The Multi-tiered Hybrid Data Center

An innovative design model for cost reduction and facilities optimization

Table of contents
Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 The data center dilemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Traditional data center designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Single-tiered design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Multi-tiered (pod) design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 HP multi-tiered hybrid design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Innovative topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Business focused design methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Prospective business outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 For further information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7

Executive summary
To accommodate increasingly dense technology environments, increasingly critical business applications, and increasingly stringent service level demands, data centers are typically engineered to deliver the highest-affordable availability levels facility-wide . Within this monolithic design approach, the same levels of mechanical, electrical, and IT infrastructure are installed to support systems and applications regardless of their criticality or business risk if unplanned downtime occurs . Typically, highredundancy designs are deployed in order to provide for all eventualities . The result, in many instances, is to unnecessarily drive up both upfront construction or retro-fitting costs and ongoing operating expenses . The HP multi-tiered hybrid design approach, by contrast, right-sizes the data center redundancy infrastructure by engineering the facility to incorporate multiple operational environments, each aligned with business priorities and the criticality of specific technology systems and applications . Technology environments that, according to rigorous business-impact analysis, require similar levels of redundancyhigh, mid-level, or low are grouped within segmented raised floor spaces (pods) supported by right-sized facilities and technology infrastructure . HP multi-tiered hybrid design can reduce capital costs typically by 15 to 25 percent* . Because this architecture conserves energy, requires fewer support resources, and operates more efficiently, it can substantially cut total cost of ownership (TCO) as well .

The majority of data centers are mixed technology environments populated with systems and applications of varying criticality and business priorities . A banks ATM technology must be available 24x7, and therefore requires Tier 4 types of availability . But applications sharing the same facilityfor example, check processing or funds transfermay not be severely affected by some downtime instances and could therefore be accommodated by lower-tier redundancy operational levels . However, as application availability and proactive downtime prevention become perceived as increasingly central to business success, the trend has been to build data centers to the highest tier the enterprises budget allows . Entire facilities are often designed to accommodate the availability requirements of just a few business processes . The high redundancy levels these processes require are extended across the data center, regardless of other systems and applications often substantially lesser availability requirements, downtime risks, and business significance . Not surprisingly, increased redundancy of mechanical/electrical/plumbing (MEP) infrastructures, computer networks and subsystems, uninterruptible power supplies (UPS), telecommunications equipment, and the like equates to increased overall costs . To further complicate matters, this substantially increased cost and complexity come at a time when executive management and IT organizations are struggling to gain control over data center operational frameworks and are operating within tight budgetary constraints .

The data center dilemma

The complexities involved in planning, designing, and deploying todays critical production data center environments have increased exponentially in recent years . This is largely attributable to a growing demand for highly available operational frameworks capable of supporting high-density technology systems, always-on applications, and aggressive business-service delivery models . The best-known data center classification system, developed by the Uptime Institute, defines data center availability as falling within four tiers (Figure 1):

Traditional data center designs

Single-tiered design Conventional data centers are designed monolithically, with technology systems and applications located together and designed to operate within a single tier level . To cover all availability contingencies and provide for future growth, these monolithic facilities are typically built to the highest-affordable facilities tier level and incorporate fully redundant MEP and technology infrastructures that feed all IT systems and applications regardless of their availability risks or business importance .
*

HP proprietary data based on customer modeling

Figure 1. Uptime Institute Tier Classification System

Tier 1

No Generator, or optional

Basic UPS for LAN Room, non-redundant N+1 UPS with redundant components N+1 UPS with redundant components

Single Utility or on Radial line from Loop

99.671% Availability

Tier 2

Generator

Single Utility Feeders, N+1 Mechanical System One Active, One Passive, Utility Source, N+1 Mechanical System Dual Active Utility Feeders, 2N Mechanical System, compartmentalization

99.741% Availability

Tier 3 Concurrently Maintainable

N+1 Generator System

99.982% Availability

Tier 4 Fault Tolerant

2N Generator System

2N UPS Systems

99.995% Availability

Tier 1Non-redundant capacity components (single uplink and servers) . Tier 2Tier 1 + redundant capacity components . Tier 3Tier 2 + dual-powered equipment and multiple uplinks . Tier 4Tier 3 + all components are fully fault-tolerant . Everything is dual-powered .

In this monolithic model, the physical raised floor space is generally designed as a single footprint comprising systems/applications within a mixed technology environment . This framework requires an anywhere anytime mindset, inasmuch as the systems/applications are deployed into any available rack/cabinet within the data center space .

requirements, the monolithic model is bound to result in wasteful over-provisioning and over-investment of precious capital . (It is estimated that as many as 50 percent of the applications running in the typical data center could be classified as less than business-critical .)

Moreover, while highly robust for accommodating technology systems throughout the data center space, the monolithic approach tends to be expensive from The single-tiered approach readily accommodates a multitude of diverse technology solutions and provides an operations perspective . In addition, it requires the entire facilities infrastructure to be continuously a consistent IT architecture throughout the entire space . But it harbors inherent efficiencies and financial operating at full throttle, with unnecessary redundancy continuously driving down energy problems . efficiency and driving up energy costs . Single-tiered designs deploy only one operational Multi-tiered (pod) design model that must accommodate the mixed-tier Many data center operators have recognized that applications environments that may exist across the higher levels of operational efficiencies could be facility . There is no way to easily segment missionrealized through a multi-tiered design model critical applications from less-critical applications so that supports high-density technology deployments as to meet varying availability and business-impact while simultaneously accommodating lower-tiered requirements . These disparate applications reside within the same raised-floor space, although their MEP systems/applications within the same data center facility footprint . This involves carving the data center operational requirements are apt to differ widely . into segmented pods that represent independent We have seen that increased redundancy inevitably raised-floor spaces capable of supporting individual drives higher upfront construction costs . By providing environments that are fed from dedicated, low- and medium-priority applications with redundancy appropriately equipped facilities infrastructures and reliability levels that far exceed their actual (Figure 2) .

Figure 2. Typical pod design modelAvailability/redundancy levels remain identical across all pods

Monolithic tier

A-side B-side

MEP Systems/Infrastructure (tier IV) A-side Pod-1 Pod-1

MEP Systems/Infrastructure (tier IV) B-side

Pods are nothing new . Co-location operational models deployed by service providers and enterprises have been utilizing the pod design concept for years . Recently, enterprises have also begun to use this approach to provide segmented raised-floor pods that can accommodate high-density IT systems and at the same time segregate them from, and negate their impact upon, blade servers, rack-mount, and/or legacy systems and applications . This approach is highly inefficient for deploying highdensity systems with the appropriate segmentation, because each pod is fed from the same high-level MEP infrastructure that is engineered for the overall facility, regardless of application criticality . Moreover, it shares many of the operational cost and efficiency drawbacks of single-tier designs .

by a structured business priority analysis process . Business-critical applications, for example, would be deployed only within pods designated for high-uptime Tier 4 facilities operational models . The multi-tiered hybrid approach carries added advantages for companies that utilize chargeback models for internal cost tracking . With the establishment of multiple data center tiers, business units can better apportion right-sized availability and redundancy to target systems/applications, and better determine how much they need to budget for their data center requirements . Various tiered applications are deployed within designated pods that provide the required operational levels . To deliver the appropriate levels of fault-tolerance, high-tier pod environments would be equipped with fully redundant feeds to all IT equipment, redundant computer and telecommunications networks, industrial-strength security hardening, and redundant power and cooling and other mechanical/ electrical infrastructure components . Less-than-critical applications would be deployed into pods engineered on lower-level operational models . Right-sizing each pods mechanical and electrical infrastructure and redundancy technology in this manner opens the door to substantial capital (CAPEX) and operational (OPEX) cost savings . Upfront data center construction costs can be reduced by some 15 to 25 percent . And important efficiencies can be realized in areas such as energy usage and day-to-day facilities and IT support requirements .

HP multi-tiered hybrid design

Innovative topology HP has introduced a future-focused innovation for data center optimization from business, technology, and financial perspectives . HP multi-tiered hybrid design takes the pod concept to the next level by engineering single data centers to deliver multiple operational environments . Using this model, data centers may be designed with tiers of varying capabilities to meet specific technology and business requirements . For example, the model can be applied to design a data center as both a Tier 2 and Tier 4 facility and enable the appropriate level of facilities infrastructure to be assigned to each designated pod (Figure 3) . Each pod would be engineered to match the availability requirements of the specific technology systems/ applications tiers deployed within it, as determined

Figure 3. HP multi-tiered hybrid design model

MEP/Infrastructure tier IV A-side

Tier II A-side Multi-tiered Tier IV A-side

MEP/Infrastructure tier II A-side Pod-1

Tier IV B-side

Saved space

Pod-2

No B-side (space saving)

Pod-1 Pod-2

MEP/Infrastructure tier IV B-side

Business focused design methodology Integral to HP multi-tiered hybrid design is a unique process methodology that involves close collaboration between the HP consultant/design teams and leaders of clients IT and facilities organizations . In fact, a key goal of the HP methodology is to break down the walls that have traditionally kept enterprises IT and facilities personnel apart, as well as those separating network, mainframe, desktop, storage, applications, and other intra-IT disciplines . HP has found that a large measure of the overprovisioning and overinvestment that have characterized data center construction can be traced to a lack of communication between these traditionally independent entities . Their collaboration and integration are vital for achieving an appropriate balance between data center availability requirements and capital and operations costsa primary objective of the HP design approach . The multi-tiered hybrid design process begins with production of traditional schematic design, design development, and construction documents for both the technology and mechanical/electrical infrastructures . However, the HP methodology then departs from traditional approaches based on a recognition that significant efficiencies and cost savings can be achieved if the nature of the applications intended to be run in the data center is considered in a systematic way during the critical planning, design, and configuration phases (Figure 4) .

Central to the process, therefore, is the determination of critical business impact and urgency levels for each prospective application . In a banking environment, for example, an application providing authorization to withdraw funds from an ATM machine may be classed as having high impact and high urgency, whereas an application providing for overnight transfer of funds from one account to another may be classed as having high impact but only medium urgency . The combination of the applications business impact and urgency defines its priority level . This analytical process is carried out within the context of overall program management for data center optimization . Once the priority of each software application has been identified, the number and type of data center tiers may be determined . Each application is mapped to a tier that outlines the level of reliability and redundancy corresponding to the applications business priority . Applications with critical priority may be populated on computer equipment within a Tier 4 environment . Applications with medium priority may be assigned to Tier 2 or Tier 3 environments . Applications with low priority may be slated for Tier 1 environments . The capacity of each tier may then be estimated based on individual performance requirements (required processing power, memory, network bandwidth, and so on), estimated physical size of the data center tier, and/or estimated power density of the data center tier .

Figure 4. Business impact analysis

Obtain list of applications

Analyze facilities, infrastructure applications

Estimate cost

Categorize/Classify based on business impact Define priority Map categories and infrastructure to (right sized) DC tier Model capacities Estimate current and go forward costs

Define business impact and urgency

Rationalize tiers

Define priority level

Estimate cost (2nd pass)

Process within overall program management

Estimate project cost: facilities, infrastructure, virtualization and migration costs Calculate projected Return on investment (ROI), obtain green light for detailed design

Determine data center tiers

Modify capacity

Create Detailed design Finalize ROI, obtain green light for implementation

Determine capacity of tiers

Final cost estimate

Implement Check Value Realization

Additional steps can enhance the data center design based on financial considerations, ultimately leading to arrival at an estimated total capital cost based on the number of data center tiers and their capacity . At this point, the previously identified data center tiers can be analyzed (rationalized) from a financial perspective to determine whether any tier consolidation may be achieved . For example, where there are a large number of low- and critical-priority applications and a small number of medium-priority applications, it may be cost-effective to design a data center with a Tier 1 section for the low-priority applications and a Tier 4 section for the critical and medium-priority applications, rather than creating an additional Tier 3 section dedicated exclusively to the relatively few medium-priority applications . This strategy is based on the fact that a minimum fixed cost is associated with the construction of each data center tier . If appropriate, a new cost estimate is generated as a result of this rationalization step . This process may be repeated numerous times, each iteration modifying various characteristics to optimize them from a business and financial perspective .

Prospective business outcomes

Enterprises deploying data centers built on the HP multi-tiered hybrid design model can look to realize immediate and long-term financial and operational advantages: Capital cost savings: Upfront center construction costs can be cut by 15 to 25 percent* . Compare, for example, a 100,000 sq . ft . raised-floor data center designed to a monolithic Tier 4 specification (with power density of 100 watts/sq . ft .) with one designed as a 50:50 split between Tier 2 and Tier 4 infrastructure based on analysis of application criticality and priorities (Figure 5) . According to HP customer data, the original capital cost of building the Tier 4 facility would have been $184 million . But the data show that moving to a multi-tiered hybrid design reduces this to $138 million, a savings of $46 million . High levels of energy efficiency: Each pod is rightsized to meet the appropriate operational model requirements . Dynamic cooling capabilities can provide targeted capacity to both high-density technology and less-demanding pod environments .

Figure 5. Multi-tiered hybrid design capital cost savings

Multi-tiered hybrid design

Monolithic: Data center with high-level availability only $184 million Align applications with facilities $46 million

Multiple availability levels within data center

High

High Medium Savings (High + Medium = $138 million)

Lower operations costs: Not only does an HP multitiered hybrid design data center save on energy outlays . Redundancy is built-in only where it is needed, and the entire data center operates not as a high-tier facility but within a hybrid mode that can incorporate high, medium, and low operational tier models . By avoiding overprovisioning and unused redundancy of IT and MEP infrastructure, this model requires less maintenance and support . In addition, it enables multiple systems/applications tiers which might otherwise be housed in entirely separate facilitiesto share a physical enclosure, security systems, access controls, and the like . Enhanced ability to meet specific applications requirements: Each pod is engineered to match the appropriate facilities operational model (Tier1 through Tier 4) . Increased adaptability: For long-term viability and cost-containment, modular blocks in the power and cooling systems can be readily reconfigured to modify various pods redundancy levels . Increased operational flexibility: The segmented raised-floor pods are capable of supporting technology applications of widely disparate business criticality . Wider scalability: To provide for future growth, the multi-tiered pods are designed with built-in scalability based on projections from the business impact analysis .

Reduced business continuity risk: Systems and equipment failures are contained within smaller, more readily addressable, areas of the data center .

Conclusion
By helping todays enterprises work within tightened budgetary constraints and realize more cost-effective and energy-efficient operationwhile accommodating their continuing need to house higher-and lower-tiered technology systems and applications within the same facilityHP multi-tiered hybrid design points the way toward the data center of the future . Working with experienced HP consultants and design teams, clients can leverage this unique approach to build right-sized data processing frameworks that deliver consistently high levels of performance, scalability, and predictability . They can substantially enhance data center energy efficiency . They stand to save millions of dollars in capital costs and operational outlays . And they can position themselves to meet their evolving business and operational requirements for years to come .

For further information

Multi-tiered hybrid design is offered as part of the HP Critical Facilities Services delivered by EYP MCF . Additional information is available at www.hp.com/go/eypmcf

Technology for better business outcomes

To learn more, visit www .hp .com/go/eypmcf
Copyright 2009 Hewlett-Packard Development Company, L .P . The information contained herein is subject to change without notice . The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services . Nothing herein should be construed as constituting an additional warranty . HP shall not be liable for technical or editorial errors or omissions contained herein . 4AA2-4791ENW, March 2009