Linley Cloud Hardware Conference 2017
Covers system design for cloud computing and networking
Held On February 8, 2017
Hyatt Regency Hotel, Santa Clara, CA
Agenda: February 8, 2017
Datacenter Networking Market Transitions
This keynote talk will cover the evolution of merchant silicon for data-center switching and routing, metro and long-haul communications, the transition in optics technologies and form factors, and the resulting rapid expansion of bandwidth in particular in cloud networking.There will be Q&A following this presentation.
|10:20am-11:50am||Session 1: Optical Interconnects at 100G and Above
This session, moderated by Jag Bolaria, Principal Analyst at The Linley Group, will focus on scaling optical bandwidth within and across data centers.
PAM4 Shaping the Future of Data-Center Optical Interconnects
PAM4 modulation has now been recognized as the modulation scheme that will take the industry over the next wave of Ethernet deployments by doubling the bits per symbol at the same baud rate. As the massive megatrends of cloud computing, Big Data, IoT, Social and Web2.0 continue to accelerate and drive bandwidth requirements, there is clear need to increase the speed of interconnect pipes while maintaining cloud economics and lowering carbon footprints. We will explore how PAM4 is enabling the next generation of interconnects for 200G and 400G.
Forces Align for 100G Serial
The networking industry is at an inflection point due to advances in semiconductor process technology coupled with a seemingly insatiable demand for bandwidth from cloud providers. Economic realities have compelled a shift from conventional NRZ signaling to multi-level signaling for both optical and copper interconnects. PAM4 signaling used in conjunction with DSP fulfills the bandwidth requirements while leveraging the advantages in 16nm CMOS and beyond.
Silicon Photonics, a Foundry Perspective
This talk will present a view on emerging cloud data-center applications based on silicon photonic interconnect switching solutions. Communication standards like 100G and 400G Ethernet and interoperable specifications are discussed, including architectures likely to drive the interconnect paradigm from electrical to photonic and from discrete components to modules. A foundry view of photonic systems' rate and pace of market adoption with Hypercloud data centers interconnectivity, including servers, switches and storage, will be presented. Finally, this talk will provide commentary on some challenges in industry engagement within the photonic supply chain.There will be Q&A and a panel discussion featuring above speakers.
|1:00pm-2:30pm||Session 2: Protocol Processing and Server Offload
This session, moderated by Bob Wheeler, Principal Analyst at The Linley Group, will discuss security-protocol processing as well as offloads implemented in server NICs
Layer 2 MACsec Security for 400GE Data-Center Interconnects
As network speeds continue to rise, efficiently protecting the heavier data traffic is an absolute must. The MACsec protocol allows efficient scaling to support extreme bandwidths to secure Data Center Interconnect links. The Optical Internetworking Forum (OIF) advocates FlexEthernet instead of link aggregation to fill in the gaps and support multiple interfaces. This presentation will explore the dedicated crypto and security-protocol acceleration required for 400-Gigabit Ethernet and FlexEthernet - while keeping die area and power consumption under control.
Chelsio Terminator: Enabling High-Speed Secure Network Interconnects
The concept of secure network convergence around 10/25/40/50/100GbE has been discussed in the industry for some time. The security aspect is being addressed by the ubiquitous TLS/SSL, SMB 3.X crypto, IPsec, and DTLS, as well as AES XTS for data at rest. In this talk, Chelsio discusses the use cases of the different crypto suites in the cloud and in the data center, and discusses how its Terminator T6, which has 100Gbps low-latency crypto features, addresses these use cases.
Scalable Server Offload for Cloud Networking “Done Right”
The deployment of Network Functions Virtualization (NFV) is on the rise and cloud operators are recognizing that enhanced network adapter cards can dramatically boost the server's networking performance while freeing CPU cycles for applications. But since the workloads vary widely within the cloud, there's no single network offload product that fits all scenarios. Mellanox will explore its scalable, interoperable offerings including ConnectX advanced network adapters, FPGA-accelerated Innova NICs, and BlueField multicore ARM "Software-Defined NICs".There will be Q&A and a panel discussion featuring above speakers.
|2:50pm-3:50pm||Session 3: Advanced Memory Technology for the Cloud
This session moderated by Bob Wheeler, Principal Analyst at The Linley Group, will examine new memory architectures and applications of new memory technologies in cloud data centers.
Searching Beyond TCAMs
Search is key to all network services. Traditionally, this has been done using exact match, LPM, and multi-field match using TCAM. With white boxes, SDN, NFV, and security, the desire for different kinds of searches, including regex and in-field flexibility, is emerging. While multi-core CPU-based servers can address this need, offload engines can provide more cost- and power-effective solutions. MoSys will present a breadth of software-defined solutions using its Programmable Search Engine (PSE) platform while highlighting architectural features.
Rethinking System Architectures: The Shifting Performance Bottlenecks Driving Future Silicon Design
Moore's Law has relentlessly delivered tremendous improvements in processing power and functionality for several decades. While processors will continue to become more feature-rich, the emergence of new computing paradigms and slower rates of improvement in other areas mean that power and performance bottlenecks are moving away from processors into other subsystems. Moving data between processing elements, memory, storage, and across networks is becoming a growing challenge, and concerns about power and security are forcing chip and system architects to rethink silicon and system architecture. This talk will discuss how these concerns are driving changes in the way we think about future silicon and systems.There will be Q&A discussion following each speaker in this session.
|3:50pm-4:40pm||Special Panel: Accelerating The Cloud
Cloud data centers perform a variety of different workloads. Some of these workloads perform well on general-purpose servers, but others can be offloaded to more efficient special-purpose processors, GPUs and ASICs. In this session, Jag Bolaria will host a panel of experts to discuss the latest trends and evolution of cloud computing hardware and infrastructure.
|4:40pm-6:00pm||Reception and Exhibits|