Global Headquarters: 5 Speen Street Framingham, MA USA P F

Transcription

1 SOLUTION BRIEF The Software-Defined Datacenter: What It Means to the CIO Sponsored by: VMware Inc. Global Headquarters: 5 Speen Street Framingham, MA USA P F Al Gillen John Daly Christine Dover July 2012 Richard L. Villars Randy Perry EXECUTIVE SUMMARY IT executives at organizations large and small have successfully virtualized much of their server infrastructure over the past five years. Today, over 50% of all applications run on virtual machines (VMs) on a virtualized server. These efforts have reduced datacenter capital costs, heightened asset utilization, and enhanced IT staff productivity across organizations. However, IDC finds that while virtualization delivers impressive initial productivity boosts, continuing results often do not meet management's expectations for further improvements in IT asset use and operational efficiency. This stall often occurs as the rapid expansion of virtual server deployments threatens to overload storage and data network facilities, resulting in overprovisioning of storage capacity and sharply increased administration workloads. As server virtualization reaches levels of hundreds, thousands, and even tens of thousands of server (and/or desktop) application instances, it becomes very clear that datacenter managers need, in addition to server virtualization, the virtualization, pooling, and management of all the other resources that interoperate with their VMs. They require virtualized network interconnects and storage and the tools to manage and automate them as an integrated datacenter system. They need a system that can evolve into a cloud infrastructure as they want it to, on their terms and according to their requirements, not on someone else's terms. In short, they need a centrally managed, software-defined datacenter. The industry is now answering these needs. Virtualized storage, "collapsed"/virtual network infrastructure for both storage interconnect and edge I/O software-defined networking and the orchestration software to manage and automate all these ingredients as a system are delivering results today. Our research with over 30 companies at different levels of convergence and virtualization experience from level 1 (low) to level 5 (high) indicates that increased IT convergence and improved asset sharing deliver substantial business benefits. Continuing results show that progress toward more centralized IT resource pooling and management, in fact, reduced IT costs per unit of workload, accelerated resource deployment, and reduced downtime. Organizations that had progressed to level 3 deployed new infrastructure in 40% less time than those at level 1. Level 3

2 organizations processed datacenter workloads (sustained throughput to the core) at less than 10% of the costs (total datacenter costs) of those at level 1. This document discusses the advances that enable these improvements. It describes the results that increased datacenter virtualization delivers, and it lays out both the promises and the potential pitfalls of the journey toward the software-defined datacenter. SITUATION OVERVIEW Continued economic volatility around the world and dramatic changes in ICT industry dynamics cloud; social marketing; ubiquitous, connected mobile devices; platform as a service; big data; and more are profoundly affecting organizations' IT investment and operations priorities. To keep pace, many of these organizations are making fundamental changes in business practices and transforming the datacenter to be better aligned with the heightened pace of both the business and the market. Over the past two years, IDC surveyed, interviewed, and consulted with a wide range of organizations around the globe to detect changing attitudes toward and priorities for IT investments. We spoke with C-level executives (e.g., CEOs, CFOs, COOs, CIOs), business unit leaders (e.g., vice presidents of sales and marketing, division heads, general managers), and IT leaders (e.g., IT vice presidents, CTOs, senior architects). Put simply, senior executives want to "operationally" transform IT so that their organizations can react quickly to major positive or negative changes in the business environment while also more reliably assessing revenue gains and cost savings from proposed investments in new technologies and services (e.g., mobile applications, use of cloud services, big data analytics). Server Virtualization: Transformation Stage 1 Despite these dramatic changes and imperatives, IT continues evolving at the pace and in the pattern of previous years. Capacity and capability continue to grow in synch with Moore's law. Server virtualization has rapidly taken hold. More than half of all applications deployed are running as virtual machines on a virtualized server. IDC believes that based upon typical server replacement cycles and new application deployments, the use of virtualization for servers will only accelerate in 2012 and 2013 (see Figure 1). 2 # IDC

3 FIGURE 1 Rapid Growth in Virtualized Applications; Increasing Pressure on IT Management/Administration Customer Spending ($B) Installed Servers (Millions) $250 $200 $150 Power & Cooling Expense Management and Administration Server Spending Logical Server Installed Base (Millions) Physical Server Installed Base (Millions) Virtualization Management Gap $ $ $0 '96 '97 '98 '99 '00 '01 '02 '03 '04 '05 '06 '07 '08 '09 '10 '11 '12 '13 Year - Source: IDC, 2012 But the relative ease with which application builders can use virtual machines to roll out new software or widen user bases has exposed and sharpened the significant challenge of acquiring, provisioning, and, especially, deploying the right level of platform (server, network I/O, and storage) capacity to enable this expansion. In the words of one IT executive we interviewed, " We were building new lines of business and onboarding customers at an incredible rate we were constantly gated as an organization by the speed at which we could then assemble the IT infrastructure behind that." To keep up with this demand without overwhelming IT staff, organizations working to virtualize their server infrastructure felt the need to change the way they acquired, configured, deployed, and maintained the rest of their IT services. In particular, they needed to deal with: Virtual server sprawl's effects on server/storage/network administrative burdens that threatened application performance 2012 IDC #

4 Overloading/overprovisioning of storage and data network facilities, which forces time-consuming, costly, and often unnecessary upgrades Unpredictable application performance and recovery behaviors, which stall plans to migrate more business-critical applications to virtual environments In our conversations with IT teams, we consistently heard that these problems could escalate as the scope of virtual server use expands beyond a few hundred virtual machines. How to Adapt: Transformation Stage 2 IDC found that a growing number of organizations are dealing with these issues as they address the "operational" mandates from senior management related to agility, increased business value, and reduced costs. They achieved improvement by investing in standardizing and converging their IT infrastructures. These converged infrastructures provided virtual resource pools of standard elements, including blade servers for compute, modular storage systems for data, unified 10GbE fabric for internode and storage connections, and even standardized software stacks with operating systems and often middleware layers, allowing the IT organization to provide a predictable amount of "IT capacity." As organizations extended this datacenter convergence, they experienced good results. They found themselves able to: Shift IT resources both funding and staff from operational tasks such as application deployment, application/data migration, performance monitoring, and business continuity to innovation and competitive differentiation Increase utilization to get more value out of both existing and new IT infrastructure investments (e.g., server, storage, and networking equipment as well as facilities and power and cooling) Accelerate the delivery of reliable IT services to support corporate business goals whether on premises or in private/hybrid/public cloud environments The Goal: Software-Defined Datacenters These organizations found that they no longer simply deployed each element (server, system software, storage, and network) individually and at the hardware level but that they deployed the optimal (e.g., densest, greenest, simplest) system from a centralized console using software tools: They had begun evolving into software-defined datacenters. Organizations that successfully embrace this approach can then manage the unified IT asset (e.g., rack, aisle, network port, storage capacity, or entire datacenter of converged IT infrastructure) by leveraging their hypervisor, network fabric manager, storage virtualization, and IT automation at the software console level. As the resources move to centralized pools, orchestration systems enable increased automation through dynamic partitioning and automatic load balancing for a pool of different business applications. As a result, IT teams can more quickly allocate large chunks of IT resources for critical services-oriented applications (e.g., virtual desktops, data analytics warehouses, or large content repositories) on an as-needed basis. In addition, this approach makes disaster recovery within a datacenter or 4 # IDC

5 across geographically dispersed datacenters not only practical but also predictable and cost effective. INDUSTRY DIRECTION: OPPORTUNITY AND FUTURE OUTLOOK We are not there yet. A majority of traditional datacenters today reflect a legacy of rapid IT evolution. That is, these datacenters continue to deal with multiple "silos" support resources dedicated to a specific domain (e.g., a technology island or a separate application or organizational domain [test and development versus financial systems versus departmental, etc.]). This world is changing, however. Positive advances in server, network, and storage virtualization enable today's IT managers to "converge" the datacenter and manage resources centrally via software. Advanced datacenters are moving a larger and larger portion of their datacenter resources (e.g., storage, memory, server nodes, network I/O, virtual OS images) into pooled collections that enable administrators to deploy resources centrally. The following technologies enable this convergence: Continued standardization. The consolidation of multiple hardware architectures to x86 servers and the deployment of increasingly consistent (standardized) operating system images and other software layers signal a movement toward standard platforms. Leading organizations today maintain a small number of "gold images" that are the basis for all virtualized server operating system deployments. Virtualized compute. Server hypervisors have allowed the pooling of server instance images, enabling the operation of multiple applications per physical server and allowing IT to respond to spiking application demand with more efficient utilization of physical server infrastructure. Virtualized storage. Networked storage that uses thin provisioning on virtual volumes allocates storage requested by the node as virtual storage. It preserves actual physical disk capacity and adds more only as needed. The system manages physical disks as a single disk pool and allocates physical disk capacity according to the amount of data written to the virtual volumes, avoiding overprovisioning. Virtualized network. Converged network switches such as UCS provide a centralized pool of virtual I/O ports that integrate core I/O for storage (SAN) and server interconnectivity (LAN, VLAN), eliminating the need for separate physical networks for each. Automation. Pooling of these resources enables dynamic partitioning and automatic load balancing for different business applications with different servicelevel demands. IT teams can allocate IT resources for critical services-oriented applications (e.g., virtual desktops, data analytics warehouses, or large content repositories) and handle disaster recovery within a datacenter or across geographically dispersed datacenters. Organizations that have leveraged all of these technologies report very positive results. Our continuing research with over 30 companies whose levels of 2012 IDC #

6 convergence and IT asset unification ranged widely indicated substantial business benefits associated with higher convergence and asset sharing. IDC interviewed these companies to assess their level of: Standardization reduction in the number of technology platforms (separate consoles) to learn and master Convergence percentage of datacenter resources (e.g., storage, memory, server nodes, network I/O, virtual OS images) that administrators can deploy from a pooled collection versus percentage of resources dedicated to a specific domain (e.g., a technology island or a separate application or organizational silo) We measured each organization's cost per datacenter workload sustained throughput to the core, speed of deployment, and availability. That cost served as one of the key inputs for our ranking, which ranged from level 1 (low) to level 5 (high). The results of the ranking are depicted in Figure 2, which shows the difference between datacenters at level 1 and those at level 3. Organizations that achieved higher levels of standardization and convergence realized reduced IT costs per unit of workload, faster deployment, and reduced downtime. 6 # IDC

7 FIGURE 2 Effect of Standardization and Convergence on Datacenter Costs per Workload, Availability, and Deployment Speed $24 Availability (%) Mean Time to Deploy New Services (Days) Annual IT Costs (000 omitted) per Gbps Workload Throughput Selected Metrics Standardization Level Percent of Server Nodes Running Hypervisor Percent of Nodes with Virtualized Network I/O Level 1 Level 3 No Centralized Asset Config DB, 70% with Centralized Asset Config DB, 50% with SLAs in Place; >14 Consoles 100% with SLAs in Place; <8 Consoles 20% 50% <10% >20%; Converged FCoE Installed $2 Source: IDC, 2012 Why Do Standardization and Convergence Deliver? Our research indicates that more standardization (i.e., fewer consoles and technologies) and more pooling of resources (i.e., virtualization and central management) allow datacenters to achieve higher levels of automated provisioning and higher utilization of networking, compute, and storage resources. This means that organizations achieve more datacenter throughput per unit of infrastructure capital cost, power, and facilities footprint. It also means that organizations experience lower datacenter administrative costs. The centralization drives the consolidation of technology support teams and the establishment of network operations centers to manage consolidated systems network, compute, memory, and storage support for critical applications. Availability increases as automation programs and datacenter administrators remediate system-critical events with "elastic" pools of available resources such as additional memory, I/O, or scale-out compute power IDC #

8 FINDING A SOFTWARE-DEFINED DATACENTER PARTNER A big challenge for organizations that want to migrate to embrace the softwaredefined datacenter infrastructure is finding the right partner(s) to help, especially in the following areas: Defining the right architecture and deployment characteristics for systems that meet performance and reliability requirements Integrating converged IT systems with existing IT environments and management systems Retraining IT staff to support a more unified approach to provisioning and managing pooled IT assets and services Organizations not only will need to look for an IT partner or solution provider with a broad product and services portfolio (server, storage, software, network, management) but also will need to be certain that this partner can help them overcome internal inertia in terms of segregated product evaluation, procurement, administration, process, and even culture. Most important, the partner will need to help them identify and sell the business value of adopting a software-defined datacenter approach when it comes to IT infrastructure investment decisions. CHALLENGES/OPPORTUNITIES Most organizations are at the beginning of the journey toward software-defined datacenters or private clouds. Even the most aggressive adoptees will likely need several years to complete this journey, and they will face several hurdles. Technical hurdles include: Security enhancements must ensure the protected separation of application data and processes as different applications increasingly operate contiguously within shared resources. Logical application security controls must replace the traditional "air gap" processing isolation that dedicated application resources formerly provided. Virtualization protocols and technologies, such as those for software-defined networking, must work for all environments. That is, they must work across multiple converged systems and across multiple datacenters. The industry must avoid the proliferation of proprietary protocols and technologies that lock in individual solutions and restrain centralized pooling and management. This can be particularly challenging for datacenters that have a mix of x86 and RISC, EPIC, or mainframe systems which is the majority of large enterprises. Centralized management and unified orchestration solutions must handle server provisioning, performance monitoring, and resource management across multiple converged systems technologies and across multiple datacenters. 8 # IDC

9 The switch away from server storage and network resources dedicated to an application to virtualized resources for all will require rethinking and redesign of the formerly dedicated backup and disaster recovery procedures for these applications. A software-defined datacenter approach presents technology challenges, but more important, it poses a significant number of challenges for IT organizations in terms of product evaluation, budgeting, roles, and IT operations management. Leading-edge adopters of software-enabled datacenters have dealt with these challenges and evolved. Other organizations with less experience in this transition will need help navigating the shift to tighter internal infrastructure standards, realigned datacenter organization, staff retraining, and financing and acquisition approaches. LEVERAGING THE POTENTIAL: THE NEXT ESSENTIAL STEPS The transition to a converged infrastructure the software-defined datacenter will play a vital role in helping IT teams meet the fast-evolving business needs of their organizations. It will help executives reduce both the capital costs and the operational costs of running datacenters and improve the resilience of the applications/information residing in them. Leading vendors are now advancing solutions that abstract and virtualize the hardware elements, provide an open operating environment, and support full orchestration of pooled resources across the entire datacenter. IT executives must adjust existing product and service selection and management practices to fully take advantage of the advancing converged IT infrastructure and the exceptional cost, speed, and resilience business benefits it delivers. Working closely with a partner or solution provider that understands this journey, its customers' needs, and the solutions of leading software-defined datacenter vendors will prove critical for the transition to the software-defined datacenter. In this context, IDC recommends that IT executives: Embrace standardization of hardware and software components as much as possible because it can simplify management and interoperability challenges (however, be sure that the approach also provides an interoperability and transition path for mission-critical applications on installed systems and SANs) Implement a mature, centralized, and automated approach to management operations with added investment in performance monitoring and analytics Revamp the IT organization structure to move away from device-specific (e.g., server, storage, and network) administration and move toward an IT resource oriented (e.g., database, collaboration, and archiving) structure Embarking on these initiatives will help organizations realize the significant efficiency and effectiveness improvements that our research has uncovered, even as the path toward the software-defined datacenter evolves, changes, and improves IDC #

10 Copyright Notice External Publication of IDC Information and Data Any IDC information that is to be used in advertising, press releases, or promotional materials requires prior written approval from the appropriate IDC Vice President or Country Manager. A draft of the proposed document should accompany any such request. IDC reserves the right to deny approval of external usage for any reason. Copyright 2012 IDC. Reproduction without written permission is completely forbidden. 10 # IDC