When : Tuesday, December 19 – 1:00 PM- 3:00PM
Azure provides on-demand true HPC capability that enable you to run large parallel and batch HPC jobs and extend your on-premises HPC cluster to the cloud when you need more capacity.
In this session we will start by introducing the Azure cloud and the Azure presence in India. After this we will do a deep dive of HPC in Azure cloud including Hardware (compute, storage, network) and software (Middleware and ISV application) stacks along with typical solution architectures in both hybrid and native cloud deployments. We will explore HPC solutions in practice by considering use case scenarios in key industries including finance , manufacturing, life sciences and educational research.
Tony has over 24 years of IT experience in enterprise and solution architecture across vertical industries ( FSI, Healthcare, Energy, CME, Manufacturing) with specialty on enterprise and cloud technology associated with High performance advance computing solutions. He currently works as principal HPC architect in Microsoft AzureCAT (customer advisory team) within the Azure engineering division and provides expert level engineering advice and technical assistance on large complex high performance parallel processing solution in Microsoft Azure native and hybrid cloud running High performance computational and data intensive workloads as well as emerging AI ( ML/DL) solution in various vertical industries.
Rangarajan Srirangam is Principal Program Manager in the Azure Customer Advisory (Azure CAT) team. He has over 21 years of experience in the IT Industry in different roles and has worked for the past 16 years at Microsoft. He has consulted on diverse projects for large customers and delivered complex solutions spanning multiple Microsoft technologies. Ranga currently focuses on cloud solutions involving the Microsoft Azure platform. He has spoken in many conferences and events such as Tech Ed, Tech Ready, Architect Summits etc and published technical papers. As part of the Azure CAT team in India he works with field teams and customers on cloud projects of significant scale , volume and complexity.
Rakesh Patil is Senior Program Manager in the Microsoft Azure Cloud Engineering division as part of Azure Customer Advisory Team (CAT). He has over 17 years of experience in the IT Industry. Rakesh works as Senior Program Manager at Microsoft for Customer Advisory team and works with customers who build and deploy cutting edge solutions on Azure that push engineering boundaries. Before joining Microsoft, Rakesh had been working for RedHat India concentrating on cloud, openstack and Devops environment. His focus area is Cloud projects involving Open source components, Microservices Architecture, linux and HPC workloads.
Narendra is the GM of Customer Success Unit, which is responsible for the Top Azure Customers of Microsoft. For the first 18 months at Microsoft, he was responsible for DX, Developer relations with startups, ISVs and tech communities. DX also worked on developer productivity with Visual studio business in India. Earlier to Microsoft, Narendra worked at Intel for 20+ years on CAD, 3D Graphics stacks and moved to India in 1998. His most recent stint was running AsiaPac & Japan for Software and Services Group. A Computer Science graduate from RV College of Engineering, Bangalore and an MS Computer Science from Arizona State university. Narendra also has published on Multi-Level Texture Caching for 3D Graphics and in IEEE on Hardware and a research testbed for switching software. He also did HPC switching systems work at CDOT Bangalore.
When : Tuesday, December 19 – 10:00 AM- 12:00 noon
Title: Accelerating your workloads with NVMe SSDs
Speaker: Esther Spanjer, Western Digital
Join Esther Spanjer from Western Digital for this presentation where she will discuss how NVMe SSDs are enabling the next generation datacenter. She will discuss the unique advantages of NVMe SSDs over alternative solutions and discuss the best use cases for NVMe SSD in a variety of environments, such as vSAN, Hadoop, Microsoft Storage Spaces Direct, SQL Server and others. She will furthermore discuss the next development in the datacenter when NVMe Over Fabric will be introduced, and how this can enable new architectures that allow for higher performance at lower cost.
Esther Spanjer is Director of Business Development EMEA at Western Digital. With more than 22 years of experience in the flash industry, Ms. Spanjer has gained valuable knowledge on the requirements for flash-based storage solutions in enterprise storage applications. She has authored more than 30 articles in international publications and conference proceedings, and has been chosen to speak at a number of industry events, including DiskCon, ARM Developers conference, Server Blade Summit, Storage Networking World and Flash Memory Summit. She holds a B.Sc. degree in Electronic Engineering from the Technical University Amsterdam (Netherlands) and an MBA from Presidio Graduate School in San Francisco (USA).
When : Wednesday, December 20 – 1:00 PM- 3:00PM
Title : Hardware Accelerators for NoSQL databases
Speaker: Chidamber Kulkarni – RENIAC
In this talk, we will discuss our experiences in building a low-latency, high-throughput data acceleration engine for NoSQL databases, such as Cassandra. We will introduce the concept of a Data Proxy that marries the concepts of a transparent data layer proxy, that can talk CQL (for Cassandra), and the concept of a storage engine, that implements caching of data across multiple memory technologies such as SRAM, DRAM, and Flash memories. We will then discuss the different accelerator technologies that can be used to build such a data acceleration engine and elucidate why FPGAs make the right choice. In this talk, we will also highlight the key design decisions and trade-offs that we made to leverage the power of FPGA devices to deliver 1/3rd to 1/10th lower latency and significantly higher throughput by saturating the 10Gb Ethernet within a server. Finally, we will present some of our benchmark results to validate our claims. To conclude, we will show how the Data Proxy can be applied to other NoSQL databases, and adjacent technologies such as Search (using Solr/ElasticSearch) both as an on-prem solution and on cloud platforms with FPGAs, such as AWS F1.
Title : Video codec acceleration on FPGAs
Speaker: Mahesh N Shukla – Xilinx
With the need to handle high-resolution video streams (1080p to 4K and beyond), advanced video compression standards such as HEVC, VP9, and AV1 are being developed and deployed. The high resolution raw streams present a huge challenge, with the need to compress enormous data in the available network bandwidth. For example, 4kp60-10bit takes around 5Gbps in its raw form, which means that in a network bandwidth of 10Mbps, raw data needs to be compressed by 500x. To cater to this need, the newer compression standards are more complex and are challenging to implement. For example, there is a 30x increase in complexity from H264 1080p30 to HEVC 4kp60. This complexity means that hardware / SoC implementations need higher area and power budgets. FPGAs are well suited to address this problem, as they are programmable.
In this talk, we will discuss the video acceleration strategy at Xilinx and the enablement of solutions for video applications. We will also talk about the video codec roadmap and how dynamic codec switching offers high performance at a significantly low cost.