Storage Area Networks (SANs) have definite measurable benefits over direct attached storage (DAS) as previously discussed in Part I of this article. They also carry additional costs over DAS as well. These costs include additional capital (SAN hardware, SAN software, Redundant Array of Independent (or Inexpensive) Disks (RAID) hardware and RAID software) and additional operating budget (maintenance, training and possibly more personnel). The key here to making a decision, is whether the measurable benefits of an enterprise moving to a SAN from a DAS environment for the first time, will outweigh the additional costs.
In other words, the combination of hard- and soft-dollar savings will yield the payback picture for the SAN infrastructure that can be used for management and technical decisions around SAN technology. Each enterprise will have to consider their infrastructure and determine which savings are hard, and which are soft. For example, hard savings are characterized by the following business planning case options:
- Asset utilization.
- Defer storage purchase.
- Reduce tape systems.
- Reduce floor space.
- Increase availability.
- On-line recovery.
- Storage on demand.
- Increase disk.
- Storage maintenance.
- Reduce number of servers.
- Avoid network costs.
- Non-disruptive growth.
- Improve LAN/WAN.
- Backup Window.
- TB-per-system application (SA) Ratio.
- Reduce backup servers.
And, soft savings are characterized by the following business planning case options:
- I/O, data movement.
- New Dr capability.
- Protect critical data.
- Reduce CPU load.
- Impact to application development/test.
- Vendor consolidation.
- Increase disk life.
- Management $dollars versus storage $dollars.
- Server clustering.
- Improve DR.
- Terabytes (TB)-per-data base administration (DBA) ratio.
- Secondary security.
- Application migration.
- Extend server life.
Estimated savings can be characterized for conservative, aggressive, or nominal states (calculations) and should be considered in the analysis. Conservative states usually will include hard savings and highly likely savings areas (seen or predicted), while an aggressive approach will take hard- and soft-savings examples together for the return on investment (ROI) calculations. Nominal cases will take a middle-of-the-road approach and will be consistent with the most acceptable/believable ROI case studies. Aggressive estimates are often used when baseline data is not verifiable or for soft-dollar savings. Each of the business planning case options identified in Part I and II of this article have different hard/soft considerations, and tend to be best applied in an approach that is clearly conservative, nominal, or aggressive. For example, the following are rules for simple ROI determination:
- To achieve simplicity in ROI, cash flow, internal rate of return (IRR), and depreciation are not factored.
- Moore’s Law (the pace of microchip technology change is such that the amount of data storage that a microchip can hold doubles every year or at least every 18 months) is used with price stability for all servers and disk price decline.
- Lease costs converted to purchase costs.
- All hardware costs can be pulled to year 0, at Net Present Value 0 (NPV0).
- Hardware costs can be spread out over the time period in which it is acquired.
- Three-to-four-year planning usually is adequate for most SAN payback models.
Calculating SAN ROI
The following steps are also part of a structured approach (methodology) for calculating SAN ROI. A detailed discussion of these steps follows:
- Gather baseline data (see examples in each case model discussed later in article).
- Define the proposed SAN topology and then determine SAN costs.
- Select ROI business planning cases that best meet the current and target optimum situations (typically, five to eight models are chosen for a business case justification).
- Run the ROI case models and create summary charts and conclusions.
- Review conclusions and make adjustments to topology type, assumptions, and parametric values of the ROI models.
Gather Baseline Data
There are several baseline costs that will be required in order to determine current- or future-year costs before starting a SAN ROI exercise. Each of these cost elements could be applied to any of the ROI business planning cases discussed in Parts I and II of this article. Also, each case will outline the necessary baseline elements needed. A listing of these cost elements are as follows:
- System administrator, database administrator (DBA), storage manager and tape operator.
- New storage (S/390., UNIX, Windows NT).
- Rate of storage price decline.
- Average unit cost of new servers (UNIX, Windows NT).
- Average unit cost of Windows NT or UNIX servers with SAN kits (h/w, s/w, middleware, setup).
- Expected increase/decrease of servers.
- Tape library systems.
- Tape media (Digital Audio Tape (DAT), digital linear tape (DLT), 8mm, other).
- Off-site tape storage.
- LAN operational costs.
- LAN degradation impact.
- LAN build-out costs.
- Data center floor space.
- Data center environmental costs.
- Application(s) downtime.
- Data off-line impact.
- Recovery time for different types of data.
- Revenue potential per person.
- Monitoring cost.
- Maintenance costs for all storage.
- Maintenance/license fee for storage s/w.
- Value of improved I/O performance.
- Time to procure/request net new disk.
Determine SAN Costs
The proposed SAN infrastructure has to have completed a logical design that will capture the design parameters and the list of components for the baseline costs in order to develop the proper payback models for the business case. This can be accomplished with a variety of services that can provide rough order magnitude (ROM) designs and costs based on information gathered from workshops or an onsite assessment. A preliminary list of SAN components (h/w, s/w, middleware, cabling, integration labor, etc.) can be summarized and priced, once the design is roughed-out and blueprints are available.
A useful comparative review of design and operational needs is provided by applying ROI models on more than one topology. Some topology design options can be very similar in price, but yield different total cost of ownership (TCO) cost streams or ROI payback terms. A balanced financial approach to the analysis and cost/benefit determination is created by having two or three topology options (network attached storage (NAS), point-to-point (PtP), FC switch or director, etc.). Since storage growth is assumed to continue regardless of the attachment architecture (small computer system interface (SCSI), FC, SAN, Enterprise Systems Connection (ESCON), etc.), some (but not all) of the SAN ROI analysis is completed without considering the detailed cost of storage. Thus, the cost of additional or replacement storage may need to be separated from any new SAN infrastructure costs when selecting and implementing a payback model/option. In other words, most ROI cases are prepared from the perspective of justifying the new SAN infrastructure, not necessarily the attached storage.
The future price of SAN infrastructure can be normalized to NPV cost or the average year costs, if the enterprise needs to migrate to SAN via PtP and small fabric installations (i.e., Fibre Channel Arbitrated Loop (FC-AL) or small switch). Future year, incremental growth will need to factor price erosion of future SAN components. It is often best to show total costs per usable SAN port as a comparison in the phased implementation, if an incremental approach to SAN is planned. Otherwise, for each SAN development phase, a separate ROI calculation would be needed.
Now, let’s very briefly look at a few of the business planning cases that enterprises can choose from (who are currently working in a DAS environment), when performing a SAN ROI exercise to determine whether they should consider a move to their first SAN. Not all cases are equal in value and applicability to the target environment. Some cases are more specific to very large or mature enterprise infrastructures. Others are best suited for small sites, any kind of service provider (XSP), decentralized sites, etc. Those considering a move to their first SAN should be familiar with the following options in order to choose those that best apply to the target environment.
SAN ROI Case Selections
In order to build a business plan/justification, most enterprise environments choose between six and nine ROI business planning cases. Depending on the payback methods needed in the analysis, each case has specific characteristics that may vary, such as:
- Achieving Service Level Agreement (SLA) targets.
- Growth attainment.
- Immediate cost reduction.
- Long-term TCO reduction.
- Operational efficiencies.
Next, let’s look at these cases. In the interest of the length, only the most popular or widely used of the cases will be discussed in detail in this article.
Disk Utilization Increase
Improved storage asset utilization is inherently provided by pooled storage. As shown below, DAS storage can vary by platform:
- Windows NT–20 to 40 percent average disk utilization.
- UNIX–30 to 40 percent average disk utilization.
- S/390–60 to 70 percent average utilization.
Disk Procurement Deference
In procurement, a pooled storage architecture provides better asset utilization and better efficiencies. However, future storage needs will emerge, regardless of how well the first consolidation effort goes. With larger, aggregated storage purchases possible, the future price of any disk bought will be cheaper per GB than storage bought today. Deferring as much capacity need fulfillment to the future as possible, will have a positive impact on near-term cash flow, since storage will be cheaper in the future.
Improvement Of Terabytes (TB)-Per-Administrator
Pooled storage increases the total capacity of data management that can be assigned to a DBA or storage administrator. Implementation of central, pooled, or SAN strategies, can reduce the staff growth requirements in comparison to staff needed to support server-attached distributed storage.
Data Center Rack/Floor Space Reduction
Server and disk consolidation represents real savings in data center floor space and environmental costs to the enterprise. Less space and power is needed to service data storage needs through aggressive consolidation and pooling of resources. These costs are true hard-dollar savings, and can start providing payback almost immediately. Since many enterprises are running out of space, the cost avoidance of building a new data center can be factored into the SAN infrastructure investment.
Tape Library Procurement Deference
Fewer tape devices will be needed to back up the enterprise data if the total disk capacity is smaller and the efficiencies of pooled storage work for the enterprise. For local backups of application data, OS, etc., some local drives that already exist can be used. However, to move off-line data from the central storage pool, larger enterprise tape libraries will be used. Many enterprise have multiple tape stackers, local digital linear tape (DLT), and 8mm drives attached to servers. Many of these can be replaced in favor of a better central library system. General estimates put a 15-to-25 percent tape unit savings through consolidation to larger silos. Depending on the decentralized state of the servers and the quantity of wasted tape space, efficiencies of 50-to-100 percent have also been observed. Since better tape utilization will be evident with pooled resources, fewer tapes will be needed. You can also calculate the effect of less time, handling, and the moving of tapes to off-site locations.
New Capabilities For Disaster Recovery (DR)
The separation of storage from the servers provides an opportunity to plan for recovery with better optimization for data, and also to manage data and processing separately. In order to create multiple copies of critical data in other parts of the SAN, data replication can be applied to storage (at the controller level, switch level, or OS level). Also, appropriate levels of protection can be applied, since specific datasets and relational database management systems (RDBMS) can be targeted according to the value.
Options For On-line Recoverability
According to recent storage industry estimates, about 80 percent of Fortune 1000. companies will implement replication-based backup methods for 20 percent of their data. Moving critical data to near-line or on-line areas for very fast restore, will require increased scalability and replication methods that can be found in some SAN topologies.
Data Path Availability Improvement
SAN and advanced storage options can increase the value of availability to the business; as well as, enterprise operations costs. The improvement value of having data more available to the enterprise can be factored into a business case for SANs.
General-Purpose UNIX. And Microsoft.Windows NT. Servers Reduction
Much of an enterprise’s Windows NT growth can be attributed to servers (up to 20 percent) that are installed for disk capacity only, not CPU cycles. This phenomenon is not as prevalent in UNIX environments. You can expect to see situations in which servers are
acquired for the sake of storage only as server growth continues. Thanks to the centralization that is enabled with the SAN, because the assumption here for this case, is that some percentage of Windows NT servers will not need to be procured. This projected number of servers that could be avoided is multiplied by the current or projected cost per server and shown as hard-dollar savings.
Avoid Upgrades: Improve LAN/WAN Performance
If data movement, backup, and replication is migrated to the IP backbone (assuming incremental data growth), network infrastructure will need constant upgrades (about every year). New FC SAN topologies off-load the traffic to the new storage infrastructure, thus relieving the network load and potentially deferring upgrades. Thirty percent of LAN traffic can be attributed to backup processes occurring over the net. You should be able to determine how many new sub-nets or virtual private networks (VPNs) will be required to support server-to-server transfers and network-based backup from projections on server growth. You can also determine the costs of new sub-nets (network interface card (NIC), hubs, and routers) from the projected growth. And, some portion of these costs can be counted as savings when avoided in favor of FC-based topologies.
Backup Servers Reduction And Elimination
Finally, there may be a specific instance in which to reduce dedicated backup servers that are servicing small sub-nets or groups of servers. Within advanced storage architectures, the role of dedicated backup servers is likely to continue, but the need for numerous small servers is diminished. By using larger pooled storage with tape connections to silos and bypassing the need for host processing (LAN-less backup), this function will determine how the need for small, dedicated backup servers can be reduced or prevented.
Summary And Conclusions
The results can be very subjective in any ROI analysis. There are no guarantees to the models, parameters, assumptions, and payback performances discussed in this Parts I and II of this article. SAN technology should not be considered on cost savings or strong ROI expectations alone. SAN is a business-enabling technology, which, when planned and deployed correctly, can have a tremendous impact on data management, growth, performance, and investment protection. There are opportunities for SANs to impact enterprise operations and expenditures financially in the future, but these should be secondary to SANs’ capabilities to help companies achieve business and operational goals.
In order to determine the best ROI results, trade-offs on values and parameters (conservative to aggressive) can be made within the spreadsheets, and the calculations run again. Summarizing savings by year and by category is the overall approach. Separation of soft-dollar savings and hard-dollar savings is highly recommended.
It is concluded that when choosing and implementing a long-term pooled storage strategy, that early cost and TCO planning in the design work is a valuable process to employ. In topology and design decisions, ROI is not the only determinant. Also, in any logical or physical pooled storage design, there are dozens of factors that need to be included. Some of the other technical and business considerations may include:
- Availability needs for the data.
- Chargeback costs to storage users.
- Geographic considerations of the servers, storage and staff.
- Improvements in data restoration.
- Performance requirements.
- SAN infrastructure serviceability and maintainability.
- Scalability and non-disruptive upgrades.
Finally, SAN ROI and TCO are important elements to consider in parallel with other design qualities. Before the design is committed to production, each topology has differing cost and payback characteristics that need to be explored.
So, after hearing all of the business case reasons why your enterprise should leave the safety of their DAS environment and move to a their first SAN, what is you decision in this matter? Do you think they should move? Of course they should, but only if the right process of choosing a vendor is followed; and, that can only be determined by vendor consolidation. In other words, pooled storage architectures can consolidate the number of vendors involved in providing infrastructure services. This reduced overhead impact can be measured and summarized. But, only time will tell!
About the Author :John Vacca is an information technology consultant and author. Since 1982, John has authored 36 technical books including The Essential Guide To Storage Area Networks, published by Prentice Hall. John was the computer security official for NASA’s space station program (Freedom) and the International Space Station Program, from 1988 until his early retirement from NASA in 1995. John can be reached at [email protected]