Some storage buyers willingly pay twice as much for consolidated enterprise storage as for decentralised, server-dependent storage. The reasons are found in user reports of the cost of managing storage, the cost of downtime, and the value of uptime.
Based on the results of numerous case studies, IDC currently believes that, in contrast to server-dependent storage, consolidated enterprise storage, properly implemented, may offer dramatic improvements in the cost of managing storage while in many cases improving application availability, avoiding costly losses, and adding business value.
Storage is a fast-changing technology, and conclusions reached today may not be valid in six months. Therefore, the long-term value of this bulletin lies in providing a framework for analysis, not in the ephemeral conclusions.
Study Background
IDC designed a study to evaluate the factors leading to an enterprise storage decision, describe, and to the extent possible, quantify the benefits that users report as well as to gain user guidance in determining the critical success factors for enterprise storage implementation.
IDC conducted more than 20 hours of in-depth interviews with data center managers and storage managers at companies where enterprise storage was installed. No attempt was made to gain a statistically significant sample size; rather, IDC employed a case study methodology of reported success stories. Participants came from multiple industries including health care, banking and finance, insurance, energy, retail, outsourcing, and government.
Installed disk capacity across all servers ranged from as little as 2TB to as much as 10TB. All participants had multiple environments: mainframe, Unix, and Windows NT. IDC included sites that were migrating applications from one operating environment to another. Virtually all sites were simultaneously connecting to multiple Unix servers, multiple Windows NT servers, or, in some sites, mainframe, as well as multiple Unix or Windows NT servers.
The results of the interviews are presented in an attempt to show data center and storage managers how to evaluate the inherent benefits of enterprise storage and determine their ability to meet the critical success factors.
Storage Strategies
For purposes of the case studies and this report, IDC evaluated three strategies for meeting the disk storage needs across an enterprise: Consolidated enterprise storage Co-located, server-dependent storage Decentralised, server-dependent storage
Enterprise Storage
Enterprise storage is consolidated storage that is capable of supporting multiple operating systems and multiple mainframe, Unix, and Windows NT servers on a single storage system.
Co-Located, Server-Dependent Storage
Co-located server-dependent storage means storage located in the data centre; each storage system is capable of supporting only a single server.
Decentralised, Server-Dependent Storage
Decentralised, server-dependent storage is located outside the data centre in multiple sites; each storage system is capable of supporting only a single server.
Bottom-Line Conclusions
Highlights of the study`s results are as follows: IDC research identifies explosive growth of data within corporations. Growth is occurring both inside and outside the data center. The nature of client/server architecture has changed from the original uses of word processing, spreadsheets, and e-mail with shared printers to business processing utilising extensive databases and networks. With these changes has come significantly greater technological complexity, and business unit managers are requesting that IS assume the technology management role. IDC has identified a desire among many data center managers to recentralise management of decentralised storage, especially in production environments for business processing and large-campus collaborative computing.
This recentralisation has come at a time when the use of distributed and networked servers is rapidly rising. The recentralisation effort usually comes under the control of the mainframe storage manager. Many of these storage managers are choosing enterprise storage, a common storage platform for mainframe, Unix, and Windows NT servers. Hardware costs for enterprise storage are comparable to and, in some cases, higher than those for mainframe storage and are often twice those of server-dependent storage. All other factors being equal, IDC believes that storage managers would purchase the cheapest storage available. But according to this study`s participants, all factors are not equal.
Storage Management Costs
Calculating Management Costs
Based upon the interviews, IDC finds two major components to storage management costs: the storage-management efficiency factor and burdened labour costs.
The Management Efficiency Factor
One motivation for the shift of data management from users to IS is that enterprise storage users see a dramatic difference in the efficiency of storage management between mainframe and enterprise storage and decentralized server storage IDC compared the storage management efficiency of sites by calculating the gigabytes of capacity managed per storage management employee . IDC found that storage management efficiency was significantly greater by a factor of 7.5 to 1 in enterprise storage than in the decentralized, server-dependent storage environment. Co-located server dependent storage achieved a ratio one-half that of enterprise storage -- 3.75:1.
Labor Costs
Participants reported that qualified storage managers and other operations staff are increasingly hard to find and keep. As a result, labour costs have risen sharply. By one participant`s estimate, the average burdened cost (salary plus benefits) of a mainframe or enterprise storage manager approaches $100,000 per year. The same participant estimated the average burdened cost of a decentralised storage manager is $50,000 per year. This 2:1 labour cost ratio in favour of decentralised storage offsets the 7.5:1 ratio of data storage per employee.
Building our cost formula, IDC finds that the management cost of storage is a function of labor cost and management efficiency as measured by gigabyte per storage manager. Doubling the labor cost per employee while increasing management efficiency by a factor of 7.5 improves management cost on a cost-per-gigabyte basis by a factor of 3.75. The formula is as follows:
Labor expense/GB per storage manager = Management expense per GB
Two of the primary contributors to storage cost (assuming equipment is under warranty) are management cost and hardware expense. Written in a formula, the relationship is expressed in the following manner:
Management expense + hardware expense = Storage expense
Assuming that server-dependent storage can be purchased for 50% of the price of enterprise storage (a figure IDC believes reflects the realities of the marketplace), the interviews demonstrate that in these sites, the lower purchase price of server-dependent storage is more than offset by the higher management expense. The labour cost is a fraction of the depreciation expenses for enterprise storage. The total cost per megabyte is $.65 per year. In contrast, the labour costs for server-dependent storage are double the depreciation expenses, resulting in a cost per megabyte of $.76 per year -- a figure 17% greater than that for enterprise storage.
In fact, if, as IDC believes, management efficiency will improve faster on enterprise storage than on decentralised, server-dependent storage, and the hardware cost for enterprise storage remains at two times the cost of decentralised storage, the total cost advantage of enterprise storage is likely to increase over time.
Factors in Management Efficiency
Factors affecting the improvement in storage manager efficiency will include asset tracking, capacity planning, and performance monitoring and tuning.
Asset Tracking
The ability for a storage manager to consolidate information on the amount of data capacity available in the company is dependent on the strategy being followed by the company. For companies depending on distributed server storage, a major challenge for the storage managers is simply knowing how much storage is associated with each server. "I can waste DASD," stated one hardware manager, "because I don`t know where it is, which makes my price much higher. I can buy cheaper DASD, but if I can`t find it or manage it, it becomes very expensive DASD." In contrast, enterprise storage offers a more manageable solution.
As a representative from one site stated, "When you take enterprise storage and you carve it up, you know how much physical DASD is in there and you know how much logical DASD, but when it`s spread all over the floor on different servers, you don`t know what you have anymore. With enterprise storage it`s very easy to visualise what you have and where it is." IDC believes co-located, server-dependent storage is also easier to track than decentralised server-dependent storage, but not to the same extent as enterprise storage.
Capacity Planning
The job of capacity planning for decentralized, server-dependent storage or even for co-located server-dependent storage is more complex than for enterprise storage. Two factors influence this job. First is the ability of available storage to connect with the proper servers. As one user complained, "I know I have available disk out there. I just can`t get to it from the right server." The second factor is predicting where the next storage requirement will come from. As another participant stated, "There are no tools to predict open-systems storage requirements." Nonetheless, these factors are more easily managed with an enterprise solution.
Performance Monitoring and Tuning
Performance management is still very much an art, but it can be greatly helped through the adoption of common monitoring and management tools. Standards in this area are particularly hard to set when business units make independent decisions regarding data storage solutions because they often result in multiple monitoring tools of varying capability. One storage manager said, with decentralised, server-dependent storage, "I can`t find it, I can`t monitor it, and I can`t manage it." In contrast, users report that enterprise storage offers tools to monitor the performance of discrete storage systems, providing efficient use of storage resources including disk capacity, cache, and host connections.
The Cost of Downtime
Regardless of the storage strategy, one goal shared by every IT organisation is to move towards a 24 x 7 operation, and this requires the reduction and eventual elimination of downtime. Downtime can occur for many reasons, both unscheduled and scheduled, and downtime can be viewed from the perspective of the storage device or the application.
Unscheduled Downtime
Participants identified several enterprise storage attributes, which reduced unscheduled downtime. These factors include RAID protection, remote copy capability, and redundant and hot-swappable components, all of which provide higher data availability. One participant stated, "These are mechanical devices, so they`re going to fail. The difference is that with enterprise storage, I don`t lose access to data." In contrast, distributed server-dependent storage is less likely to have remote copy capability. Some business units may not have selected storage with RAID characteristics, for whatever reason, and are less likely to have hot swappable components nearby when a hardware failure occurs.
At first glance, 99.98% uptime appears to be excellent availability. In a 7 x 24 x 365 operating environment, 99.98% uptime translates to one hour of downtime per year. Analysis of the data reported by one participant demonstrates that a single hour of storage-related, unscheduled downtime per year, if it occurs on a broadly shared dataset, could equal almost 7% of the annual hardware cost of storage. In this participant`s case, the additional cost comes from lost productivity associated with 2,000 application programmers.
Scheduled Downtime
The single factor most affecting scheduled storage downtime is the time needed to install, reconfigure, or upgrade a storage system. The challenge for enterprise storage is that, for devices with multiple hosts attached, more than one host and more than one application may be affected. Changes to the storage device must be co-ordinated around the application availability requirements of the attached servers. Most participants reported that the scheduling requirements were manageable through careful selection of the applications attached to a single enterprise storage device.
The Qualitative Value of Uptime and Performance Data center and storage managers identified a number of qualitative benefits of enterprise storage, including customer satisfaction, customer retention, improved competitiveness, and job security.
Several outsourcing companies reported that they were competing against other outsourcing companies that claimed to provide continuous availability. Now, internal IT organisations are finding that they also must offer continuous availability to compete against potential outsourcing pressures.
Globalisation and the Shrinking Backup Window
The biggest improvement for uptime reported by participants came from reducing backup time. The most extreme case was a health care service bureau in which the time required to complete backups prior to installing enterprise storage had exceeded the hours in the day. After migrating to enterprise storage, the company completed backups nightly in two hours. The improvement allowed this organization to compete effectively in North America, Europe, and Australia. Other sites reported that by reducing the time for backup and other batch processes, they increased the availability of their online processes, providing better services for their customers and increasing their competitive advantage. This advantage can be achieved only through the enterprise storage strategy.
Service-Level Agreements -- Buying a Little Insurance
Service-level agreements, especially among outsourcing companies, increase the potential value of uptime and application performance. When pressed, however, participants with outsourcing companies stated that they have never had a customer claim against a service-level agreement. Rather, they viewed the enhanced performance and availability as "buying a little insurance" against the unexpected. "You never know," said a participant from one outsourcing company, "when a customer will decide to exercise a penalty provision."
Preventing Decentralised Disasters
More than just completing backups on a timely basis, some study participants worried about the reliability, completeness, and currency of decentralised data backups. One participant referred to the "unmanageability of client/server storage" and stated that with decentralised, server-dependent storage, "I`m trying to think of what it is that I`m not backing up that has not `gotten me` yet. A server that I`ve missed something on, a backup cycle that is not the right cycle. With enterprise storage, I know exactly what`s on that box." Participants reported that enterprise storage reduced the errors and omissions often associated with backing up decentralised data.
Critical Success Factors
Study participants identified a number of factors as being critical to successful enterprise storage implementation. These factors included the following: Consolidating servers (both mainframe and open systems) to common sites; matching application availability requirements to storage systems; performing a complete inventory of current installed storage (DASD, tape, and optical); choosing applications of appropriate size
Consolidating at a Common Site
Study participants reported that before installing enterprise storage, participants first co-located servers to the mainframe data centre because of SCSI distance limitations. This co-location brought with it the added benefits of improved asset tracking, environmental and power conditions, and security. These benefits were offset somewhat by the higher cost of data-centre floor space compared with that for office floor space.
Matching Application Availability Requirements
Participants reported that reconfiguration of enterprise storage required scheduled downtime and the involvement of the vendor`s customer or systems engineer. Therefore, they concluded that consolidation on enterprise storage requires careful selection of applications that share common application online requirements. As an example, some service bureaus might avoid consolidating the storage requirements for applications supporting European daytime business applications with those supporting U.S. or Australian daytime business applications. Otherwise, scheduled downtime required for system reconfiguration for one application may conflict with online demands of another application. Enterprise storage users are especially careful regarding the placement of applications where storage requirements change frequently.
Completing an Inventory
Sites reported that completing a thorough inventory of servers, storage, and application workload type and performance requirements was necessary before they attempted to consolidate onto enterprise storage. One user added that because of the high cost of tape operations in decentralised environments, data stored on tape should also be included in the inventory. Another user viewed enterprise storage as preferable to optical from the perspectives of both management cost and performance.
Choosing the Right Applications
Until the reconfiguration process becomes less disruptive, several participants expressed reluctance at consolidating small storage requirements from a large number of servers. The definition of small varied among study participants, but ranged as high as 100GB. These participants determined that enterprise storage was most appropriate for servers with large storage requirements. Still, these participants were consolidating multiple servers on a single storage device.
Looking Ahead
Increasing the Distance
Distance limitations between servers and storage remain a challenge for enterprise storage. Storage managers are looking forward to upper-channel serial interfaces to solve the problem of needing to locate the storage near the servers.
Eliminating Reconfiguration Disruptions
Users of enterprise storage are asking for tools that will allow them to reconfigure storage systems without disruption to the operation of the storage systems. Some also desire the ability to reconfigure storage systems without the involvement of the vendor`s customer engineer. Users report that tools that will simplify management and reduce disruptions associated with enterprise storage reconfiguration are under development.
Data Extraction and Data Sharing
Users report that new functions allowing the automatic extraction and loading of mainframe data to open-systems databases will provide value in the form of faster and more management-efficient database loads for processes that are scheduled and often repeated. Users report that the value of these capabilities will increase proportional to the number of data types supported and to the number of server types supported.
Conclusions
For many application types and environments, the argument for consolidated enterprise storage is compelling. The justification for enterprise storage comes not from simplistic cost-per-megabyte measures, but rather from the improvement in storage management efficiency, the avoidance of costly losses, and the enhancement of business value.
Based on the results of these case studies, IDC concludes that storage management expense may significantly surpass storage hardware expense in many data centres. Storage buyers and manufacturers that focus solely on hardware cost-per-megabyte measures and not on overall value will find themselves at a significant competitive disadvantage.
Share