NFV, SDN, and the Need for Optimized Server and Processor Architectures

Wednesday, May 13th, 2015 at 6:30 PM

TI Auditorium


6:30 - 7:00 PM Networking & Refreshments
7:00 - 8:00 PM Talks
8:00 - 8:30 PM Panel Session
8:30 - 8:45 PM Speaker Appreciation & Adjournment

Chair: Mehran Nekuii
Organizer: Mehran Nekuii, Saurabh Sureka

Session Abstract:Software Defined Networks (SDN) and Network Function Virtualization (NFV) which advance the evolution of cloud computing are technologies that will be playing a critical role in the next generation data centers and enterprise networks. Driven by continued demand for low-cost computing, the proliferation of NFV & SDN will place additional burden on both compute and communication infrastructure of datacenters. On one end, compute resources will be expected to shoulder additional demands of network workloads, whereas on the other, there will be higher demands on scalability, lower latency, increased I/O bandwidth demands, to mention a few. Two experts from Cavium and Freescale will be reviewing these issues and presenting new optimized server architectures to respond to the higher demands enabled by NFV & SDN.

Speaker: Gopal Hegde, VP/GM, Data Center Processor Group, Cavium Inc.

Bio: Gopal is the VP/GM for Data Center Processor Group at Cavium where he is responsible for Caviumís Data Center Processor strategy, product and business. Gopal has over 22 years of experience driving business, technology and product innovations in silicon, software and systems. Gopal joined Cavium from Calxeda where he was the COO. Gopal revamped Calxeda solutions team and moved the company to an ODM-driven and workload specific model for delivery of ARM based servers. Prior to Calxeda, Gopal worked at Cisco as Senior Director of Engineering responsible for Ciscoís UCS platforms. At Cisco, Gopal played instrumental role to drive products that helped UCS grow from zero to $1.3B+ 3 years. Prior to Cisco, Gopal worked for Adaptec and Intel. At Intel, he served as General Manager of Ethernet switching division in Communications group for 4 years. Later in Intel Server IO Platform, Gopal led the development of technologies enabling integration of PCIe into Sandybridge CPUs, IO virtualization for servers, DCE and FCoE. Gopal holds MSECE from University of Massachusetts at Amherst, and ME from Indian Institute of Science, Bangalore, India.

Title: Workload Optimized Processor Architectures for Next Generation Data Centers and Cloud Applications

Abstract: Traditional server architectures were designed a decade ago to support general purpose applications ranging from home/small business to super computers. They were designed to support a wide variety of applications, each of which is used by limited number of users. Next generation HyperScale Data Centers require highly scalable, power efficient and dense infrastructure. Applications deployed are inherently scale out, requiring scalable compute, high memory performance and virtualization. In addition, Cloud applications that are typically deployed are highly parallel, requiring processors with high memory bandwidth, low latency and tremendous IO capabilities along with support for virtualize functions. With tens of millions of users using the same application, accelerating them in hardware provides significant TCO benefits. This presentation discusses a set of scale out workloads, their differences compared to traditional workloads and how traditional benchmarks are no longer representative for them. It introduces the concept of workload optimized servers to better address the needs of Hyper Scale Data Centers and describes benefits of such a platform.

Speaker: Bharat Mota, Director of Engineering, Networking Applications & Advanced Technology Group, Freescale Inc.

Bio: Bharat is currently Director of Engineering at Freescale's Networking Applications & Advanced Technology Group. He has over 20 years of R&D and entrepreneurial experience in Networking and Telecom Software, Systems and Solutions, including 10+ years in Multicore. Prior to Freescale, Bharat held Engineering management and leadership positions at various Silicon and Systems companies like RMI, Tasman Networks, Digital Link and ROLM Siemens, including founding member of Tiara/Tasman Networks that later got acquired by Nortel. He has delivered many products and solutions to market, ranging from Subsystems and Software Products to CPE, Enterprise and Provider equipment. Bharat was awarded an MS in Computer Engineering from the University of Southern California and a BSEE from the Birla Institute of Technology & Science in Pilani, India.

Title: Acceleration of NFV - Integrated NFV Nodes and Intelligent NICs

Abstract: The rapid migration of network functions from physical machines to the cloud and widespread deployment of virtualized network functions (VNFs) add substantial new demands on compute servers that are often designed to handle non-network workloads. This presentation will cover ways to optimize NFV servers using an innovative Layerscape architecture to achieve higher capacity and performance.