Driven by new applications requiring more and more bandwidth, faster response times, and lower latency to deliver improved quality of user experience, Telecom, Data Center and overall IT infrastructure is evolving closer to the “edge”, moving into highly specialized and optimized (space, power, cooling) facilities. One of these applications is 5G, with a reference architecture “disaggregated” into multiple components, one of the most important being UPF (User Plane Function).
How much more bandwidth? If you consider what Verizon anticipates (in their published “currencies of 5G”), one of these Edge Data Centers, could handle multiple Terabits per second of traffic (considering not only mid-range frequencies, but the new millimeter waves with massive 32T32R/64T64R MIMO antennas), while requiring very low end to end latency (~5ms). Based on these considerations, any traditional approach would not be capable of providing not only the required performance characteristics, but the right business/financial model (as cost per Gb/s or now cost per Tb/s).
Meeting these requirements led Kaloom on a different path, capable of not only addressing the specific technical challenges – high BW, low latency – but the strong need for new economics, leveraging a complete end to end Intel HW based approach (Intel/BXD “Tofino” as Edgecore Wedge100 32QS, Intel FPGA “Stratix 10”, Intel x86) with P4 as the most flexible language to program these targets on a regular basis, integrating and assuring protocol evolutions and specific customer requirements. Associated to this innovative and differentiated HW architecture, Kaloom offers a fully containerized solution, leveraging Red Hat OpenShift not only as the foundation for this edge fabric, but to host customer workloads using the same “cluster” further reducing the footprint and overall operational complexity and cost”.