Disk arrays are at the core of any enterprise data center. But enterprise-class disk arrays carry high prices. These storage devices often use redundant components, such as power supplies, network ports and disk controllers to maintain operation when problems arise. Large storage platforms filled with spinning disks also demand a great deal of energy for power and cooling. In addition, annual maintenance contracts can be exorbitant.
SearchStorage.com has already covered the issues involved in purchasing strategies for a storage consolidation project. Below you'll find a list of the criteria for purchasing disk arrays from vendors such as 3PARdata Inc., EMC Corp., Hewlett-Packard Co., IBM and others.
Total storage capacity. Select a new storage array with enough capacity to replace several existing storage systems, as well as accommodate enterprise storage demand through the new array's expected service lifetime. Consolidating three arrays with 10 TB each would require at least 30 TB of capacity in the new array. But if storage demands have been increasing at 30% annually and you plan to keep the new array for five years, the new array
should be expandable to over 110 TB. Performing a capacity planning exercise up front can help predict storage capacity growth trends and needs.
Potential for clustering. Clustering allows storage administrators to interconnect several disk arrays so that multiple arrays can share capacity and connectivity. With clustering, if one box in the cluster fails, the other boxes can maintain storage availability. Clustering can be implemented in phases, so an enterprise can add arrays to the cluster as needed over time (thereby distributing the costs). However, consolidating several existing storage arrays into a new cluster can undermine savings in power and floor space.
Connectivity and traffic performance. Because consolidated storage concentrates storage traffic, network bottlenecks may appear at or around the new array. You'll want to evaluate the new array up front in your specific environment and use network performance tools to gauge traffic performance. While the array should offer an adequate number of ports for aggregation and failover, you may need to implement network upgrades or reconfigurations to ease any potential congestion. For example, a new array with more Fibre Channel ports may require a SAN switch upgrade to accommodate the added connectivity.
Internal disk storage performance. Data performance inside the box is just as important as I/O handling outside the box. Storage requests must be processed efficiently at the disk level. When you have multiple storage systems, the user traffic and disk activity is distributed among several systems. But keeping performance at an acceptable level when all internal disk activity and user traffic is consolidated in one box is a challenge, especially for transaction-oriented traffic.
Tiering. More and more disk arrays support storage tiering, meaning one array can provide online (Tier 1), nearline (Tier 2) and archival/backup storage, while automatically migrating data between the tiers. Tiering presents another issue since data will often need to be migrated between tiers within the same box, while handling outside traffic simultaneously. Large data migrations between internal tiers can multiply the internal disk traffic and impair outside traffic performance. Test the prospective array under load and use performance testing tools to gauge actual disk behaviors.
Reliability and availability features. The downside of deploying more storage in fewer arrays is that a serious fault in the consolidated storage array might render mission-critical data inaccessible. You'll want to consider all the reliability features. These include redundant components, such as power supplies, disk controllers and I/O ports; disk reliability features, such as vibration-isolated disk trays and RAID support; and RAID recovery features, such as pre-emptive rebuilds. The last feature reduces rebuild times by starting a disk rebuild process the moment a disk reports possible problems.
Power conservation. With consolidation, many organizations also realize savings in cooling/facilities requirements. Still, some consolidated arrays host a huge number of disks, so it's wise to evaluate disk power management and other green storage features. For example, some arrays can throttle back, spin down or completely power down inactive nearline disks to save energy.
Heterogeneity of bundled storage tools. You'll want to look at the management tools that come with the new array and determine its interoperability with the remaining arrays and other storage/switching devices. Ideally, the consolidation process will not add more management overhead. Array consolidation may make some existing management tools obsolete, so you'll want to know which management tools can be decommissioned by using the new management software.
Service contract needs. The total cost of ownership (TCO) for a new array should include purchase cost, projected upgrade costs, software licensing fees and annual service contract expenses throughout the anticipated lifetime of the new array. With more storage on fewer arrays, serious faults will have a greater impact. Take the time to understand the service contract and be sure that the manufacturer's service levels meet your requirements. You may need to negotiate for additional service levels, such as dedicated service representatives or shorter technical response times.
The disk array product specifications page in this chapter covers the following products:
- 3PARdata Inc.; InServ E200 Storage Server
- 3PARdata Inc.; InServ S400 Storage Server
- 3PARdata Inc.; InServ S800 Storage Server
- EMC Corp; Symmetrix DMX-4
- Hewlett-Packard; StorageWorks XP20000 Disk Array
- Hewlett-Packard; StorageWorks XP24000 Disk Array
- Hitachi Data Systems; TagmaStore AMS1000
- Hitachi Data Systems; TagmaStore USP1100
- IBM; System Storage DS8000 Series
- Sun Microsystems; StorageTek 9985 System
Sun Microsystems; StorageTek 9990 System