Virtualizácia Dátového centra v Slovak Telekom e FOCUS konferencia TRENDY, STRATEGIE A IT TECHNOLOGIE pre 2008 až 2010 Doc. Ing. Oto Malý, PhD Vrchný riaditeľ Divízie IT Bratislava, 5.3.2008 February, 2008 1
Agenda ST who we are in Slovakia and in DT group People are driver and - Technology is only a tool Storage story living reality Servers story - new concept and challenge February, 2008 2
ST who we are in Slovakia and in DT group One of 3 largest players in Slovak Telecommunication market Only one of DT Family members without revenue drop YoY unique even in the European market Modern product strategy Leader in customer satisfaction - TRIM index in DT group and in Europe Technology standardization based on DT standards Leader inside DT affiliates in IT operational efficiency and maturity Technology evolution is long-term a basis for innovations, effectiveness, efficiency, productivity and operational stability February, 2008 3
People are driver - Technology is only a tool Department for Infrastructural Operation + Department for IT Strategy Conceptual design Financial study Midterm Storage Management Strategy Server virtualization strategy Common ST IT and HP team Department for Infrastructural Operation / Team for System Support + HP professional services and HP Servers and storages High level technical design Detailed technical design Migration concept Test cases Real implementation / Storage story Solution strategy, Analysis and design running, test environment in creation/ Server story February, 2008 4
Storage optimization story February, 2008 5
Storage Management Strategy Mid term scope Executive Summary Status of storage infrastructure before Storages installed in 2002 and 2003 were going out of warranty and support Storage demand determined by business requirements. Storage demand classified into 3 categories: Business Critical, Critical, Ordinary. Disk utilization much higher then 80% of allocated space (optimal utilization ought to be at 80%). Higher utilization would lead to rapid loss of performance. Primal storage infrastructure and capacity could not cover the anticipated annual capacity business demand via upgrade (ongoing and new projects, Disaster recovery/sox compliance, regulatory requirements) Traditional extensive grows was out of real and reasonable financial possibilities February, 2008 6 Solution reengineering and optimization was strongly required
Initial status As is: To be: Data on the old disk arrays To restructure data and migrate to HP XP12000 DC Kosice DC Presov IT IT environment environment IT IT environment environment February, 2008 7
Mid -Term Storage Management Strategy Executive Summary Mid-Term Storage Management Strategy a)introduce virtualization program as a best market practice b)scale storage solution by capacity upgrade and optimize, CAPEX and OPEX efficiency c) Redefine architectural principles and IT standards, simplify and optimize SAN structure in terms of capacity and utilization efficiency d)reduce operational storage demands by optimized usage of old disk arrays, applying optimized IT standards solution and using cheaper Mid range solution e)cover future capacity gap by systematic YoY reengineering of storage space and process, utilizing above measures 2 years program February, 2008 8
Mid -Term Storage Management Strategy Consideration Operational IT efficiency = significant decrease of TCO for solutions and services Standard cost-cutting approaches have been applied and were to greatest extent more-or-less exhausted, however NEW drivers can enable potential savings, namely: Exploitation of newly available technologies and solutions Qualitative (careful) transition to less robust IT architecture and standards, resulting in: Adjusted SLA (process of all stakeholders) and lower TCO In order not to cause disruptions of IT operations, Disaster Recovery and Business Continuity support (in other words existing SLA level), full commitment of stakeholders: Businesses, Controlling and IT is needed to reengineer SLA to the less robust architecture. This process (and solution deployment) is not attainable in few weeks time, it is a mid-term process. The measurable KPI is TCO today versus TCO target model (Gartner source: Standard approaches INDUSTRY statistics for IT budget savings over 5 Year period: revision of contracts with vendors (84%), HC reduction (42%), licenses reduction (32%), centralization of servers (32%), centralization of development (26%), etc) February, 2008 9
Mid -Term Storage Management Strategy Consideration There are 7 basic drivers for architecture and standard optimization Virtualization Implemented in Disk array optimization program Disks with lower speed (rpm) Have been tested in HP project and are to be used as standard for secondary site Higher RAID Have been tested in HP project. To be implemented in Disk array optimization program immediately Business continuity volumes Provision of BCV on cheaper technology > To be tested and implemented with virtualization Remote replicas Provide mirroring on cheaper technology > To be tested and implemented in Disk array optimization program Back up reengineering New back-up engineering and SW solution together with virtualized storage capacity Server consolidation and Virtualization Accelerator of innovations February, 2008 and 10 efficiency next step
Mid -Term Storage Management Strategy Short Term Actions 2006/7 Resolve capacity gap and expiring warranty: Extend installed High-End HP storage capacity according to optimized demand Implement only lower speed disks for secondary site Continue Disk array optimization program for SAN based on RAID reengineering 3/4 and 7/8 according to criticality of site and relevant data Deploy cheaper Mid Range disk technology (like EVA) for applications of appropriate business criticality Prolong post-warranty support for old technology under revised and optimized conditions Assess impact of deployment of archiving solution and old data archiving February, 2008 11
Mid -Term Storage Management Strategy Mid Term Actions 2007/9 Based on best market practice start virtualization program Analyze and test Cross platform virtualization (XP and EVA + others) Test Cross platform virtualization, Utilization of old disk arrays for Business Copy and Continuous Access As an alternative test HP virtualization Cross HP platform virtualization (XP and EVA) Implement virtualization solution Start Business Driven Disaster recovery and Business Continuity reengineering program Start Continuous Access reengineering on Mid range (cheaper) technology Reengineer back up solution optimize back up processes and Review allocation of applications to appropriate range of disk arrays Formalize demand/requirement process (incl. support tools) with focus on prioritization and capacity allocation management Implement reviewed internal regulations and standards driving storage infrastructure and setup balancing cost versus business impact February, 2008 12
Migration to new disk arrays HP XP12000 DC Presov migration of 5,4TB preparation, systems patching + SAN 20 days + 5days, data migration 5days DC Kosice migration 10,5TB preparation, systems patching + SAN 20 days + 5days, data migration 12days DC Kosice CA DC Presov IT IT environment environment 42 42 servers servers (*) (*) IT IT environment environment 11 11 servers servers (*) (*) * - connected to XP February, 2008 13
Capacity upgrade of XP12000 and virtualization Project plan Start: 01/2007 Finish: 08/2007 fulfilled Project finished in time, in scope and in budget frame. Approved main milestones - Capacity upgrade of XP12000 in DC Kosice and DC Presov - Introduction of virtualization technology on the platform of disk array of EVA8000 Activities - Upgrade of hardware XP12000 - Installation of EVA8000 - Virtualization of EVA disk space / enhancement of the capacity via virtualization - Preparation of migration - BC (business copy) from XP12000 to EVA8000 - Testing of backup by means of Dataprotector Zero Downtime Back up (DP ZDB) - Migration of BC (Business Copy) from XP12000 to EVA8000 (virtualization and optimization of data classification methodology) - Data migration from EMC to XP12000 - Performance testing of disk arrays and data availability for Application owners - Removing old connections to servers - Operational tuning February, 2008 14
Final status XP12000 is used as an Virtualisation Engine enabling Backup Data to go to External Virtualised midrange Storage on EVA8000 Prešov Košice SAN SAN SAN SAN EVA8000 CHIP CHIP XP12000 Continuous Access EVA8000 CHIP CHIP XP12000 External Storage Business Copy External Storage Business Copy 3x Direct Connection 3x Direct Connection February, 2008 15
Main project business achievements HP storage reengineering project leads to significant billing improvements Business segment billing run average 3:48 hours against 10 hours before Residential Segment billing run average 12:45hours against 30 hours before February, 2008 16
Results Higher availability based on new technology Better performance SAP - improvement of average time for disk operations by more than 50% Reporting DB had time schedule problems before migration, there is enough defined work time. (Daily, weekly, monthly 30%) Shorter response to requirements (provision of new/change disk space) Better management - implementation of virtualization (disk arrays XP see other disk arrays HP EVA as their own disks) Consolidated and optimized backups (from D2T to D2D2T)** Lower backup frequency, faster backup, cheaper backup (less backup drives) Better and faster restore (in case of IS release, hot fix, upgrade, disaster) Business copies (used for backups) use cheaper disks (midrange arrays in final status, high-end in a past) Run of Test/Dev environment on lower level support storages, lower cost in operation ** - D2D = disk to disk, D2D2T = disk to disk to tape February, 2008 17
Proposed next steps Reuse of old disks in low cost operational mode for non critical use Replacement of old disk arrays (out-of warranty) To change technology (mid-term or long-term vision) To use for test, dev environment virtualized disk space from EVA Ad-hoc requirement (temporary) to cover by BC volumes Prepare capacity management and standard storage volume enhancement 2009++ February, 2008 18
Disk Array Life Cycle Management Implementation Team ST Team HP Team Project sponsor: Oto Malý Project owner: Gejza Tomko Project manager: Juraj Lengyel Project delivery manager: Peter Izsóf Project members: Peter Kiss Ľuboš Polgár Dana Nagyová Martin Gablik Yveta Kollárová February, 2008 19
Server Virtualization Story February, 2008 20
Executive Summary Why does ST need virtualized HP-UX server infrastructure? BC Applications running on server environments older then 5 years EURO transition project requires, space, performance and reliability Majority of servers running out of warranty Strategic migration from PA-RISC to Itanium platform due to PA-RISC end of sales in 2008 Improve server efficiency by creating unified, scalable infrastructure with flexible pool of shared resources Strategic migration from HP-UX 11 v1 to newer version due to end of sales in 2008 for old one Simplification management of server infrastructure Increasing support costs of existing servers after standard warranty to be eliminated Reducing IT costs is required February, 2008 21
Benefits of virtualization Unified and flexible infrastructure easier to manage easier to deploy application shared pool of computing resources applications benefit from virtualized environment Cost saving on licensing less number of CPUs (more powerfull CPU) and better utilization e.g. better Oracle licensing factor (0.75 PA-RISC vs. 0.5 Itanium) Updated infrastructure supported hardware for long period supported OS for long period (ability to use new features of the newest OS) February, 2008 22
Implementation Phase I Quick Wins implementation virtualization of server infrastructure for SAP, DAS, UTP information systems Phase II, Phase III Virtualization of the rest of Unix server infrastructure February, 2008 23
Phase I-III Summary Investment + Opex 5-y Current 100% Phase I-III Total 66,5% February, 2008 24
Phase I Quick Wins Three scenarios differing in target platform sizing and migration approach: 1. Box-to-box: Current servers will be migrated in 1:1 fashion (target hardware is a mix of low-end and mid-range servers of various models) 2. Virtualized - option I: Current servers will be virtualized on mid-range platform (rx8640) with exactly the same performance as current applications need 3. Virtualized option II: Current servers will be virtualized on mid-range (rx8640) with reserved capacity to cover migration needs of other applications February, 2008 25
Scenario Evaluation (5-Year TCO) February, 2008 26
Phase I - 5 Years IT Spend by Category February, 2008 27
Recommended scenario for Phase I (Virtualized option II) This option will cover virtualized infrastructure for SAP, DAS, UTP plus additional capacity reserved for starting virtualization in Phase II. February, 2008 28
Target solution architecture for Phase I (Virtualized option II) SITE A Košice SITE B Prešov February, 2008 29
Phase II and Phase III Realization in 2008-2009 Phase II (2008) - Virtualization of 27 servers onto 4 rx8640 Phase III (2009) - Virtualization of 39 servers onto 6 rx8640 Migration Phase I - III in 2008-2010 February, 2008 30
Phase II and Phase III February, 2008 31
Phase II - 5 Years IT Spend by Category February, 2008 32
Phase III - 5 Years IT Spend by Category February, 2008 33
TCO/ROI Assumptions TCO/ROI method is NPV (Net Present Value) Migration cost estimated as 20% - 30% of hw and sw cost Annual refreshment cost of current technology: 35% of servers refreshed cost is 50k USD per refreshed server Refreshment of new environment in last 2 of 5 years Real migration timeline has not been taken into account as it depends on: ST s business needs (outage windows, impact, new releases...) Application interdependencies February, 2008 34
Thank you for attention Question??? February, 2008 35