Build Systems that Drive Business
June 11–12, 2018: Training
June 12–14, 2018: Tutorials & Conference
San Jose, CA

Steering the Edgecast CDN with Heteractis

Marcel Flores (Verizon Digital Media Services)
2:10pm–2:50pm Wednesday, June 13, 2018
Hardware, Storage, and Datacenters, Systems Engineering & Architecture
 Location: LL21 C/D Level: Intermediate
Secondary topics: Systems Architecture & Infrastructure
Average rating: *****
(5.00, 1 rating)
Download slides (PDF)

Prerequisite knowledge

  • Familiarity with internet protocols and limitations (TCP/IP, DNS, anycast routing, etc.)

What you'll learn

  • Explore Heteractis, the traffic management system Verizon Digital Media Services uses to turn network telemetry data into automated decisions in an automated fashion

Description

Heteractis is the automated traffic management system that was built by the Verizon Digital Media Services traffic management and research teams. The system has been live and actively managing Tbps of live traffic for the CDN for over a year now.

Marcel Flores explores the design and implementation details of the Heteractis system. Marcel offers an overview of the underlying logic that drives the system and the actions that it takes, including high-level descriptions of traffic management techniques and approaches as well as some of the nitty-gritty details associated with managing CDN traffic in particular. Marcel then explains how Heteractis collects and stores information about what it sees on the network and examines the decision-making capabilities of Heteractis, demonstrating how it uses the available data to consider what-if scenarios. Marcel also covers Heteractis’s scoring system, showing how it ultimately makes decisions based on the outcome of its considerations.

Photo of Marcel Flores

Marcel Flores

Verizon Digital Media Services

Marcel Flores is a research scientist at Verizon Digital Media Services, where he explores transport layer optimizations, large-scale traffic management strategies, and cache optimizations. He holds a PhD from Northwestern University, where his research focused on enabling additional channels of communication in existing networks and improving network performance.