GradientGraph™ Analytics is a new optimization framework that provides high-speed network administrators and operators with a new analytical platform to analyze, understand and act upon distributed bottlenecks and flow performance. Based on a recent mathematical breakthrough on the understanding of the bottleneck structure of data networks, GradientGraph provides a new network model and advanced algorithms capable of characterizing with high precision the performance of bottlenecks and flows.
Due to the inherent topological structure of each data network, not all bottlenecks are of equal importance. The GradientGraph is able to capture the fundamental bottleneck structure of each network, revealing key operational insights such as the regions of influence of bottlenecks and the ripple effects of small perturbations in the network. It also provides a novel analytical framework to quantify such effects and make bottleneck-optimized traffic engineering decisions. Because of the dualism between bottlenecks and flows, the same analytical framework provides a powerful platform to optimize flow performance. This includes the identification of flows that, while they may not be among the top heavy hitters in the traditional sense, they lead to severe system-wide performance degradation due to their traversing of high impact hotspot regions revealed by the GradientGraph.
GradientGraph provides a new paradigm to qualitatively and quantitatively understand distributed bottlenecks and flows, enabling real-time traffic engineering, high-performance baselining, and a framework for high-precision network capacity planning.
mCore – Packet Path Accelerator
As the capacity of network links and multicore CPU/GPU processors continue to increase, a new bottleneck located at the network interface is arising that prevents performing network security at scale. Today a single fiber link can carry about 100 terabits per second of data traffic, while state-of-the-art multicore CPU/GPU processors can aggregately scale at a similar rate. Yet traditional network stacks implemented by the operating system are limited to rates up to 100s of gigabits per second. Further, as computational parallelism increases, this bottleneck has dramatically worsened.
This situation has been recognized by both the scientific community as well as the industry. For instance, state-of-the-art artificial intelligent applications using massive amounts of data are finding the network interfaces a major obstacle to scale. Similarly, companies that manage very large-scale distributed systems require parsing packets at the rates of multiple terabits per second in order to secure and identify potential cyber attacks.
To address this gap, Reservoir Labs has designed and patented mCore, a suite of cybersecurity-aware data structures and algorithms addressed to resolve the network interface bottleneck.
Experimental benchmarks demonstrate that at input rates of 10Gbps, aggregately these optimizations increase application performance up to 500% while packet drops are reduced up to 200%.