What We Learned: Next Generation Infrastructure Packet and Orchestration Partner Presentations
During the week of May 1, Internet2 staff and community members met near San Jose, California, to hear presentations by potential partners in response to the call for information regarding the Next Generation Infrastructure (NGI) packet and orchestration plans. Eight potential partners responded, including: Anuta, Arista, Ciena, Cisco, Cloudify, Itential, Juniper, and Nokia. Over the course of two days, we participated in seven presentations that lasted more than two hours each. The experience was enlightening on several fronts and also confirmed some planning direction that have already been in consideration.
Some highlights to report back on:
2019/2020 Arrival of 400G Routers/Switches – The hardware manufacturers all plan small (2” high but very deep) 400G packet transport routers that will arrive in late 2019 or early 2020. These devices will have 24-48 ports in combinations of 100G and 400G and are expected to support internet-scale route tables. While these boxes will still require large power feeds, it looks like they will have a substantial (4X+) improvement in bits per watt and will allow consolidation of expensive colocation throughout the Internet2 Network. The availability of affordable 400G solutions are important for a number of connectors and research collaborators who want substantial research headroom above regular traffic levels that are beginning to fill their 100G links.
Pluggable 100G-400G DWDM Optics – Motivated by cloud-scale requirements, many vendors are accelerating their investments in coherent DWDM pluggable optics that can interoperate with line systems similar to the one Internet2 will be deploying. The promised benefits include cost and space improvements and reduced complexity. While CFP2 200G DWDM optics have been prevalent on the market for some time, they will be joined by coherent 400G QSFP-DD and OSFP modules with enhanced DSPs in late 2020 or early 2021; this is expected to increase the achievable speeds and distances within more compact packet transport platforms. Assuming the developing standards hold these may have some point-to-point applications within metropolitan areas, however their low launch powers may limit their usefulness in 400G backbone segments. This helps solidify our short-term strategy to use sled-based transponders for the next several years.
Segment Routing – Each of the packet vendors who could readily support it talked about industry migrations to segment routing (SR), currently based on the MPLS label-enabled implementation as the emerging best practice to deliver routed services. For the Internet2 community, SR promises greater control and overall simplification in support of individual research flows, as well as the ability to service chain security, performance, caching, and other support resources in to individual flows without changing the underlying routed infrastructure. The big advancement of SR is the ability to encode routing intent in the packet instead of the router. This means that a packet can be crafted to follow a specific path separate from the other packets routed through the same device. The benefits for security, large research flows, and other traffic management is significant.
Ethernet Virtual Private Network (EVPN) – All of the packet vendors who could readily support it also confirmed their recommendation that EVPN is the natural evolution from VLAN and MPLS-VPN technologies. For Internet2, EVPN may simplify configuration and better position for inter-domain automation and orchestration with regional partners. There is more work to do to nail down specific configurations, appropriate use cases, and transitional strategies for Internet2 related to EVPN.
Orchestration Software – We heard updates on the “state of the market” for tools that can help automate and manage the lifecycle of common network services and tasks across an enterprise. Many of these tools appear to have evolved significantly since our last review a year ago. We were impressed with some of the proposals for service-delivery and service-management capabilities that the products promised, including some network management and cloud provisioning capabilities that could benefit the community. One area that requires more development is how to conceptualize the use of these toolsets in the federation of networks and resources like the U.S. R&E community, and how to enable the capabilities these tools present to the end-to-end use case.
The team of staff and community leaders are now busy working through the information gleaned from these presentations and turning them into concrete plans for the NGI packet implementation phase. We are looking forward to seeing many of you at the Community Architect’s workshop in July 2019 to hear feedback on these draft plans. An optical deployment update as well as plans for segment routing, EVPN, and topology implications will all be a focus of the community discussions that will happen at the July workshop.
Rob Vietzke, Vice President of Network Services
Scott Taylor, Optical Network Platform Manager and Architect
Chris Wilkinson, Director of Network Engineering and Planning
Karl Newell, Cyberinfrastructure Security Engineer