Virtualisation Consolidation and Application Aware Storage The New Mandates for

Virtualisation, Consolidation and Application-Aware Storage The New Mandates for Datacentre Efficiency Adrian Groeneveld – Product Marketing, EMEA 2 Today’s Business Landscape Data is still growing There is increasing economic uncertainty We have all been impacted by the credit crunch Our financial belts are being tightened Competition is getting tougher The business is demanding every bit of assistance it can get (more applications and intelligence) Applications need infrastructure Infrastructure costs money 3 Today’s Storage Landscape Multiple Technologies Storage Silos Training Costs Service Costs Admin Time Complexity Expensive to Manage Expensive to Own Low Utilisation Power and Space Point Products Performance falls behind capacity Space constraints Hidden costs 4 How are Your Resources Being Used? Poor server utilisation Fragmented or silo d infrastructure silo’d Project driven infrastructure deployments Performance at the cost of inefficiency What about the power? What about the cooling? What about the space? 5 What Do Applications Need First? Capacity or Performance? Infrastructure decisions for applications driven by SLAs based around performance and availability p y Capacity without performance means missed application SLAs Storage decisions should start with the applications 6 Understanding the Industry Problem The Problem: Getting optimal p g p performance & capacity utilisation p y from your array 9GB Fibre Channel 15K RPM Drive circa 1999 Approx 190 IOPS 300GB Fibre Channel 15K RPM Drive circa 2007 Approx 190 IOPS The Application’s need for performance, not capacity is creating an industry average utilisation sub 40% - but you are still buying arrays on one metric - $/GB. 7 Evolution Driving Change Storage requirements for the Next Generation Data Centre: – Asset consolidation based on modular architectures – Automated operations using highly integrated IT Virtualization – Unified storage protocols for server and storage connectivity – Service-oriented approach for application delivery – Advanced low-cost scalability – Non-disruptive upgrades and error handling – A t Automated data classification, retention and management t dd t l ifi ti t ti d t – Shift of emphasis from $/GB to $/IOP and $/Usable GB Source: various analyst firms 8 9 Application-Aware Storage Key Differentiation Differentiated Storage Services aligned to Application Needs – All LUNs should not be created equal Highest utilisation Rates in the Industry – Buy less storage - use it more efficiently Optimal Performance, No utilisation Penalty p , y – Remove the capacity versus spindle decision Provision Storage like Servers – You like VMWare for servers?...you’ll like Pillar for storage Dynamic storage resource (re)assignment – Eas to change priorities based on b siness needs Easy business 10 The Pillar Axiom Storage Platform One Single Consolidated System – SAN, NAS or Both Service Level A Tier 1 Application – Multiple workloads and performance levels – M d l growth Modular th Tier 2 Application Service Level B – Scale capacity AND performance – Hi h availability th High il bilit throughout h t – Six click provisioning Tier 3 Application Service Level C – Flexible data protection and disaster recovery 11 12 System Wide “Quality of Service” Pilot Slammer Embedded Switching Brick Software Virtualization maps QoS policy management Predictive modeling I/O prioritization I/O allocation Cache prioritization Dynamic cache algorithms Slammer load balancing Fabric bandwidth prioritization and allocation RAID configuration Mirroring Brick load balancing Disk access prioritization Data layout Disk block partitioning Disk block prioritization Minimum % of Queue Allocation M Medium A Archive High Low High Medium Low Archive Logical Volumes Data Layout & Block Prioritization Bands 13 What’s the Approach The Axiom Problem? 10 QoS Level 1 Application A Lower performance means missed SLA Performance of Provisioned Storage Per rformanc ce Higher performance means wasted money 5 QoS Level 2 Application B 0 14 Pillar Axiom System Scalability Scalable Capacity and Performance in a Single Virtual Storage Grid Storage CPUs IOPS TB IOPS TB IOPS TB IOPS TB x 64 Bandwidth IOPs Capacity 15 Easy Provisioning Application Profiles 16 Most efficient storage system in the industry Shared resource with minimal contention (QoS) Thin Provisioning with QoS features Best Efficiency Quotient (EQ) vs. competition 17 Achieving Higher Utilisation Consolidation – Eliminate multiple physical tiers – Eliminate separate archive and content based storage Intelligent data layout g y – More intelligent use of spindles and available capacity – Distributed RAID to deliver better performance from p existing capacity Capacity on demand g p y growth – Intelligent thin provisioning across all data tiers or p priorities 18 Utilisation Without Compromise “We “W are consistently i t tl delivering 80% utilisation. g GUARANTEED!” 19 Linear VM performance scaling –VMware, Xen or HyperV H perV Supports 50% more VMs than other systems VMotion VM ti or X M ti optimized XenMotion ti i d Site Recovery Manager (SRM) ready Provides “Virtualization Insurance” P id “Vi t li ti I ” 20 Storage in a Virtualised Environment Typical approach to storage in virtualised environment – Centralised storage array – All virtual machines treated the same – One size fits all 21 The Issue of Capacity vs. Spindles 300GB Drives set as Raid 5 (4+1) provides a 1 2TB LUN 1.2TB Multiple VM’s are usually supported by this “stacked array” Result: 100% random I/O, overloading cache The Solution – Application Aware Storage Each LUN is striped across four RAID 5 disk groups Spindle count increased to a least 24 disks Performance of individual LUN optimized for each VM 22 Pillar in a Virtualised Environment Capital Savings – Optimization Removes the Capacity “white space” white-space Capacity lost to underutilised Server Capacity lost to various RAID configurations – L Lower Initial Acquisition Cost due to higher Utilisation I iti l A i iti C t d t hi h Utili ti $/GB Value assigned for each VM $/IOP Value assigned for each VM Thin Provisioning reduces initial capacity need 23 Application-Aware Storage Virtualisation Virtual Server Environment E i t Physical Server Pillar Axiom Storage Ser ices Services Appropriate storage SLA mapped against VM Service Level A Virtual Machine 1 Performance mapped against virtual application Storage instantly tuned for Virtual Machine 2 Service Level B Virtual Machine 3 CPU (Priority) Memory (Cache) Network (Priority) Service Level C peaks and troughs in demand CPU (Priority) Disk Memory (Cache) (Layout) 24 Superior Performance NAS/SAN : TPC-C TPC H TPC C TPC-H 1MB Wide Stripe – Unique in the industry Pooled P l d RAID 10 – U i Unique i the industry in th i d t Plug in to Oracle Enterprise Manager ASM, ILM, BI/DW ECM ASM ILM BI/DW, ECM, RAC optimization ti i ti 25 Linear Scalability of 10g/11g RAC Scaling RAC – Scaling ASM Scale storage controllers Scalable RAC cluster Total CPU Server Count = 18 Scale capacity and RAID controllers 26 Simple Deployment and Management Oracle validated Configs with Pillar Axiom Pill A i Oracle Enterprise Manager Plug-in Plug in 27 Optimised Grid, Data Warehouse and Applications Single Data Warehouse with performance differentiation Managed via Oracle Enterprise Manager for lower Operating Costs Native 1MB Stripes for superior performance under ASM Flashback 15K Recovery Area Operational 15K 10g/11g Data Warehouse 1TB - Archive B >1 days 180 Data Warehouse D t W h >180 days Best $/GB and Tightest integration with Oracle in the industry. Hot Stand-by 15K 10g/11g 28 28 Application-Aware Storage from Pillar 29 Delivering Tomorrow’s Data Centre Today Storage requirements for the Next Generation Data Centre: – Asset consolidation based on modular architectures – Automated operations using highly integrated IT Virtualization – Unified storage protocols for server and storage connectivity – Service-oriented approach for application delivery – Advanced low-cost scalability – Non-disruptive upgrades and error handling – A t Automated data classification, retention and management t dd t l ifi ti t ti d t – Shift of emphasis from $/GB to $/IOP and $/Usable GB Source: various analyst firms 30 Thank You Questions & Answers