SMART HIGH-PERFORMANCE NETWORKS Workshop

Introduction

Links

Introduction

Smart High-Performance Networks - Towards a New Generation of Intelligent Networking Infrastructure for Distributed Science Environment Workshop

SPONSORED BY THE U.S. DEPARTMENT OF ENERGY,
OFFICE OF ADVANCED SCIENTIFIC COMPUTING RESEARCH
ROCKVILLE, MARYLAND,
DECEMBER 8-9, 2016

The purpose of this workshop is to bring together network researchers and operators from national laboratories, academia, and industry to identify and discuss emerging opportunities and challenges in the design and development of a new generation of smart high-performance network infrastructures to support distributed extreme-scale science.

Scientists increasingly depend on highly reliable and secure high-performance networks to access critical science facilities, collaborate, and share massive volume of data. As this trend continues, networks are expected to provide more sophisticated easy to use, secure, and predictable intelligent services. These expectations translate to a new generation of high-performance networks with intelligent capabilities delivered to scientists in the form of just-in-time Network-as- a-Smart service (NaaSS).

The workshop will explore emerging concepts such as cognitive networking, machine learning, network science, Software Design Networking (SDN) as critical technologies for designing smart networks capable of delivering NaaSS. The desired features of these networks include self-awareness, auto-defensiveness, self- reconfiguration, and self-healing that collectively define a new networking paradigm known as AUTOGNOSTIC networking - namely self-aware networks designed and instrumented to continuously adapt to its operational conditions, traffic loads, inter- domain policies, cyber treads, and performance conditions. The ultimate goal is a smart network infrastructure that can monitor itself, diagnose and resolve problems, defend itself from cyber-attacks, and provide intelligent services to scientists. The following key characteristics of smart networks are noted:

Self-Aware & Self-Managed Networks:
- Organization
- Optimization o Configuration o Adaptation
- Healing
- Defense/Protection o Contextualization

This workshop is organized around four key Technical Focus Areas:

Smart Networks (Leads: Tom Lehman, Inder Monga, Bryan Lyles)
Smart Applications (Leads: Ian Foster, Raju Vatsavai)
Artificial Intelligence Based Technology for Smart Networked Systems (Leads: Prasanna Balaprakash, Kalyan Perumalla, Nagi Rao)
Smart Cyber Security Sub-systems (Leads: Stacy Prowell)

There will be separate breakout groups for each of these areas where the key issues, concepts, and recommendations will be discussed. The discussion in each of these breakout groups should be held in the context of smart networks. As an example there should be discussion about the interface between smart networks and smart applications, artificial intelligence systems, smart cybersecurity infrastructure. There should be discussion regarding what the smart network needs to provide to these other smart systems, and vise versa. The time frames considered should be near term (2-4 years) and longer term (5-10 years).
Each breakout group may start with one or more talks by a technical area expert. The outputs of these breakout groups should include the following:

Describe the current state of the art and what communities/companies are leading
Define differences between commercial and R&E needs, directions, issues
Identify the key research problems
Recommend approaches for DOE, and other Government Research organization, to make progress on smart networks and associated cross discipline systems