Fabric-Based Infrastructure Enablers and Inhibitors Through the Lens of User Experiences
Our discussions with clients and users identify five major impacts of fabric-based infrastructure to assist in evaluations, versus traditional data center approaches.
- External service providers justify fabric-based infrastructure (FBI) based on operating cost savings and density (for greater revenue per square foot), while enterprises base their FBI acquisitions primarily on capital cost savings.
- Gartner clients found that FBI's use of templates and profiles improves resiliency because, in the event of infrastructure failure, they can recreate servers in minutes.
- Virtually all clients with FBI in production found a reduction in time to provision from two to three months to a few hours to three days.
- Nearly all clients acknowledged that their purchase created a lock-in-type dependency on the vendor (and related partners), but were willing to accept the consequences for the benefits gained.
- All IT organizations acknowledged that they had to redefine data center roles for server, network and storage administrators from the conventional serial handoffs and delays to synchronous and collaborative management.
- Enterprises should assess FBI capital cost reductions in your ROI analysis but realize your mileage will vary.
- External service providers should assess how increased virtual machine (VM) density will improve your margins by increasing revenue per square foot of data center space.
- To maximize FBI value, implement infrastructure profiles to quickly recover after failure.
- Design Internet Protocol (IP) and storage-addressing schemes for automated provisioning and reclamation through FBI profiles and templates.
- Use the annual number of server builds and labor cost as an input into your FBI ROI.
- To maximize FBI value, develop workload/service technology and architecture patterns for repeatable engineering.
- Develop software configuration standards for the infrastructure patterns identified.
- Before FBI preintegrated systems (e.g., blocks) are agreed on by IT management, perform a time-to-value impact analysis.
- Form a competency center, define your automation objectives and define processes across functional groups.
- Add development skills to infrastructure and operations (I&O) for straight-through processing; this may increase initial costs, but will reduce ongoing operational costs.
Gartner interviewed more than a dozen IT organizations and service providers that selected a FBI-based on Cisco, Dell, HP and VCE offerings. Based on these interviews, we offer insight into key value propositions gained by successful implementation. We complement those with key challenges of FBI and methods that can be used to mitigate these risks. We offer recommendations to aid IT organizations in deciding for or against FBI, and in mitigating risks should they decide to acquire FBI.
Source: Gartner (April 2012)
Impact: External service providers justify fabric-based infrastructure (FBI) based on operating cost savings and density (for greater revenue per square foot), while enterprises base their FBI acquisitions primarily on capital cost savings
Cost reduction was cited as a major benefit of the FBI approach for both service providers and enterprises, although their justification differed. For enterprises, the attraction was capital cost savings, especially through the implementation of network virtualization, which simplified and reduced the cost of cabling and physical network interface cards. Those switching from racks to blades had the highest degree of capital cost reduction, because the enterprises that had implemented blades had already done some network and server consolidation. Even so, this was cited as the No. 1 ROI and value proposition (for project justification) for end user organizations. Floor space reduction and energy cost savings, as well as improving server provisioning times, were also cited as benefits.
Service providers favored the labor cost reductions that come with FBI as a result of programmability of the infrastructure (for provisioning customers) and dynamic integration of additional physical capacity, with little to no labor required after installation in the data center. One service provider we spoke with indicated that its new data center based on FBI incurred two-thirds less labor costs than its traditional data centers, which allowed it to improve its service delivery margins through reduced service delivery costs. Moreover, service providers justified their FBI investments based on greater density, thus achieving greater revenue per square foot of floor space. In addition, many FBI vendors offer service providers marketing, sales and business development funds to reduce risk and acquire customers, which also factored into the FBI ROI.
- Enterprises: Assess FBI capital cost reductions in your ROI analysis, but realize that your mileage will vary based on where you are in implementing blades and infrastructure consolidation and virtualization, specifically on the network side. Those with older infrastructure will see more benefits than those with newer infrastructure. Don't forget to assess operating cost savings as well as the benefits of increased provisioning speed to your ROI analysis.
- Enterprises and service providers: Assess labor cost reduction opportunities as a result of automating physical infrastructure capacity increases, quickly and efficiently resizing resource pools, and automating infrastructure provisioning for customers. Be prepared to reassign administrative personnel to other activities from the labor savings achieved through FBI.
- Service providers: Assess how increased VM density will improve your margins by increasing revenue per square foot of data center space. Enterprises should also assess whether improved VM density will reduce their data center floor space requirements and enable them to further consolidate facilities, thus driving cost savings.
- While cost reductions are always desirable and a key way to justify investments, keep in mind that FBI is fairly new and that there are negatives that should be considered as well (such as the degree of lock-in to the platform through specific and required processes that work differently for each vendor's platform).
Impact: Gartner clients found that FBI's use of templates and profiles improves resiliency because, in the event of infrastructure failure, they can recreate servers in minutes
FBI offerings have the ability to abstract runtime variables, such as CPU and memory required, connectivity to storage, host names, media access control (MAC) addresses, and IP addresses, from the software that is required to be run. As a result, if a server fails, it can be recreated in minutes on another server by applying the failed server's profile. This is most commonly implemented to automatically build or image a hypervisor or OS environment should a physical server fail, although it can be applied to virtual servers also. Clients indicated that this feature turns a two-hour labor project (to rebuild a server) to an automated one in minutes. It is also useful in terms of migration to new hardware, such as another blade or rack (but typically only to hardware from the same vendor).
- Use infrastructure profiles as part of the FBI management and automation tools to quickly recover after failure.
- Design IP and storage-addressing schemes for automated provisioning and reclamation through FBI profiles and templates.
- Use the annual number of server builds times the average labor cost as an input into your FBI ROI calculation.
Impact: Virtually all clients with FBI in production found a reduction in time to provision from two to three months to a few hours to three days
Cloud computing enables speed and agility primarily due to standardization — of the services offered, and of the components underlying those services — as well as automation of service fulfillment and delivery. To get the agility value proposition of FBI requires preplanning and implementation of standards. Clients must assess their types of services and workloads for patterns to develop a catalog of patterns that can be requested (for example, by application developers and testers for infrastructure as a service [IaaS] offerings, or by end customers for software as a service [SaaS] offerings), such as the infrastructure required for small, medium or large workloads and/or online transaction processing (OLTP) or batch services. These patterns become component services in a catalog that may be requested for automated delivery. In addition, the software to be installed must also be predefined by workload or service, so that a request in the catalog delivers not just the hardware infrastructure, but also the required software (such as a Linux, Apache, MySQL, PHP [LAMP] stack, a Microsoft stack, a particular application or a selectable list of heterogeneous software components). Failure to effectively develop hardware and software standards will result in high administrative, operational and maintenance costs, effectively negating many of the benefits of FBI. FBI vendors have varying degrees of capabilities in software deployment and orchestration, and often require more sophisticated third-party IT operations management software layered on top (which integrate with their server templates/profiles).
- Develop workload/service technology and architecture patterns for repeatable engineering of infrastructure. For example, such patterns would include the amount of infrastructure required for compute, memory and storage based on the size or type of workload or service. Determine whether overrides would be required by your user base when designing the service request entry form.
- Develop software configuration standards for the infrastructure patterns identified. For example, every time a request is made from a developer for a small OLTP service with the LAMP stack, the identical configuration of the software stack is deployed, thus enabling reuse of the configuration automation, and reducing ongoing support and maintenance costs.
- Enterprises that are unable or unwilling to develop and implement standard service/workload patterns should not put operational cost savings into their FBI ROI analysis.
- Before FBI preintegrated systems are agreed on by IT management, perform a time-to-value impact analysis.
Impact: Nearly all clients acknowledged that their purchase created a lock-in-type dependency on the vendor (and related partners), but were willing to accept the consequences for the benefits gained
All FBI solutions place limits on the supported hardware infrastructure (and possibly software infrastructure, such as the hypervisor software). They do this for two reasons:
- To reduce the complexity, and make it easier to write automation that supports the less diverse hardware infrastructure, which has the potential, therefore, to improve the time to value of the solution.
- To make their offerings more sticky, and encourage their customers to buy more of their infrastructures. Moreover, each solution depends on specific processes to be implemented, such as server profiles or templates, which are done in a specific way that also locks you into a process or way of doing things.
Some offerings have strict requirements on hardware and version support, while others are looser and support multivendor hardware infrastructure (but may require additional services to write and test automation). It is critical that enterprises weigh the degree of lock-in and the cost of such lock-in before they buy, as well as how they will minimize it, and prepare for contingencies if they find they do not get the value out of the solution that they expected. Many of the service providers and enterprises with which we spoke were willing to accept the lock-in for the perceived or real benefits gained. Others specifically purchased from vendors that supported multivendor hardware to manage their lock-in risk. Yet others added a higher-level management software and process layer to abstract the differences between the offerings and to enable an abstraction so that they could design and implement a single set of processes across multivendor FBI offerings.
- Know what you are getting into in terms of infrastructure (server, storage, network, hypervisor, OS) support when you buy an FBI solution. Assess whether it meets your current and future needs, the degree of heterogeneity supported, and its impact on support and timing of new infrastructure, upgradability (given varying asset life spans) and your management processes.
- When assessing FBI solutions, consider the openness and documentation around APIs. FBI solutions with open and published APIs will be less risky than those without them because, in the event you want to manage their functionality from a higher-level perspective (to reduce lock-in and the changes required to processes should you change infrastructure), it will be easier to do so. This is also useful for integrating automation solutions in a heterogeneous environment.
- Define and implement higher-level processes for provisioning and configuration management so that multiple heterogeneous FBI solutions can be supported, thus reducing potential lock-in because infrastructure can be changed underneath, without impacting the process.
Impact: All IT organizations acknowledged that they had to redefine data center roles for server, network and storage administrators from the conventional serial handoffs and delays to synchronous and collaborative management
FBI can both reduce cost and improve agility; however, to achieve both requires culture changes to be successful. Most IT organizations that have implemented virtualization have already started their move from a pure functional organization toward an end-to-end process orientation. When they do so, however, functional processes change, because policies are established across the new process. By doing so, IT organizations can begin to think about automating the entire process — or straight-through processing without human intervention (see Figure 2). This was the same evolution that manufacturing organizations experienced 30-plus years ago, when they implemented ERP systems and found that in order to get significant value, they needed to do things differently, not just automate what they had been doing in the functional groups.
Source: Gartner (April 2012)
- Form a cross-organizational competency center, define your automation objectives and begin to define processes across functional groups. Once the process is defined, policies can be established and automation can be developed that significantly reduces or eliminates much of the manual labor, thus reducing the cost of service delivery. This is what cloud providers do regularly — enable interaction with their systems and automate the delivery without human intervention. That is why they can deliver with speed, and their cost model appears to be attractive. Design IT processes for straight-through processing.
- Designing for straight-through processing will require development skills in the I&O organization to write automation. Add them if you have not already done so. This will have the effect of increasing your costs initially during design and implementation, but reducing your operational costs after implementation.
- impactappraisal... (44KB)