11/09/2015

Reservoir Labs at Supercomputing 2015 (SC15)

Reservoir Labs will be attending Supercomputing 2015 (SC15) on Saturday, November 14 through Thursday, November 19, in Austin, TX. Please join us at the following venues or contact us for a private briefing.

SCinet The Fastest Network Connecting the Fastest Computers
Annually, Supercomputing hosts of one of the most powerful and advanced networks in the world—SCinet. SCinet brings to life a very high-capacity network that supports the revolutionary applications and experiments that are a hallmark of the SC conference. SCinet will link the convention center to research and commercial networks around the world. In doing so, SCinet serves as the platform for exhibitors to demonstrate the advanced computing resources of their home institutions and elsewhere by supporting a wide variety of bandwidth-driven applications including supercomputing and cloud computing. Reservoir Labs donated R-Scope Pace™ advanced threat systems and security expertise to protect the network and its users.

SCinet Briefing—Security in a High Performance Environment
(aka, Lessons from the Trenches)

Tuesday, November 17 at 4 pm, at the Austin Convention Center, RM 5A-5B.
During this session, thought leaders from the SCinet team will highlight the infrastructure created for SCinet and share insights gathered from architecting, supporting and securing SC15’s communication backbone. Following presentations, a good amount of time has been allocated for interactive discussion.

SCinet Demonstrations are available through NRE, Network Research Exhibition, throughout the week.
Each year the SCinet NRE showcases the leading network-based solutions that were created to deliver the highest performance available within the greatest security footprint. The goal of the NRE is to showcase technologies that will impact HPC in general and SCinet in particular. Topics for SC15’s NRE demonstrations and experiments range from software-defined networking (SDN) to security/encryption and resilience.

For NRE, Reservoir Labs will be demonstrating our ENSIGN technology. ENSIGN is a new technology for machine learning that specializes in discovering deep patterns in complex multi-aspect datasets. The unique contribution of ENSIGN is its ability to scale the unsupervised learning methods enabled by tensor decomposition techniques to industrial-scale problems. For complex multi-aspect datasets, these decompositions provide a level of insight into cross-aspect correlation that cannot be replicated with traditional machine learning methods.

Please contact Reservoir Labs if you would like to schedule a demonstration.

HP-CAST25 HP Consortium for Advanced Scientific and Technical Computing.
Saturday, November 14, 13:30 at the JW Marriott Hotel (two blocks west of the convention center)
Reservoir Labs: Advanced Graph Analytics Using HPC

Ms. Ann Johnson, Vice President of Engineering, will introduce the latest advances in Reservoir Labs’ ENSIGN technology, a multi-aspect analysis tool for Big Data that leverages hypergraph decomposition technology. Using Reservoir’s patented algorithms and data structures, ENSIGN enables parallel processing of large hypergraphs by processing sparse tensors through multi-linear algebraic decompositions.

INDIS Innovating the Network for Data Intensive Science
Monday, November 16, 11:15AM at the Hilton, RM 410 (across from the convention center)
Reservoir Labs: High-Performance Many-Core Networking: Design and Implementation

Reservoir Labs invites you to attend our presentation for the SCinet 2015 INDIS Workshop. Dr. Jordi Ros-Giralt, Managing Engineer, will introduce DNAC, Dynamic Network Acceleration for Many-Core, a high performance abstraction layer designed to target the maximum network performance available from the network interface in many-core general purpose operating systems. Reservoir will share the performance gains that can be achieved in a state-of-the-art many-core system.

The second annual INDIS workshop is part of the SC15 Tech Program. INDIS showcases both Demonstrations and Technical Papers highlighting important developments in high-performance networking.

RESPA Runtime Systems for Extreme Scale Programming Models and Architectures
Monday, November 16, 15:30—16:00 at the Hilton, RM 400-402 (across from the convention center)
Reservoir Labs: Automatic Code Generation for an Asynchronous Task-Based Runtime

Dr. Muthu Baskaran, Managing Engineer, will describe capabilities developed at Reservoir Labs within R-Stream—an automatic source-to-source optimization compiler—for automatic generation and optimization of code targeted towards Open Community Runtime (OCR)—an exascale-ready asynchronous task-based runtime. Hardware scaling considerations associated with the quest for exascale and extreme scale computing are driving system designers to consider event-driven-task (EDT)-oriented execution models for executing on deep hardware hierarchies. Further, for performance, productivity, and code sustainability reasons, there is an increasing demand for auto-parallelizing compiler technologies to automatically produce code for such asynchronous EDT-based runtimes. Dr. Baskaran will discuss some of the key features in R-Stream for enhancing EDT runtime execution, specifically features that extract more concurrency, data locality, and asynchrony in programs, and manage data movement and memory explicitly to save energy.

OCR Open Community Runtime: Birds of a Feather (BOF)
Wednesday, November 18, 12:15PM—1:15PM at Austin Convention Center, RM 18AB
Reservoir Labs: Experiences With OCR as a Target for an Optimizing Compiler

The OCR is a standard for event driven task execution in high performance computing, being developed in a collaboration that spans several continents. Reservoir Labs has developed compiler technology to parallelize programs into the event driven task form to achieve performance and power efficiency much higher than can be achieved with traditional execution models.

Dr. Benoit Meister, Managing Engineer, will present Reservoir Labs’ R-Stream compiler technology and show how it can utilize advanced features of OCR such as data block management and dynamic task graph spawning. Reservoir Labs’ R-Stream compiler has produced more code, and better code, than any other path toward OCR that exists today.