Private Cloud on Pure Systems Smarter Computing Solutions 2012 Briefing
Overwhelming IT Complexity Drives Cost Up $217 B $247 B est. $100 B 8% 29% 63% $130 B 11% 43% 46% $175 B 16% 33% 51% 12% 22% 66% 11% 20% 69% 1996 2001 2006 2011 2013 New Server spending Electricity and cooling costs Server management and admin costs Cloud computing a way to help lower costs! Source: IDC, 2012 Private Cloud on Pure Systems v1.2 2
Introducing PureSystems - A New Family Of Expert Integrated Systems That Drives Costs Down Built-in expertise to address complex business and operational tasks automatically Integration by design to accelerate system setup and application management Simplified experience from purchase to maintenance PureFlex System Pre-integrated and optimized infrastructure Management integration across compute, storage and networking both physical and virtual No compromise design with system level upgradeability Designed for cloud with 3 flexibility and simplicity PureApplication System Optimized for performance and virtualized for efficiency Designed for transactional web applications and enabled for cloud Application-aware workload management Private Cloud on Pure Systems v1.2 3
IBM Has A Wide Range Of Workload Optimized Systems Tuned For Specific Workloads Flexible Focused Custom Built with Optimized Components Expert Integrated Systems PureSystems Appliances DB2 on Power and Storwize IBM Smart Analytics on Power IBM PureFlex System IBM PureApplication System Netezza Private Cloud on Pure Systems v1.2 4
PureFlex Starts With Choices Of Compute Nodes And Operating Systems No compromise designs for full performance Support multiple architectures: POWER7 or x86 nodes Support for applications across 4 operating environments Support up to 14 compute nodes per chassis Support up to 4 racks for the largest enterprise needs Linux = SUSE 11 and RHEL 5 or 6 Windows = 2008 R2 AIX = 6 and 7 IBM I = 6.1 and 7.1 2-socket 2-16 cores 32 threads Intel Sandy Bridge-EP Up to 8 cores per socket Up to 16 threads/socket with hyper-threading 2-socket 4-16 cores 64 threads 4-socket 4-32 cores 128 threads IBM POWER7 8 cores per socket 32 threads / socket Private Cloud on Pure Systems v1.2 5
PureFlex Integrates More Choices For Hypervisors, Networking And Storage Multiple Hypervisors supported PowerVM, VMWare, KVM, Hyper-V Cloud supported in Standard and Enterprise systems IBM SmartCloud Entry Advanced Networking Choices Ethernet Scalable 10Gb Switch / 40Gb Uplinks & Scalable 1Gb Switch 10Gb Uplinks 1/10Gb Pass Thru Fibre Channel (16Gb, 8Gb and FCoE) Infiniband (QDR and FDR) Enterprise Class Block Storage (Storwize V7000) Enterprise class storage virtuialization and copy services Solid state storage optimized with Easy Tier SSD and HDD support Direct attach PCI storage expansion Attaches directly to compute nodes Expand storage capacity available to nodes FCoE = Fibre Channel over Ethernet QDR = Infiniband Quad Data Rate FDR = Infiniband Fourteen Data Rate Private Cloud on Pure Systems v1.2 6
PureFlex Also Delivers A Simplified Experience By Integrating All The Components PureFlex Building Blocks Chassis Compute Nodes Storage Nodes Management Appliance Networking Expansion 14 half-wide bays for nodes x86 2S Power 2S/4S V7000Expansion in/out of chassis Flex System Manager (FSM) 10/40GbE, FCoE, IB 8/16Gb FC PCIe, Dedicated Storage Pre-Configured Chassis + Choice of Compute Nodes + Integrated Networking + Integrated Storage + Integrated Management Up to 640 Cores PDU PDU Cable ingress / egress BNT 64 PT Enet SW BNT 64 PT Enet SW V 7000 V 7000 V 7000 V 7000 Expansion Controller Expansion Controller Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute MGMT Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute Node Compute MGMTNode Compute Node Cable ingress / egress PDU PDU 42 U 41 U 40 U 39 U 38 U 37 U 36 U 35 U 34 U 33 U 32 U 31 U 30 U 29 U 28 U 27 U 26 U 25 U 24 U 23 U 22 U 21 U 20 U 19 U 18 U 17 U 16 U 15 U 14 U 13 U 12 U 11 U 10 U 9 U 8 U 7 U 6 U 5 U 4 U 3 U 2 U 1 U Private Cloud on Pure Systems v1.2 7
PureFlex Offers Unlimited Compute Power Choices Compared To Oracle Exalogic Number of Cores IBM PureFlex 4 Cores to 3,300+ PDU PDU PDU PDUFrom 4-640 Increments of 2 cores 4, 8, 12, 16, 20, 24, 28, 32. 3,328 Oracle Exalogic Four sizes fit all? PDU PDU 3,328 Cores PDU PDU PDU PDU PDU PDU PDU PDU PDU PDU PDU PDU PDU PDU 48 96 192 360 1/8 Rack 1/4 Rack 1/2 Rack 1 Rack Private Cloud on Pure Systems v1.2 8
Integrated Flex System Manager Single Pane of Glass System Management Server Network Storage Security Virtualization Mobile UI Flex System Manager Integrated appliance supports single pane-of-glass for managing all physical and virtual resources Auto-discovery and inventory of hardware Photo-realistic chassis map shows integrated view of all health information Dynamic provisioning of virtualized resources (via VMcontrol or vcenter) Policy-driven, automated virtual machine placement due to utilization, energy, or failure conditions Centrally perform non-disruptive system updates Private Cloud on Pure Systems v1.2 9
PureFlex Power Has Better Hypervisor Efficiencies Built In Which Deliver Lowest Cost Per Workload Hypervisor Efficiency Lowest Cost Per Workload Workload Density IBM PureFlex System Private Cloud on Pure Systems v1.2 10
PureFlex System with Power Delivers Better Workload Density For Virtualized Workloads Compare the Number of Virtual Machine Workloads Supported Online Banking Workload: Average Rate of 80 User Interactions per Second Average CPU% Utilization per Server 80 70 60 50 40 30 20 10 0 OracleVM VMWare V5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 Number of VMs 70.2% Pre-integrated 61.8% Software Competitor Hypervisor Hypervisor on Intel W-EP 12C PureFlex System 21 VMs Intel SB-EP 16C 48.8% 63 VMs PowerVM on PureFlex System POWER7 32C 144 VMs PowerVM 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Aggregate User Interactions per Second This is an IBM internal study of PureFlex System solution designed to replicate a typical IBM customer workload usage in the marketplace. The results were obtained under laboratory conditions, and not in an actual customer environment. IBM's internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or testing conditions may produce different results and may vary based on actual configuration, applications, specific queries and other variables in a production environment. Private Cloud on Pure Systems v1.2 11
What Is Workload Density? Workload density refers to how many virtual machines can be run on a virtualized server Greater workload density reduces cost per workload Hardware cost is shared by more workloads Software license costs are shared by more workloads Reduces footprint required Workload density is determined by Size of the workload versus capacity of the physical server Variability of the workload demand Efficiency of the hypervisor Private Cloud on Pure Systems v1.2 12
Cost Per Workload On Power Is Less Than VMWare on Intel 728 User Interactions per Second WAS 8 PowerVM / POWER7 4 Socket /32 Core POWER7 (3.55 GHz) 17 workloads $42,436 per workload Greater workload density yields 32% lower cost per workload WAS 8 VMware / Intel Sandy Bridge 2 Socket /16 Core (2.70 GHz) 5 workloads $62,623 per workload Source: IBM CPO internal studies Private Cloud on Pure Systems v1.2 13
IBM PureApplication System Adds Built In Expertise To Drive Labor Costs Down Even More IBM PureApplication System Deep integration and optimization Autonomic and Dynamic Workload Management + Integrated Workload Deployment + Web Application Middleware Platform + PureFlex Infrastructure Foundation Built-in workload elasticity using predefined scaling policies Built-in expertise via web workload patterns; Self-service, automated provisioning of workloads Pre-entitled licenses included Compute + Network + Storage + Management Private Cloud on Pure Systems v1.2 14
Compare IBM PureApplication System To The Competition IBM PureApplication System State-of-the-art management Fastest deployment Best practices workload patterns Dynamic workload management Lowest cost of labor for private clouds Oracle Exalogic / Exadata Vendor setup team No virtualization for Exadata v1 Virtualization is new on Exalogic v2 - July 2012 Full management support not yet delivered Private Cloud on Pure Systems v1.2 15
Case Study - Compare The Customer Labor To Manage 48 Web-Facing Workloads 6 blades, 96 cores TradeLite App Web DB App 48 Web-facing Workloads Do-It-Yourself (DIY) 1/8 Rack Compute (4 nodes) + 1/2 Rack Data (4 nodes) Oracle Exalogic/Exadata* PureApplication System - Small (6 nodes, 96 cores) PureApplication System** Which option requires the least labor? * Performance ratio used based on experiments - Nodes required PureAS: Exalogic :Exadata = 5.28:2:4 * Oracle minimum 1/8 rack with 4 compute nodes (only 2 needed); db instances on ½ rack database machine with 4 nodes **PureApplication System ¼ rack with 96 cores Intel Private Cloud on Pure Systems v1.2 16
Case Study - Five Key IT Processes For Infrastructure Administration Change Management Based on ITIL Industry best practice for life cycle management Security Management 20% 10% 22% Deployment Management Deployment Management Hardware set-up and software deployment Incident/Capacity Management Monitor and respond automatically 12% Asset Management 36% Incident/Capacity Management Typical percentage of time spent on each task category Asset Management Hardware and software asset tracking Security Management Access control Change Management Hardware and software changes % Allocation based on customer data from IBM study ITIL = Information Technology Infrastructure Library Private Cloud on Pure Systems v1.2 17
PureApplication Deployment Cost Reduction Reduce Virtual Machine And Software Deployment Costs Deployment System set-up time Pre-loaded best practice images and patterns for quick-starts Self-service console to request workload provisioning Drag and drop tooling to create new images or patterns Intelligent placement algorithm to map workloads to physical servers Private Cloud on Pure Systems v1.2 18
PureApplication System Complete System And Applications Up-And-Running In Hours Integration lowers start up time Hardware Integration / Testing Install Components Cable Hardware Configure Hardware System Management Configuration Software Integration / Testing Preload Images Install Pre-Entitled Patterns Final Functional test of entire system Private Cloud on Pure Systems v1.2 19
PureApplication Delivers Much Faster Initial Set Up Times Than Oracle IBM PureApplication System IBM PEP Team + Customer <4 hours Includes (at no additional cost): Hardware set-up and configuration Software set-up and configuration First workload deployment Oracle HW Engineer Oracle SW Engineer Customer Oracle Exalogic Hardware Set-Up and Configuration (2 days*) Software Set-Up and Configuration (2 days*) First Workload Deployment (1 day*) 5 days* * Actual times as experienced by customer Oracle team manually sets up hardware and software for each node Oracle Reference Implementation used to manually set-up and configured at customer site Oracle Enterprise Manager software is not installed by Oracle team Private Cloud on Pure Systems v1.2 20
PureApplication Integrated System Management Set Up Time IBM PureApplication System Oracle Software Management Add-On Packs Hardware Management Add-On Packs No set-up time Pre-installed, pre-integrated management stack Single pane-of-glass management * Source : Actual times reported by customer working with competitor setup team 8 hour* set-up time Management stack requires separate purchase, install and configuration Different user interfaces for software and hardware management Private Cloud on Pure Systems v1.2 21
IT Production Requirements And Deployments Can Get Expensive Very Fast 1. What the business wants 2. What s actually required Application DB 3. Then it suddenly grows bigger tomorrow Private Cloud on Pure Systems v1.2 22
Setting Up A Production Environment Manually Can Be A Time Consuming Laborious Process Typical Manual Process Steps Small Cluster of Two Application Servers and DB 1. Install Operating System with fixes and test 2. Install Application Servers and required fixes and test 3. Configure the HTTP server and security settings 4. Configure HTTP servers for high availability 5. Create an application cluster with session replication to support failover 6. Install a database and required fixes and test 7. Install the DB schema and populate the DB 8. Configure DB for high availability (cluster) 9. Connect the Application Server with the DB with JDBC drivers and test 10. Deploy the application to the cluster and test 11. Set up the access rights for the cluster, log files and test 12. Final test and performance tuning This process can easily take 15 hours or more * All of these manual steps can be included in a Pattern! * IBM Internal Study Private Cloud on Pure Systems v1.2 23
PureApplication Provides Breakthrough Economics With Patterns Of Expertise What is a Pattern? The pre-defined architecture of a system and application For each component of the application (i.e. database, web server, etc) Pre-installation on an operating system Pre-integration across components Pre-configured & tuned and Pre-configured Monitoring Pre-configured Security Lifecycle Management In a deployable form, resulting in repeatable deployment with full lifecycle management Example of a WAS pattern with Operating System IBM patterns are shipped as an OVF virtual image, ready to run on a hypervisor WAS V7 HV WAS V8 HV WebSphere Application Server OVF = Open Virtualization Format Operating System Private Cloud on Pure Systems v1.2 24
PureApplication System - Workload Deployment Methods Virtual Appliance Single virtual machine image with software IBM supplies Hypervisor Editions Automated deployment Virtual System Patterns Multiple virtual machines with software Explicit topologies IBM supplies predefined, best practice patterns Automated deployment of pattern Virtual Application Patterns Define workload policies Implicit topologies are automatically generated and deployed based on defined policies Container-like services at run time Configuration flexibility Skills required More labor Easiest to use Less skills required Administrator productivity Private Cloud on Pure Systems v1.2 25
PureApplication Manager Orchestrates Deployment Process For Each Virtual Machine In The Pattern Workload Up and Running PureApplication Manager Start Application Install Script Copy Application Script Package Boot VM Create VM Assign files to storage App Scripts Application Install Scripts OS Init sequence Invokes Activation Engine (configure OS; start and configure WAS and DB2) Mount WAS and DB2 Volumes Virtual Machine Hypervisor Install Kernel OS Boot Record Begin Boot Boot OS OS WAS Map Map Map DB2 Cloud Group Volume AE Scripts OS WAS DB2 V7000 Storage Private Cloud on Pure Systems v1.2 26
PureApplication Manager Orchestrates Deployment Process For Each Virtual Machine In The Pattern Workload Up and Running PureApplication Manager Start Application Install Script Copy Application Script Package Create VM Assign files to storage App Scripts Orchestration via remote commands Boot VM Application Install Scripts OS Init sequence Invokes Activation Engine (configure OS; start and configure WAS and DB2) Mount WAS and DB2 Volumes Hypervisor Install Kernel OS Boot Record No change to Begin Boot Virtual Machine hypervisor Boot OS OS No change to normal OS boot process WAS Activation Engine and configuration scripts for Map Map Map middleware DB2 Cloud Group Volume AE Scripts OS WAS DB2 V7000 Storage Private Cloud on Pure Systems v1.2 27
PureApplication System Virtual System Patterns Define Explicit System Topologies A Virtual System Pattern (VSP) is one or more virtual images with script packages ready to automatically deploy as a collection of virtual machines Self- Service Portal select & configure automated deployment VM VSP Image and Pattern Repository VM VM Operating System Hypervisor Virtual System Pattern Examples Single WebSphere Server Standalone node WebSphere cluster DMgr WebSphere Advanced Cluster DMgr WAS Node WAS Node WAS Node WAS Node WAS Node WAS Node WebSphere and DB2 WAS Node DB IHS IHS IHS IHS ODR ODR Hypervisor Edition virtual images preloaded in software catalog repository Pre-entitled Images: Red Hat Linux OS WebSphere Application Server DB2 Automation Framework HV (for migrating applications) Optional Images: WebSphere MQ WebSphere Message Broker Business Process Management Cast Iron SOA Policy Managed Gate Portal Hypervisor Edition Informix Ultimate Hypervisor Edition Business Partner/ISV: 55 ISV partners with 100+ applications Private Cloud on Pure Systems v1.2 28
DEMO: Using PureApplication Virtual System Patterns To Build And Deploy A Web Application A full production Web Application is quickly built by dragging and dropping components on to the pattern editor A Virtual System pattern is created and deployed in less that 10 minutes Private Cloud on Pure Systems v1.2 29
Virtual Application Patterns Can Help You Build And Deploy Even Faster Captures years of customer experience and codifies this in patterns of expertise (best practices) Bring your data and application code, select the type of application you want and everything else is handled for you in the background What do you NOT have to do.. Understand the interdependencies and connections between your database, application server, management, security, and the rest of the middleware Manually engage in the real-time management of your infrastructure Result: Result: Speed Speed deployment deployment of Web of Web applications applications by 100x! Private Cloud on Pure Systems v1.2 30
PureApplication Virtual Application Pattern Fastest Way To Economic Breakthrough And Life Cycle Management A Virtual Application Pattern defines workload policies. PureApplication automatically deploys a specific pattern of virtual machines designed to satisfy those policies Virtual Web Application Pattern Simply define your application Application (.ear) Scaling Policy DB2 (.ddl) Define scaling and routing policies Routing Policy PureApplication automatically provides all the components necessary to deploy an instance of the pattern Self- Service Portal configure Define application options and policies Infrastructure automatically created Run time Monitoring, Automatic resource adjustment automated deployment VM VM VM Proxy Service WebSphere (.ear) DB2 (.ddl) Operating System Hypervisor Proxy Service WebSphere (.ear) Caching Service (WXS) Caching Service (WXS) Private Cloud on Pure Systems v1.2 31
Virtual Application Pattern Includes Automated Management Functions Full Functions Proxy service, Web cluster with failover, database, data grid, external connections Load Balancing Monitoring Auto Scaling Web requests are automatically load balanced across multiple virtual application servers All components of virtual application environments are monitored by PureApplication System Managed environments scale up and down based upon business policies you specify Monitoring Lifecycle Management Resiliency Security Failed virtual machines are replaced with new VMs which are configured with the old VM s identity ACL s for application sharing and management access; LDAP integration for application security Private Cloud on Pure Systems v1.2 32
DEMO: Using PureApplication Virtual Application Patterns To Quickly Deploy With A Service Policy 1. Build a Virtual Application Pattern using PureApplication Expertise 2. Assign a service policy (scaling) to assure your image will maintain Service Level Agreements (SLA) Private Cloud on Pure Systems v1.2 33
Exalogic Lacks Patterns Virtual Appliance Single virtual machine image with software IBM supplies Hypervisor Editions Automated deployment Virtualization is NEW as of July 2012 Virtual System Patterns Multiple virtual machines with software Explicit topologies IBM supplies predefined, best practice patterns Automated deployment of pattern Virtual Application Patterns Define workload policies Implicit topologies are automatically generated and deployed based on defined policies Container-like services at run time Configuration flexibility Skills required More labor Easiest to use Less skills required Administrator productivity Private Cloud on Pure Systems v1.2 34
Compare Deployment Labor Times For A Simple Web Application 180 Software Labor (Minutes) 160 140 120 100 80 60 40 20 0 Do-It-Yourself 89% Less Time PureApplication System Virtual System Pattern Web App DB Workload deployed 98% Less Time PureApplication System Virtual Application Pattern 50X faster Private Cloud on Pure Systems v1.2 35
Labor Time To Deploy A Simple Web Application Is Also Significantly Less Than Oracle 40 Software Labor (Minutes) 35 30 25 20 15 10 5 52% Less Time Web App DB Workload deployed 89% Less Time 10X faster 0 Oracle Exalogic Exadata PureApplication System Virtual System Pattern PureApplication System Virtual Application Pattern Private Cloud on Pure Systems v1.2 36
PureApplication Incident And Capacity Cost Reduction Deployment Reduced by 98% Incident/ Capacity Reduce incident/capacity management costs Monitor from a single-pane of glass to quickly isolate issues with hardware and workloads running on multiple virtual servers Automatic resource adjustments for workloads to meet performance goals Private Cloud on Pure Systems v1.2 37
Incident And Capacity Management In Action - Elasticity With Policies Create Policies That Provide Guaranteed Highest Quality Of Service Define service level goals with policies Routing Policy, Log Policy, JVM Policy, Scaling Policy Scaling policies specify the response time goals and the relative importance of the service policy relative to other service policies Static, CPU based, Response time based, Web to DB Application requests are mapped to service policies based on rules Policies define the relative importance and response time goals of application services WebLogic cannot do this! Exalogic cannot do this! Private Cloud on Pure Systems v1.2 38
DEMO: Scaling Policies Can Dynamically Meet Workload Performance Goals PureApplication System automatically scales the WebSphere application platform up/down as the load changes according to policies defined 1 proxy proxy 2 proxy proxy 3 proxy proxy WAS WAS WAS WAS WAS Hypervisor cache cache Hypervisor cache cache Hypervisor cache cache Workload exceeds threshold PureApplication System instantiates new VM to offload workload Workload falls below threshold PureApplication System deletes unneeded VM Private Cloud on Pure Systems v1.2 39
PureApplication Cuts Incident And Capacity Management Labor Costs 1400 1200 Total Labor Hours (Hardware and Software) 1000 800 600 400 200 83% fewer labor hours 93% fewer labor hours Incident 0 DIY Oracle PureApplication System Virtual Application Pattern Private Cloud on Pure Systems v1.2 40
PureApplication Asset Management Cost Reduction Reduce Asset management costs Keeps track of license usage of products in creating virtual images Track usage by users and group and automatically alert when product requests exceed available entitlements Generate audit reports and enforce placement decisions based on available licenses Asset Reduced by 19% Deployment Reduced by 98% Incident/ Capacity Reduced by 93% Private Cloud on Pure Systems v1.2 41
PureApplication Security Management Cost Reduction Reduce security management costs Images, patterns and permissions are stored securely Centralized granular access control, allowing to restrict visibility, pattern creation privileges and deployment rights to individuals and groups Security Reduced by 50% Asset Reduced by 19% Deployment Reduced by 98% Incident/ Capacity Reduced by 93% Private Cloud on Pure Systems v1.2 42
PureApplication Streamlines the Change Management Process PureApplication provides visibility into relationships of various components of a workload View virtual machines in context of a workload Reduces time to determine the potential impact of a change Update images and patterns once in the repository for future use Security Reduced by 50% Reduced by 74% Change Deployment Reduced by 98% PureApplication can apply changes across all platform components in a staged manner PureApplication can automatically deploy a change to selected virtual machines Asset Reduced by 19% Incident/ Capacity Reduced by 93% Private Cloud on Pure Systems v1.2 43
PureApplication Delivers Labor Savings And Reduces Overall Administration Costs Change Management Standardization of images and firmware, visibility into relationships among components; coordinated automated changes Security Management Centralized fine-grain access control Asset Management Discovery and license management for products 3648 total hours per year reduced by 75% to 924 hours per year Reduced by 50% Reduced by 19% Reduced by 74% Reduced by 98% Reduced by 93% Deployment Management Automated intelligent deployment of complete application environment Incident/Capacity Management Automation to isolate and fix issues by scaling Private Cloud on Pure Systems v1.2 44
IBM PureApplication System Significantly Reduces Overall Labor Costs Yearly Labor Hours (hardware and software) 4000 3500 3000 2500 2000 1500 1000 500 3,648 hours 75% fewer labor hours 1,969 hours 53% fewer labor hours 924 hours Change Mgmt Security Mgmt Asset Mgmt Incident/Capacity Mgmt Deployment Mgmt 0 DIY Oracle PureApplication System Case study included 48 simple deployment pattern workloads This is an IBM internal study of IBM PureAS solution designed to replicate actual IBM customer usage in the marketplace. It is not a benchmark application, nor is it based on a benchmark standard. As such, customer applications, differences in stack deployed and other systems variations may produce different results and may vary based on actual configuration, applications, specific queries and other variables in a production environment based on published standard labor rates for IT staff. Private Cloud on Pure Systems v1.2 45
Summary - IBM PureApplication Delivers IT Services Faster And At Lower Costs Compared To Oracle IBM Oracle Initial Set up: Faster time-to-value Yes No Integrated management Silicon to Applications (Single Pane of Glass) Yes No Built in best practice pattern expertise Yes No Elasticity built into the patterns Yes No Reduces labor required for entire life cycle management Yes No Private Cloud on Pure Systems v1.2 46