Escolar Documentos
Profissional Documentos
Cultura Documentos
August 2013
IT Expenditures
It has now reached the point that administrative costs are crowding out innovation. According to a 2011 IDC study, just keeping the lights on (system planning, deploying, maintaining) takes up 77% of ITs budget while only 23% goes to supporting business innovation.
24% 23%
the Lights On
77%
Originally, many IT professionals expected that moving to a shared x86 infrastructure would be the solution to their administration problem. In theory this should simplify IT management. Yet IT professionals have found that if they merely take the traditional paradigm of building custom installations consisting of siloed pools of compute, network and storage and apply it to virtual systems, their administrative load does not decrease. They are still plagued by the need to design, assemble, configure, optimize, test, patch, support and upgrade each component individually. However, there is an opportunity to fix this problem.
1 IDC White Paper. Converging the Datacenter Infrastructure: Why, How, So What? May 2012
Converged Infrastructure Simplifies Infrastructure Management, Dramatically Reducing the Administrative Cost to Just Keep the Lights On
Area Customized Integration In the Field (Do-It-Yourself or Integrator-Built) Custom implementations consisting of separate pools for compute, network and storage, are designed for individual environments and for deployment by IT staff and vendor services. The system is delivered as components to the data center where it is physically and logically configured on site by vendor services and/or IT staff. Patches for each component must be thoroughly tested by IT staff against the custom installation to ensure compatibility. Expanding capacity requires careful testing of newer versions of components by the IT department. Vendors silo-based support organizations specialize in individual components and place the burden on IT of overall problem resolution and mediating finger pointing among vendors or among a vendors product divisions. Converged System: Vendor-Delivered Standardized System The system is engineered by the vendor as a single integrated pool for optimal performance, scalability, and availability and configured to meet customers unique needs. The complete system is manufactured and logically configured at the factory using standardized processes. Patches are preselected, pretested, and prepackaged by the vendor for interoperability and compatibility with installed configurations, and are ready for immediate non-disruptive deployment. Newer versions of components are preselected, pretested and certified for interoperability and design integrity and are ready for immediate non-disruptive deployment. There is a single point of ownership that is expert in all aspects of the system. All deployed system configurations are fully supported to accelerate problem resolution.
Design
Deploy
Maintain
Upgrade
Support
Evolve
Vendor engineers next generation system leveraging The process starts over from scratch at the design latest advances from each component while phase, relying on IT or service providers. Migration also providing a migration path that interlocks roadmaps requires customized services. from each component.
Business As Usual
But despite the similar marketing messages, not all converged infrastructure offerings are created equal. The key question for IT is: how far does the vendor truly go toward delivering on the promise of a vendor-delivered and supported system vs. providing systems that are simply assemblages of components where each deployment is a custom installation? In the remainder of this paper, we will compare two converged systems: IBM PureSystems and VCE Vblock Systems (Vblock). We will evaluate them on how well they deliver on the promise of transforming IT infrastructure into a vendor-delivered system that lowers the administrative burden, thereby freeing resources up for innovation.2
Innovation
About VCE:
VCE is a joint venture formed in November 2009 by EMC, Cisco and VMware, with investment from Intel. The company brought together IT infrastructure from the three industry leaders and sought to deliver it to customers as a single entity through the newly-created VCE. VCE introduced the first Vblock in 2010 and has since expanded its portfolio significantly. Product Portfolio VCE offers 4 general purpose Vblock series (100, 200, 300 and 700) as well as several models within each series. Each series and model is designed to meet different price, performance and scalability needs. The company also offers a specialized system for SAP HANA. Components All Vblocks are built on a standardized infrastructure that uses best-in-class components as follows: Compute and Network: Vblocks compute component is based on Ciscos UCS product line, and its network component is based on Cisco Nexus and MDS switches. Released in 2009, Cisco UCS and Nexus were designed from the ground up for virtualization and feature innovations such as unified fabric, embedded management, and policy computing that distinguish them from the competition. Cisco has long been the dominant networking vendor and in less than 4 years, it has grown to the #2 position in the US blade server market from scratch according to IDC.3 Storage: All Vblocks use storage from EMC, the leading provider of storage with over a 40% share. Each Vblock series uses the EMC product that matches its target price, performance and scalability parameters, with the top of the line 700 series using enterprise-class VMAX storage, the 200 and 300 using the mid-tier VNX product, and the 100 using the more affordable VNXe. Virtualization: VMware ESX is included in the solution package and supported by VCE. VMware is the leader in virtualization with 82% market share. Management: As part of its package VCE offers its Vision Intelligent Operations software (VCE Vision). VCE Visions features include virtualization optimization, converged operation and an open API that enables users to use their own management tools of choice.
2 This study was funded by VCE 3 http://blogs.cisco.com/datacenter/cisco-achieves-2-worldwide-in-the-x86-blade-server-market/
Converged Infrastructure
The success of these early converged infrastructure deployments coupled with increasing demand for converged infrastructure systems by customers attracted additional players. The market is now fast-moving and filled with vendors with very similar marketing messages.
Early Converged Infrastructure Systems Were Very Successful in Lowering the Costs of Keeping the Lights On and Freeing Up Resources for Innovation
Market Experience According to a recent Gartner study, VCE currently controls 57% of the Integrated Infrastructure Systems market.4 The company has over 500 customers worldwide and its customers have deployed over 1,100 systems. An IDC analysis of VCEs customers shows that they have achieved major reductions in administrative costs and have freed up resources for innovation. On average, customers report:5 5x faster deployment 3x lower costs 83x better availability 4x improved staff availability
4 http://www.vce.com/about/market-leader 5 IDC White Paper. Converging the Datacenter Infrastructure: Why, How, So What? May 2012
Virtualization: IBM offers four virtualization choices including VMware, Microsoft Hyper-V, Red Hat KVM, and its proprietary PowerVM. Customers need to sign separate contracts with respective virtualization software vendors. Management: Flex System Manager (FSM) is a dedicated appliance for system management with a focus on managing hardware components leveraging Tivoli and other software. In terms of integration with virtualization, it offers VMware plugins for storage, but not for its compute or network elements. Market Experience While IBM has claimed that it has 4,000 customer installations of PureSystems,6 both industry analyst research reports and Topline Strategys own research suggests that this figure is wildly inflated. The data suggest IBM has very few deployments of truly converged systems. Topline Strategys own research corroborated industry reports that placed IBMs market share of Converged Infrastructure in the single digits. It also revealed that IBMs claim of 4,000 systems is the result of marketing sleight of hand. While there may be 4,000 deployed systems carrying the PureSystems brand, our research shows that very few of them are actually converged systems. Most of these systems are: Stand-alone compute deployments: When the PureFlex platform, consisting of only the Flex System Chassis and blade compute nodes, is deployed on its own and connected to customers existing storage and networking, it is no different than a deployment of any traditional server platform. While IBM is counting these stand-alone deployments as PureSystems installations, these installations are not converged infrastructure. These deployments account for the majority of the PureSystems installed base. Field-integrated custom installations: Even when customers purchase a bundle of compute, network and storage to deploy a system, frequently they are still not getting a truly converged system. Often for scalability and other reasons, customers choose to replace the Pure Systems Storwize midrange storage with external storage from another vendor. This external storage is integrated into the system in the field and results in a custom installation that breaks the CI paradigm. Deployments of PureData: While the PureData products carry the Pure name, they are based on a fundamentally different technology than the rest of the PureSystems family. For example, the PureData data warehousing and analytics appliances are a rebranding of IBMs SmartAnalytics and Netezza products. These deployments are not comparable to general purpose CI systems and cannot be fairly counted as CI installations.
As a business, VCEs sole focus is on creating vendor-delivered systems from the best-of-breed components and software from Cisco, EMC and VMware. Our analysis shows that they have done so successfully. As described in detail below, VCE delivers the simplified experience across the lifecycle that is very close to the vendor-delivered system vision described on page 2. In comparison, IBM PureSystems appears to be driven as much by its need to leverage the products it already owns and support its large installed base of customers who have different legacy environments as it is by the desire to design an optimized vendor-delivered system. While IBM does offer a vendor-delivered system option, our analysis suggests that it was not designed to be widely adopted. By choosing to use its own Storwize 7000 midrange storage as its standard offering, IBM virtually guaranteed that all enterprise-class systems would end up as customized installations in the field rather than standardized, factory integrated systems. By selling the PureFlex System chassis and compute nodes on their own and leaving the storage and networking components to be integrated later, IBM is continuing to do business as usual, selling components that are integrated in the field into customized, non-standard installations. In fact, as stated in the last section, we estimate that just a fraction of the 4,000 PureSystems deployments IBM reports are vendor-delivered systems. Furthermore, even when PureSystems is deployed as a vendor-delivered system, our analysis found that IBM had not progressed nearly as far as VCE in simplifying the design, deployment, maintenance, support, and upgrade experience. Again, our analysis suggests that this is largely by design. Within the PureSystems family, IBM offers 2 processors, 3 networking devices, 4 hypervisors and a host of other options. With such a complex array of possible configurations, we do not see how IBM can provide the same level of optimization and support for its vendor-delivered system as VCE has done for its system.
Design:
Evaluation Criteria: The system is engineered by the vendor as a single integrated pool for optimal performance,
VCEs Vblock was conceived of and engineered as a vendor-delivered system for highly virtualized and cloud environments. The companys vision of creating an optimized vendor-delivered system is evidenced in 1) its selection of best of breed components and 2) providing a family of systems that match the varying price, performance and scalability needs of different customers.
In comparison IBM PureSystems design appears to compromise between the competing demands of: Building a scalable, optimized system that adheres to the vendor-delivered system paradigm Leveraging the products it owns and supporting its installed base.
VCE Offers 4 Series of Vblocks Designed to Meet a Wide Range of Price and Performance Needs
Vblock 200 Vblock 100 Vblock 300 Vblock 700
Lower End
Enterprise
A closer comparison of the constituent components of both systems further illustrates the differences. Server: VCE Vblocks use Cisco UCS Servers, which emerged in 2009 with an innovative design to reduce complexity while providing the highest density of VMs per core. This resulted in best price performance in the industry and enabled Cisco to surpass IBM as the number two provider of x86 blade servers in the US.7 While newly released in 2012, IBMs Flex System chassis lacks the networking and management unification of UCS, leading to extra switches and cabling. It is designed to support two types of processors - PowerPC and x86, which inherently trades off optimizing for either one. Storage: VCE uses EMCs VNXe, VNX, and VMAX products, all of which are tightly integrated to VMware. This provides users flexibility to choose the best storage for the needs of the pool, including: EMC VMAX, which provides unique capabilities such as federated storage to support the largest and most demanding applications. VNX, the market leading mid-tier storage system, which has advanced features such as Fully Automated Storage Tiering that provide a much more efficient use of expensive SSD resources. VNXe, a more cost effective option for pools supporting less demanding applications. IBM uses Storwize V7000, which is a midrange solution. If an installation requires enterprise class storage, IBM will substitute another storage product in the field, breaking the integrity of a standardized and premanufactured product and turning it into a custom integration in the field.
7 http://blogs.cisco.com/datacenter/cisco-achieves-2-worldwide-in-the-x86-blade-server-market/
Hypervisors: VCE is designed and optimized for one hypervisor, industry leader VMware. PureSystems offers a choice of four different hypervisors.8 While it is possible, it is highly unlikely that IBM can deliver the same degree of system optimization and support for four hypervisors that VCE can provide for one. Questions for Evaluating System Design As a potential buyer of a converged infrastructure system, here are questions regarding the design of the system we recommend you ask as part of your evaluation: 1. Has the vendor truly engineered systems optimized for virtualized and cloud infrastructures or are they compromising their designs to support their legacy components, processors and software? 2. Does the vendor offer solutions designed for each tier of performance and availability?
Deploy:
Evaluation Criteria: The complete system is manufactured and logically configured at the factory using
standardized processes.
Speed of deployment is a key outcome of VCEs focus on design-level standardization and integration. In most cases, VCE Vblocks are capable of shouldering operational loads within 45 days from time of customer order and within 48 hours of delivery. Both physical and logical builds are completed at the factory, so when a Vblock system reaches its destination: Components are assembled and fully tested System is logically configured using a repeatable process with extensive automation VMware and naming conventions are installed and IP addresses and VLANs are defined The net result is that VCE Vblocks arrive quickly and can be put into service in a short time span with low administrative effort for IT staff and minimal vendor services. In most cases, since IBM PureSystems deployments involve extensive customization in the field, very little of the build takes place at the factory. By necessity, when a customer reuses existing components or opts for storage other than the midrange Storwize product, the physical integration takes place in the field, requiring significant IT resources and/or professional services. Even the small number of complete vendor-delivered systems IBM deploys still require logical configuration in the field by IT staff or vendor professional services teams. Questions for Evaluating Deployment 1. Does the system require physical or logical configuration on site and how much of your internal labor or their professional services will be required to deploy it? 2. Does the vendor provide a guaranteed deployment window for the system? 3. How much has deployment risk been reduced by the vendor?
Maintain:
Evaluation Criteria: Patches are preselected, pretested, and prepackaged by the vendor for interoperability and
compatibility with installed configurations, and are ready for immediate risk-free deployment.
With a constant stream of updates for each component, patching can be an administrative nightmare. Not every patch is relevant to every deployed configuration, and each patch introduces its interoperability and compatibility risk in its own way. VCE simplifies the process and mitigates the risk in three important steps: 1. Since VCE only ships Vblocks in standard configurations, the company supports the entire configuration for every system it ships. As a result, VCE is able to evaluate each patch to determine which patches actually need to be installed on which systems. 2. VCE then tests each patch that does need to be installed against the entire Vblock, certifying compatibility at system level. 3. Finally, VCE bundles the patches into a single release. Through this process, VCE delivers risk-free patch bundles with predictable outcomes on a regular basis, eliminating the need for customers to select and test each patch, released at unpredictable times, against their specific configuration. One of the consequences of PureSystems design complexity (2 types of processors, 3 networking options, 4 hypervisors, etc.) is that IBM cannot provide nearly the level of patch support as VCE. To patch their system, customers have to go to the IBM Fix Center where they must choose which component to upgrade (chassis, storage, server, etc.), determine which version it is, and then download and install fixes on their own. Even though IBM does provide compatibility information, for the overwhelming majority of customers who have customized their installation, compatibility for their specific system has not been tested. As a result, burden and risk of ensuring compatibility falls on IT.9 Questions for Evaluating Maintenance 1. Does the vendor issue system releases of patches that are pre-selected, fully tested and validated against the entire system or do they issue component-by-component patches that you need to validate on your own custom installation?
Upgrade:
Evaluation Criteria: Newer versions of components are tested and certified for interoperability and design integrity
Since component vendors release new versions of their products every few months, it is virtually guaranteed that when a customer seeks to expand the capacity of their system, they will be installing newer component versions. VCE handles the integration of new hardware component versions in a way similar to how it handles patches each new component release is tested against each shipped system configuration to ensure system-level compatibility. In addition, the company also tests the design integrity of the complete system, ensuring that it has the headroom for and can support the power and cooling needs of the new component. As with patches, by the time new components reach the customer they are ready to install risk-free. Similar to patching, PureSystems design complexity and share of field-customized installations means that it is virtually impossible for IBM to fully certify each new component release against customers individual deployments, leaving the burden of testing each upgrade to the customer, as well as the risk carried by each release. Questions for Evaluating Upgrades 1. Does the vendor preselect, fully test and validate upgraded components against your deployed configurations beforehand at its own facilities, or do these components need to be tried out in your systems by your own IT resources, or your vendor at your time and cost?
9 IBM Fix Center: http://www-933.ibm.com/support/fixcentral/options
Support:
Evaluation Criteria: There is a single point of ownership that is expert in all aspects of the system. All deployed
system configurations are fully supported to accelerate problem resolution. VCE Support features a single point of contact for service request management and responsibility, including escalation support tied directly into Cisco, EMC and VMware engineering support. VCEs cross-trained support staff handles compute, network, storage and virtualization, and in 2012 they were able to resolve 70% of calls without escalation to more specialized support. VCEs emphasis on standardization also means problem-solving is shared across its customer base solutions are engineered in labs dedicated to problem re-creation and then distributed through the support network as needed. IBMs support model is more akin to triage than integration. While IBM may offer a single point of contact, it is the silo-based support divisions that actually provide support. These silo-based support organizations put the burden of isolating the problem and proving which component is causing the issue on IT staff. IBMs emphasis on customization also means more problems are unique in nature, limiting shared problem-solving knowledge base and resulting in slower resolutions. Questions for Evaluating Support 1. Is there a true single point of ownership for support or just a single point of contact that serves as a front end to traditional siloed support?
Evolve:
Evaluation Criteria: Vendor engineers next generation system leveraging latest advances from each component
while providing a migration path that interlocks roadmaps from each component. VCE has a proven record of releasing new versions of its Vblocks that include the latest advances from each component vendor. The first Vblock 700 was released in 2010 and the company is currently shipping its third generation system. The company maintains close roadmap coordination with Cisco, EMC and VMware to provide customers with visibility into new capabilities as they are being developed and a clear migration path for implementing new technologies. PureSystems is too new to have established a track record for us to analyze. Questions for Evaluating Upgrades 1. Does the vendor have a proven record of introducing new generations of its solutions? 2. If it does, does it offer a migration path for its installed customers?
Conclusion:
The key challenge for IT remains accelerating innovation by reducing the time, money and resources spent on deploying and administering infrastructure systems. Converged infrastructure solutions offer a way to do just that, but only if the system is designed with the needs of IT in mind. By creating a system that, on the surface, looks like a converged infrastructure, but underneath is mostly just a platform for delivering customized components, IBM does little to address the administrative load that is prevalent throughout the system lifecycle. By contrast, VCE Vblock truly changed the status quo. It built an integrated system that simplifies purchasing and setup, shortens the time it takes to bring new capabilities on line, and reduces the cost and complexity throughout the lifecycle of the product. In short, VCE is in the business of supporting a whole system that is much more than just the sum of its parts.
Final Question:
Will the system enable you to truly transform your IT operations, shorten the time-to-market for new capabilities and free up resources for innovation?