UBS Group Technology Public UBS Data Center Efficiency Strategy Steps in Efficiency, Automation leading to enabling the Cloud October 18, 2011
UBS one of the leading financial firms UBS draws on its 150-year heritage to serve private, institutional and corporate clients worldwide, as well as retail clients in Switzerland. We combine our wealth management, investment banking and asset management businesses with our Swiss operations to deliver superior financial solutions. UBS is present in all major financial centers worldwide. Headquartered in Zurich and Basel, Switzerland, UBS has offices in over 50 countries. UBS employs about 65,000 people around the world in the four regions as depicted below.. Under Swiss company law, UBS is organized as an Aktiengesellschaft, a corporation that has issued shares of common stock to investors. UBS AG is the parent company of the UBS Group (Group). The operational structure of the Group comprises the Corporate Center and four business divisions: Wealth Management & Swiss Bank, Wealth Management Americas, Global Asset Management and the Investment Bank Its shares are listed on the SIX Swiss Exchange and the New York Stock Exchange (NYSE). 1
UBS Data Center Services Facts and Figures Responsibilities: UBS Data Center Services (DCS) provides robust, cost efficient and scalable data center capacity to the hosted computer and communication environment, as well as the services to operate the data centers. DCS maximizes operational efficiency and reduces risk by minimizing the number of data centers globally. Robust: Engineered to meet the firm s operating standards for power, cooling, connectivity and physical integrity Cost Efficient: Efficient buy/build decisions based on the economics of provisioning and operating data center capacity Scalable: Agreed workload projections into the future and plans in place allowing us to meet these needs Minimal Numbers: Located in each major region, minimal number to maximize operational efficiency and reduce risk Business Continuity Program (BCP): Data center capacity to support regional and inter-regional BCP requirements Data centers in 12 countries hosting (in % of used capacity): Servers 59% 2011 Mainframes/Midrange Systems 6% Storage 22% Network 12% Other equipment 1% 15 MW data center IT load creating an average utilization of 63% Annual DC Cost Spread: Power 25% Real Estate Costs 53% Operating costs 14% Other Costs 8 % Americas 8 UK 4 Continental Europe 3 Switzerland 5 APAC 12 Total Data Centers: 32 2
UBS APAC Data Centers Facts and Figures Core Data Center service is designed to provide high availability to UBS s main server footprint and currently runs in the following 8 Data Centers in APAC Beijing Seoul Mumbai Tokyo 1 Taipei Tokyo 2 Hong Kong 1 Hong Kong 2 Singapore 1 Singapore 2 Key Stats Core DC's 8 (4x2) Edge DC's 4 APAC ~4.1MW Melbourne Sydney
Data Center Energy Efficiency Measures Energy efficiency avoids costs and reduces carbon footprint. The potential for more IT energy efficiency is high. In 2010, IT consumed approximately 55% of the total UBS intermediate electricity usage, of which three quarters are consumed in UBS data centers Improving data center efficiency will make a difference!! Start and manage a data center efficiency program across all data centers Improve airflow Optimize operating parameters Implement equipment adjustments and modifications Conduct feasibility studies and visionary investigations Consolidate datacenter portfolio Support data center users in their efficiency programs to reduce energy and/or to reduce CO2 emissions Server and Storage Virtualization (improving utilization) Energy settings on desktop management (sleep/hibernate) Hardware lifecycle programs (increased energy efficiency in new hardware) Implement/allow for video conferencing (substantially reduce travel) Printer reduction program Grid / Cloud computing (application remediation, resiliency consideration) CPU clock speed throttling for development servers 1,500 1,000 500 - (500) (1,000) (1,500) (2,000) (2,500) (3,000) (3,500) 2010 DEEP Actual Power Reduction Including only Refresh Activity (Fig 1) Power Installed Power Decommed Net Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Target 2010 DEEP Actual Net Power Consumption Including Refresh and New Demand Activity (Fig 2) 2,000 1,500 1,000 500 - (500) (1,000) Power Installed (1,500) Power Decommed (2,000) Net Jan-10 Feb-10 Mar- Apr-10 May- Jun-10 Jul-10 Aug- Sep-10 Oct-10 Nov- Dec-10 10 10 10 10 4
UBS Virtualization Efforts Distributed Engineering & Efficiency Program (DEEP): UBS has used VMware and other Virtualization technologies for several years in a range of forms (server virtualization, VDI), but we needed an across the board push to make it the dominant platform. Virtualization is an important component of future plans which include key goals such as Automation and eventually Cloud Integration DEEP is a multi-year and multi-stage program that is focused on our Compute Platform Evolution Large Pod = 280 to 840 VM's Medium Pod = 80 to 280 VM's Small Pod = up to 36 VM's Hybrid Pod 5
Path to Data Center Automation Enabling the Cloud How do we propose to deliver automation? Our approach emphasizes flexibility, scalability, and constant quality reviews We will focus on consistent evaluation & prioritization of automation candidates Delivery of automation will be via a sequence of managed scope efforts focused on controlling risk & tracking efficiencies Maturity Layer Goals: Targets Automation Layer Area of Automation Focus Saving Time & Reducing Errors: Rapid Deployment of Fully Pre-configured Virtual Machines (self service Dev env and Databases) Physical Server initial configuration management (transition from a largely manual process) Fast automated database failover for BCM scenarios Improve Stability & Utilization: Data mining of allocation, performance, alert, and incident data for trend and pattern analysis Incident correlation analysis Dynamic resource allocation and right-sizing in Virtual environments Target Automation Maturity Model Level of Automation 6 End State Cloud-based automation Business value Dynamic multi environment automation 5 Interim Data center automation IT efficiency Lights-out computing 4 Integrated service automation Business efficiency Business-focused automation 3 Intelligent workload Automation IT complexity Self-learning based automation 2 Financial Industry Event-led workload automation Process-focus Policy-based task level automation 1 Automated batch processing Task focus Basic task-level 6
Implications of Cloud Based Services for Financial Services Questions How do I secure, track, location control and comprehensively Audit what s in the cloud? Can cloud be a tiered offering as it develops Public, Semi-Private or Private? How do I organize my data so it s classification for a tiered service? I know why I need it and where to use it but can t until full range of services and controls are available What controls do I need Location Guarantee, Comprehensive Audit, Privacy, Availability? Regulators Regulatory certification for cloud enabled solution controls needs clearer outlining by country regulatory bodies Private Cloud Is this an more than just a private network with mature virtualization and applications and how does this leverage a crowd-sourced model? 1.Develop a set financial requirements definition for each country Header 2. Collaborate within the finance industry to leverage the crowdsource model 3. Engage a list of third party providers to develop a service Semi-Private Cloud Is this a specific vertical semi private cloud, in this case built for the financial industry to address control issues and certified by the regulators. Use a crowd-source or consortium model? Public Cloud Were all using it anyway; many retail financial services are all done on the internet now (Paypal, Etrade etc.) But it s a best effort service by design. 4. Engage the Regulatory body for certification of solution controls 5. Connect and use cloud service 7
Data Center Placement Optimization Market Maturity Questions How do we deal with Proximity Hosting and Market Collocation? Market data and market execution space is becoming fragmented as providers compete and develop new products We cannot ignore this space but it is fragmenting our server base and challenging the DC placement & consolidation strategy It is expensive to locate inside a Colo and connect it back to main Data Centers Industry needs an "Econo-metric" model for Colo Mature markets presence of competition for execution combined with Infrastructure to facilitate a platform building a financial ecosystem Developing markets in the process of developing a mature market strategy but either hasn t built the infrastructure or the economic structure is limiting Immature markets no current plans to develop into mature market category in the near term. Relies on old revenue models. Algo Market Proximity General Trading Market Proximity Algo General Trading Eco System Network Cost Network Cost Data Centre Cost Shared Service Clearing Central Service Data Centre Cost Shared Service Clearing Eco System Central Service Not Header Mature Developing Mature
In Conclusion Traditional Data Center Efficiency Programs (consolidation, virtualization, and release of underutilized resource) have been very cost-effective in the past, but the role of the Data Center itself is changing. Cloud Services and Colo facilities offer the Industry the opportunity to avoid each participant having to building large and expensive fortresses to fight yesterdays wars, however the transition path to these newer offerings still requires considerable thought and commitment. Key milestones on this path include: a mature Virtualization Program, an Automation framework to handle dynamic loads and risks, and a clear market execution strategy (incl. Regulatory Guidance) for where we REALLY need to be. The Mature Markets in APAC offer the opportunity for Financial firms to develop right-sized blended models that will allow us to hit these milestones. Ultimately, we expect the industry to settle into blended models that combine the "appropriate mix" of Data Center, Colo, and Cloud Model footprints that meet both regulatory and customer requirements. It's as important to look at the goals of Green IT and Managing Emissions and not just focus on execution. A blended model where people dynamically use only the resources they need will yield more Green IT benefits than just focusing on who can build the greenest fortress. 9