and feedback, with Equinix Slough and Interxion Brick Lane the most likely candidates.
Future Plans
Our future plans will centre around supporting 400GE in the core and the edge in LON1, testing and deploying 400GE in the Juniper MX10k platform and looking at the 400G transport to support traffic growth in our core especially growth in West London.
As for LON2 and the exchange networks based on our disaggregated Edgecore IP Infusion model (LON2, Manchester, NoVA), enable 400GE in the edge and the core.
If you would like to learn more about the integration of 400GE at LINX, and how your network can benefit, please contact our Member Relations team. Alternatively, please view Mariano’s presentation at the recent LINX114 member meeting.
< Go Back