Session 1: Market and Technology Trends

Jag Bolaria, senior analyst at The Linley Group, presented an overview of market, technology, equipment-design, and silicon trends for designers of enterprise and data-center networking equipment.

Session 2: Data Center Interconnects

This session, moderated by Jag Bolaria, discussed the different form factors and types of interconnects needed for 10GbE, 40GbE, and 100GbE.

Demystifying Interconnect Form Factor Options in Next-Generation Data Centers
     Jan Meise, Director of Strategic Technology, Finisar

Data center server connectivity is finally on the verge of moving to 10GbE with the availability of single-chip LAN on motherboard MAC solutions. Leaf and core switches will need to handle aggregation and upgrade to 40GbE and 100GbE quickly. This presentation will focus on recent developments in interconnect form factors and standards for next-generation data center network architectures.

Energy-Efficient 10GBASE-T - Enabling a New Generation of 10G Interconnects for Green Data Centers
     Sid Yenamandra, Vice President, Plato Networks

Delivering 10 Gbits/s of performance up to 100 m over Cat6A or Cat7 copper cables, 10GBASE-T offers lower link costs than optical interconnects, provides greater wiring flexibility than direct-attach copper cables, and delivers the architectural elegance of a single standards-based interconnect technology backwards compatible with incumbent 100/1000BASE-T networks. However, several technical challenges and deployment issues related to power and performance still need to be addressed. In this presentation, Plato Networks addresses these challenges in its newest generation 10GBASE-T PHY device. Additionally, with new standards such as Energy Efficient Ethernet (EEE), as defined by IEEE802.3az, rapidly taking shape, they will analyze the impact and benefits these specifications will have on Plato’s 10GBASE-T PHY architecture.

HyperTransport – System-Wide Data Management Interconnect Standard
     Mario Cavalli, General Manager, HyperTransport Consortium

HyperTransport interconnect technology is ideally suited for data center and HPC environments, which require scalable computing architectures, virtualized resource provisioning, and the best performance. This presentation provides an overview of HyperTransport capabilities and examples of how HyperTransport delivers leading scalability, bandwidth and latency, while giving the industry the opportunity to tap into an open product and support ecosystem.

Session 3: Packet Processing in Virtualized Data Centers

This session, moderated by Bob Wheeler, senior analyst at The Linley Group, explored the role of network processor and hardware accelerators for packet processing and deep packet inspection in data center and networking.

Using NPUs in Cloud Computing to Scale Service Delivery
     Anders Wirkestrand, Director of Technical Marketing, Xelerated

Cloud computing will come in two phases where the first phase is in the centralized data centers, and the second phase is more distributed with service delivery integrated with the switching and routing equipment. Using Network Processors, NPUs, to do the packet processing in combination with dedicated service processors raises the efficiency in the data center and enables better scaling of the services. This presentation will explain the role of NPUs in next generation data center networking equipment.

Unified Computing - I/O Requirements for the Convergence of Virtualized Servers and Networking in Data Centers
     Nabil Damouny, Senior Director, Strategic Marketing, Netronome

This presentation will describe the major engineering design implications of virtualized I/O in servers, appliances, and networking equipment as well as the requirements to support 10Gbps data rates. To meet these stringent requirements, it will also introduce a heterogeneous multicore architecture that combines multicore x86 cores with networking optimized cores for virtual switching, security processing, classification, and deep-packet inspection over a high-speed virtualized PCIe datapath via SR-IOV. The presentation will discuss network throughput, latency, memory utilization, security, management, and optimal power consumption possible with this solution.

Accelerating Layer 4 to 7 Content Processing
     Kin-Yip Liu, Director of Customer Solutions Architecture, Cavium Networks

Data center applications increasingly require up to Layer 7 content and deep packet inspection (DPI). This requirement is driven by network security requirements and content-based QoS processing of triple-play network traffic. High performance L4-L7 processing at low power consumption requires hardware acceleration of TCP, SSL, compression/decompression, and regular expression matching. Traditionally, regular expression based pattern matching is performed using either Deterministic Finite Automata (DFA) or Non-deterministic Finite Automata (NFA) although neither by itself is ideal. Cavium’s Hyper Finite Automata (HFA) utilizes an optimal combination of DFA and NFA to offer deterministic low latency while avoiding graph size explosion. This presentation illustrates the HFA architecture and its advantages, as well as other L4-L7 processing hardware acceleration engines in key applications.

Session 4: Data-Center Switch and Network Architecture

This session, moderated by Jag Bolaria, provided an overview of how Ethernet is being enhanced to enable convergence in the data center. It will also detail different switch architectures for data centers and cloud computing as well as the key requirements in these applications.

Convergence: The New Killer App
     Brad Booth, Director of Technology, CTO Office, AppliedMicro

Convergence within the data center holds the promise of reduced capital and operating expenses. Everything from storage, video, voice, can be converged onto one networking technology: Ethernet. This presentation takes a peek at the converged Ethernet technologies at play and how AppliedMicro is tackling those challenges.

Moving to a Flat Data Center
     Ofer Iny, CTO, Dune Networks

New silicon architectures are enabling the transition from tiered data center network architectures to flat designs addressing challenges and complexities that exist in today's data centers while improving performance and manageability. Dune will present architectural considerations in flattening networks and building scalable flat switches.

Data Center Trends
     Uri Cummings, CTO, Fulcrum Microsystems

Whether cloud computing, supercomputing, or enterprise computing, the trend is toward data centers of ever-increasing size. Low power, increased compute performance, and greater bandwidth requirements are driving data centers to adopt new scalable, cost-efficient network architectures. Latency and jitter between nodes, aggregate bandwidth, and price per bandwidth are the new data center performance metrics. As a result, networks are becoming flatter yet more scalable; new protocols, such as TRILL, are emerging to simplify network control; and emerging standard APIs, such as OpenFlow, promise to make the network infrastructure less proprietary and more cost effective. This presentation examines the new data-center networking architectures and the underlying switch chip technology that is powering the new breed of data-center networking systems.

Request a free copy of the presentations by completing the registration form.

The seminar was intended for network-equipment vendors, server OEMs, system designers, network service providers, enterprise-network managers, software developers, press, and the financial community.

Information collected for this event will be shared with the sponsors paying for this seminar. This information will not be shared with companies other than the sponsors of this event.

Further questions?   Contact The Linley Group:
Phone: 1.800.413.2881 (toll free in US) or 1.408.281.1947 or email: customer service