Presentation is loading. Please wait.

Presentation is loading. Please wait.

Azure HPC Solution pitch deck.

Similar presentations


Presentation on theme: "Azure HPC Solution pitch deck."— Presentation transcript:

1 Azure HPC Solution pitch deck

2 500+ million core hours running HPC on Azure
High Performance Computing 500+ million core hours running HPC on Azure Download the commercial story and ambition stories at

3 Evolution of HPC on Azure
Engineering Insight Timeline Cloud Computing, 2012 Windows HPC on Azure, 2014 GPU on Azure, August 2016 (private preview) New HPC instances soon Linux RDMA (infiniband) on Azure, July 2015 Enhanced networking to support any MPI Very large tenant sizes

4 Hyperscale infrastructure
30+ regions worldwide—huge capacity around the world—growing every year UK East UK West North Central US Illinois West Europe Netherlands North Europe Ireland Canada Central Toronto Central US Iowa Canada East Quebec City Germany West  Germany East  China North * Beijing US Gov Iowa Japan East Saitama China South * Shanghai West US California East US Virginia India Central Pune Japan West Osaka East US 2 Virginia South Central US Texas India South Chennai US Gov Virginia India West Mumbai East Asia Hong Kong Why this Slide: This is SUCH a big investment – it’s a game for only very few. It’s not new for us – we have been doing this for our own services and our consumer/web properties for 20+ years Key Points: Where are we – EVERYWHERE…! How big is this - $15+ B and counting – this is serious, we continue to bet big and you can count on us Talk about DC innovation – DC Efficiency and Gen 5 data centers. Scale – at this scale you do get efficiencies – the main one being POWER Remember our “strategy” – we will be in the major places, but not everywhere – we have Azure Stack/Hosters for that. Transition to NEXT Slide: This is the physical infrastructure that Azure sits on, now lets talk about Azure the PLATFORM SE Asia Singapore 100+ datacenters 2x of Amazon, 6x of Google $15 billion investment Australia East New South Wales Australia South East Victoria Brazil South Sao Paulo Operational Announced * Operated by 21Vianet  Operated by Deutsche Telekom

5 Customers are moving their HPC workloads to Azure
Maturity Advanced security and cost competitive Current and high-end GPUs for deep learning, compute, and visualization Dedicated HPC VMs - close to bare metal clusters Agility and scale Quick time to market for any workload Scale at hand and flexibility on demand Cost No infrastructure investments Pay only for what you use

6 Value proposition for HPC in the Cloud
Single system Most time Value: Achieve more Faster time to value! Value: Reduce time On premises clusters are: High CAPEX High OPEX Resources limited by space and allocations Azure provides: Zero CAPEX Low and competitive OPEX On Demand capacity when you need it Pay as you go usage model Cost of 1,000 cores for 10 hours = cost of 10,000 cores for 1 hour Calculation time Time to result is a real constrain in Big Compute/HPC, with Azure customers can provision as many core as they need in order to have a fast results. => With Azure you are billed per minutes, so If our customers need cores to get result fast, the Azure cost will be the same as waiting 10hrs with 1000 cores Computers/Nodes Multi-system Least time

7 DYNAMIC CONFIGURATION
Value proposition for HPC in the Cloud Azure Instances Best value for your investment! HPC apps profiles: Parallel Distributed IO Intensive Varying needs: Cores and Memory Networking and Storage Azure provides: On Demand capacity Different configuration capabilities InfiniBand Investment = Application investments As customer can deploy the right configuration they need, they will have the best ROI FIXED CONFIGURATION DYNAMIC CONFIGURATION

8 Microsoft HPC value proposition
10/23/2017 5:36 AM Linux and Windows machines tuned for HPC Provision thousand of cores in few minutes around the world Latest GPUs in the cloud Building blocks to deploy HPC clusters that meet your specific needs Big Data platform services to gain insight from HPC jobs Pay only for what you use by the minute Today, most organizations have on-premises high- performance computing infrastructure. We see everything from a set of servers under someone’s desk to large datacenter environments, and everywhere in between, depending on what the company is trying to accomplish. While on-premises HPC is a good option for many scenarios, it also comes with inherent constraints. Generally speaking, an HPC infrastructure involves millions of dollars of capital cost (or expense, if it’s leased infrastructure), given the significant infrastructure required. <Speaker Guidance: you may not need to elaborate on what this capital investment includes – below is optional reference if needed> This investment includes: X86 servers with the latest chip sets and high-speed, low- latency interconnects Expensive application licensing Sophisticated scheduling software to ensure clusters are fully utilized High-speed storage or parallel file system to access single name space from multiple servers In some cases, remote visualization solutions are also involved And these massive investments are hard to keep current – hardware often goes out of date before it can be refreshed. Operational costs like power, cooling and maintenance also add up, and it’s common to run into constraints, especially with power and capacity. All of this assumes that there is physical space for more compute infrastructure. We’ve heard of firms taking over parking lots and other facilities in order to build out their HPC infrastructure, but that’s not always an option. Dedicating more space to infrastructure also comes with its own costs and hassles. An even greater issue is constraints on engineers’ work efforts. When we talk to engineers, they often tell us about long wait times for shared capacity. When an engineer needs a job run, they are in ‘wait mode’ while their jobs sit in a queue, waiting to be run This can be a multi-day day wait until they get to the front of the line, impacting productivity When you consider the allocation challenges of balancing small jobs with massive ones, the problem gets even trickier. As a result, engineers are constrained to using HPC to validate design choices. Using HPC for innovative purposes, like experimenting with design direction, isn’t even an option due to the capacity constraints. This type of environment is also inflexible – it’s not designed to scale up and down as needed In many firms, there are peaks and valleys of hardware utilization – for example, there might be large projects or seasonal work where your hardware is running at maximum capacity and still can’t keep up. At some organizations, it can take months to open a purchase order and add service, so if there are unexpected peaks in demand, there’s no way to quickly cope And there is also inflexibility when workloads are reduced – but you are still paying for the downtime. The hardware is often running and available whether you use it or not, and must be maintained. Finding space for HPC infrastructure, investing in it, keeping it current, and maintaining ends up being time- consuming and extremely complex. And efficiently meeting demand is often not possible. Engineers are limited to using HPC for design validation, rather than using it for strategic purposes. Transition: Microsoft Azure provides an alternative approach to on-premises infrastructure. © 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

9 Azure value proposition
Azure features Enterprise-ready Large-scale on-demand Infrastructure and Hardware built and operated for HPC RDMA for fast MPI performance Parallel file systems Support for Intel Lustre and Red Hat GlusterFS Azure Marketplace Azure Marketplace offers variety of vendor certified solutions and a robust HPC ecosystem MPI support Intel MPI, Platform MPI (IBM) in Beta Core differentiators Hybrid HPC mode Microsoft offers hybrid “Burst to Azure” HPC-as-a-service Azure offers HPC-as-a-service through Azure Batch Multi-OS support Azure supports both Windows and Linux OS options in HPC Clusters and Batch Better engineering Advanced engineering, faster infrastructure and competitive pricing

10 The Microsoft Azure HPC ecosystem
Hardware & Software Infrastructure Big Compute Marketplace Certified Applications

11 The soft bits Solvers OS support Schedulers MPI support
SLES 12, CENT OS 7.1 Windows Server 2012 R2 Schedulers PBS-PRO/COMPUTE MANAGER IBM Platform LSF & Symfoni Tibco Data Synapse Microsoft HPC_PACK MPI support Intel MPI Platform MPI (in prototype phase) MS MPI on Windows Parallel file systems Intel Lustre Red Hat GlusterFS Solvers Intersect (Schlumberger) NAMD LS-DYNA FLUENT STAR-CCM RADIOSS ACU-SOLVE OPENFOAM PAM-CRASH LANDMARK (Halliburton) ABAQUS MSC NASTRAN OPTISTRUCT

12 Deployment options for every scenario
10/23/2017 5:36 AM Deployment options for every scenario All in cloud No compromise “all in cloud” deployments, using the dedicated Azure HPC components with your preferred application and middleware. Hybrid—Burst to cloud Address changing capacity needs. Extend your HPC jobs to Azure for on-demand scale and flexibility. There are two key scenarios that we would like to explore for a minute. First, is our enterprise scale scenario, where you leverage Azure for additional capacity or new workloads. Second, is Azure burst, allowing you to burst to the cloud for more capacity. Both scenarios are truly pay as you go without any upfront cost of infrastructure. Transition: Let’s look at our enterprise scale model. As a service Easy planning, deployment and orchestration of your HPC jobs. Batch is a fully automated “as a service” engine for HPC type workloads— ISV or end user. © 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

13 Batch (HPC as a service)
Rich-client experience for users Use resources depending on workload needs Burst only when required Leverage existing applications and algorithms Manage “pools” of VMs for different type of users Get clear insights into application and service usage Azure Batch VM Pool RETURN RESULTS Azure BLAST Portal App EXE App EXE App EXE VMs SUBMIT JOB UPLOAD DATA

14 Azure Batch and Compute Stack
Service/Solution Get and Manage VMs Task failure? Task frozen? Install Task Applications Manage and Authenticate Users Azure Batch VM Management and Job Scheduling Start the tasks Batch processing platform Avoid building infrastructure Queue tasks Move task input and output Scale up and down PaaS Cloud Services IaaS VM/VMSS Hardware

15 Solutions

16 Finance and Capital Markets
Compliance Regulations and legislation Products get more complex Run on regular basis (quarterly, daily) Risk management and analytics Risk Compute Grids Modeling & Analytics Tests of new algorithm at scale Home made applications Operations Disaster Recovery Optimal TCO Flexibility for business line Compliance: Insurance = Solvency 2 Bank = BASEL 2/3/4 Risk Management: Use case are Counterparty Risk management , CVA (Credit Valuation Adjustment), VaR (Value At Risk), Stress testing, tes & dev for Quants, Algorithmc models For Solvency 2: Leader applications are TOWER WATSON Moses, SUNGARD Prophet, MILIMAN Mg Alpha All can run in Azure FLEXIBLE = when customer need compliance run (quaterly, daily) or react on a market event (for instance black swan event with the Swiss Franc in January 2015) SECURE = Secured storage and disks, secure network connections. TCO = pay only when you deploy ressources Microsoft Azure based solution enables Financial Institutions deploy and manage grids for risk management and related compute needs in a flexible, secure, and controlled manner for optimal TCO.

17 Improving operational insights
Top Japanese bank in the UK gained advanced analytics through the cloud Challenge Gain flexibility to analyze advanced capital scenarios. Decrease risk and build greater confidence across the business. Batch Implemented a bursting scenario using Microsoft HPC Server with Azure Moved 2,000 cores per day to the cloud Results Avoided the additional capital expenditure for a new datacenter Obtained consistent, reliable core deployments on demand Gained business agility for better competitiveness Key points: Customer profile: This customer is the European capital markets hub of one of the largest financial institutions in the world. Headquartered in London, this bank is active throughout the international capital markets, focusing on debt, equity, derivatives and structured products. Business goal: The risk group needed their information ready to go each morning but there was not enough time to complete these more advanced models overnight, given the already taxed production environment. Further, the risk department could not confidently predict precise capacity required as markets are increasingly volatile. The bank did not have the capacity on their own grid and they wanted to avoid capital expenditures from a large datacenter purchase for expansion. Tactics: To meet these expansion demands and to speed up the time to delivery for the business, the bank selected a Microsoft HPC Server and Azure solution. Their goal was to move at least 2,000 cores per day to the cloud, used in concert with on-premises datacenters allowing maximum flexibility. Results: By avoiding the capital expenditure of a new datacenter, the bank was able to meet growth expectations and align their costs to their business strategy. They are now running Azure daily and can obtain consistent, reliable core deployments on demand. The solution has provided the bank with the business agility it needs to be competitive today and in the future. Hybrid solution Consuming 2K cores per day

18 New financial modeling software
10/23/2017 5:36 AM Tower Watson Risk Agility FM Challenge Designed for Actuaries to develop and run models Projection system to model asset and liability cash flows over the term of policies Batch Auto-scale to meet demand Lightweight REST interface and API libraries in various programming languages Results Reduced costs and administration. Large scale deployments without infrastructure Opens up grid computing to smaller clients Increased business agility they have a application called Risk Agility FM which is traditionally a cluster application and is essentially designed for Actuaries to develop and run models to manage risk and safegaurd solvency and its primary focus is on life insurance. Their software allows them to deploy sophisticated analytical calaculations based on realistic economic principles. One of their biggest motivations is that this service opens up a host of new opportunities for their smaller clients who can't afford to build on premise clusters and can now use Risk Agility FM as a service. It also helps that Towers watson is providing and concentrating on business value which is their analysis and now spending cost and time on managing infrastructure. They also went through the evolution of running on premise, then going hybrid and then finally evolving to the SaaS model. What would you do with 100,000 cores? Big compute at global scale Manage risk and safeguard solvency 2 Azure Batch © 2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

19 Manufacturing Engineering design Applications Operations
Drive design of products Design, Simulate, Analyze results Simulated problems are complex Applications ISV applications Linux Specialized hardware Operations Disaster Recovery Optimal TCO Flexibility for business line FLEXIBLE = when customer have a new and important project or an unexpected simulation to be done POWERFUL = Azure provide HPC instance (Infinband, GPU, high end core) TCO = pay only when you deploy ressources Microsoft Azure based solutions enable Manufacturing Companies to deploy and manage clusters for physical simulation in a flexible, powerful and controlled manner for optimal TCO

20 Simulation and analytics in Azure
Business transformation in Motorsports Challenge Make better decisions on car setup and race strategy through simulation and data analysis Desktop to garage to racetrack Batch CFD in Azure to model airflow Azure Machine Learning and Cortana Analytics for race strategy Results CFD results in half the time Highly accurate models for fuel and tires Tools for the simulator and at race day

21 Oil and Gas Seismic data processing Reservoir simulation Operations
Exploration of new oil field Process large amount of data Complex problem Reservoir simulation Development of new fields Developed fields for oil forecasts ISV applications Linux Specialized hardware Operations Disaster Recovery Optimal TCO Flexibility for business line Microsoft Azure based solution enables Oil and Gas company to deploy and manage clusters in a flexible, powerful and controlled fashion for optimal TCO

22 High resolution reservoir simulator in the cloud
By the largest oilfield services company in the world Challenge Find the right cloud solution Models keep getting larger Ever-growing demand for more and faster compute resources Difficulty of having access to the right type of resources Strategy Accessibility and scalability True HPC capabilities on Azure Results Schlumberger’s customers can now submit simulations to the cloud directly from Petrel Capabilities are available on a subscription-based model Schlumberger: Schlumberger is the largest oilfield services company in the world. It employs 115,000 people in more than 85 countries. Their products and services span from exploration through production. It supplies technology, integrated project management and information solutions to customers working in the oil and gas industry worldwide. This is Schlumberger’s first fully commercial software as a service (SaaS) solution for cloud, and they decided that Azure is the right cloud platform to host it on for the initial launch in North America and Europe. Challenge: INTERSECT is their high-resolution reservoir simulator for large and complex engineering models. These models have not been traditionally possible in the cloud (at least not efficiently) because they require specialized hardware and an array of different technologies that are not part of a “normal” cloud offering. But, at the same time, these models are starting to outgrow the on-premises resources that companies and service providers have at their disposal So, in essence, SLB & his customers are confronted with is a combination of an ever-growing demand for more and faster compute resources (driven by the ever-growing characteristic of the simulations), paired with the difficulty of having access to the right type of resources. Fortunately, this is just the kind of challenge that we have been working hard to solve in Azure. Strategy: The accessibility and scalability offered by cloud computing should be the answer to the challenges mentioned earlier, but before this is true, the backend infrastructure must have the right technology to enable the simulator to really perform.  This is where Schlumberger and Microsoft working together have created and delivered the most appropriate technology solution Results: Thanks to the availability of this service, Schlumberger’s customers can now submit simulations to the cloud directly from Petrel, the industry leading reservoir modelling tool and the simulation environment that Schlumberger provides for generating the models, as well as for visualizing the results. This means that a reservoir engineer never leaves the environment that she or he is familiar with. The engineer just needs to select where the simulation should run: locally, on an on-premises cluster, or on the cloud. After the simulation has run, the engineer can analyze the results back in Petrel, as usual. But perhaps what is most exciting about this solution is that these capabilities are available on a subscription-based model, making it possible for companies of all sizes, on all parts of the world—including those customers who are unable to afford their own HPC environment—to have access to high science, on an enterprise-class simulation solution like INTERSECT. This, in particular, is one of the greatest advantages of a cloud solution. Learn more To learn more about Big Compute solutions on Azure, see: To learn more about Schlumberger INTERSECT Simulator in the Cloud, see: To learn more about Schlumberger INTERSECT, see: Azure Batch

23 Media and Entertainment
Rendering 3D production Movies and gaming industry HD Images Need many cores Applications Software ISV Rendering ISV Linux/Windows Operations Disaster Recovery Optimal TCO Flexibility for business line Major Software ISVs : Autodesk: 3DS max, Maya, Lightwave OSS : Blender Major Rendering ISVs: RenderMan – Pixar Arnold – Solid Angle V-Ray – Chaos Group Maxwell – Next Limit CINEMA 4D – Cinema4D Modo, Katana, Nuke – The Foundry Mental Ray – NVIDIA FLEXIBLE : Ramp up & down quickly for a lumpy, unpredictable and potentially long running Workloads SECURED : Data/models are secured TCO = Pay only when you deploy ressources Microsoft Azure based solution enables Media Companies to deploy and manage clusters for physical simulation in a flexible, secure and controlled fashion for optimal TCO

24 Animated feature film rendering
PROAN Entertainment leverages Azure Batch for rendering in the cloud for full length 3D animation feature Challenge 90 minute 3D animated feature film Would have taken over 13 years to render using existing on-premises cluster Find the best rendering program within the budget Strategy Blender to Azure batch plugin Close relationship with customer Azure compute Results 1,600 cores at peaks 4 months of production and 3 months in post-production Rendered entirely in Azure using Blender in about 3 months PROAN Entertainment, a division of Grupo PROAN and a leading digital studio based in Mexico ‘Pepito – La Película’

25 Other verticals Bio/genetics Academics Other
Exploration of new drugs Gene exploration Linux Academics Research in any topics Linux Other Need compute power Solving many problems Solving complex problems Microsoft Azure based solution enable access to unlimited power in a flexible, secured manner, at your terms with optimal TCO

26 Molecular modeling in Microsoft Azure
By one of the largest chemical companies in the world Challenge Reduce compute costs while continuing growth Custom application written by DuPont Scientist on molecular modeling Strategy Opportunity found by CSA during Azure briefing Engage Azure black- belt + engineering early in the sales cycle Results Mid-Atlantic district is going to use this success and apply learnings to other accounts This project facilitated Azure HPC deep dive for their R&D scientist which will drive more workloads DuPont Customer: One of the largest chemical companies in the world US based, $35 billion in annual revenue, 65,000 employees across 90 countries Team: Louis Berman CSA, Steve Roach Big Compute TSP Black belt, John Fitch ATS Workload Custom application written by DuPont Scientist on molecular modeling Customer willingly recompiled application to Intel MPI on Azure Using ARM templates spun up 2 clusters 100 A9 instances each (1600 cores) One MPI job/application running across 1600 cores for weeks 3,215,708 compute core hours i.e. $241,178 dollars consumption in less than 3 months

27 Conclusion and additional resources
10/23/2017 5:36 AM On VMs DIY Customers setting up separate cloud environment for additional capacity and new workloads Burst Burst to cloud for more capacity Utilize existing scheduler to burst to Azure and use compute on demand SaaS ISVs and Managed service providers launch their own SaaS solutions. Pay for compute and application software per hour Marketplace Utilize the scale of Azure Applications and compute capacity sold through Azure marketplace All above scenarios are truly pay as you go without any upfront cost of infrastructure. Resources Azure Big Compute Azure Batch Overview Azure Batch Technical Overview HPC Pack for Linux Azure compute-intensive instances - A8, A9, A10 and A11 specs Microsoft MPI (Message Passing Interface) Documentation resources for HPC & Batch Azure HPC and Batch Blog © 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

28 Big Compute case studies
10/23/2017 5:36 AM Milliman (actuarial and financial risk modeling using Batch): Towers Watson (actuarial and financial risk modeling using Batch): ANEO (SI for financial services using HPC Pack IaaS on Azure): d3VIEW (SI for engineering simulations providing SaaS with A8/A9 VMs): Schlumberger (ISV for oil & gas engineering modeling providing SaaS with A8/A9 VMs): Microsoft Research FaST-LMM project (genetics research using HPC Pack IaaS on Azure): Ludwig Institute of Cancer Research (genetics research using HPC Pack on premises): Check for new ones by going here: Search for “HPC”: Search for “Big Compute”: compute” Search for “HPC Pack”: Pack” © 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

29 10/23/2017 5:36 AM © 2016 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. © 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.


Download ppt "Azure HPC Solution pitch deck."

Similar presentations


Ads by Google