Storage Pipeline: Special Report: Ultimate Enterprise Storage

Although the concept of instant access to all data all the time is extremely appealing, the key is to balance needs against costs. Only the most critical data must be available immediately in the event of a network or system failure, and priority is best determined during business-continuity planning. Costs can be reduced significantly by backing up noncritical data to tape, optical disc and other offline storage systems as opposed to replicating data in real time throughout the enterprise.

Enterprise RAID solutions and other redundant SAN-based technology is not cost-effective for small companies with relatively few employees, but tape backup and low-cost NAS (network-attached storage) systems can do the trick in most cases. An accounting firm or medical office may need immediate restore capability, for instance, while a service business--an office cleaning company, say--may be able to survive longer without instant data access.

Large companies with hundreds or thousands of employees typically need significantly more robust--and more expensive--storage systems that comprise a variety of technologies. A RAID array can suffer a hard-drive failure without losing any data, but it won't do much good when data loss stems from fire, theft, virus or user error. Here again, tape-backup solutions are essential in providing recallable static storage that can reside on- or off-site.

Still, tape alone does not a storage system make. Tape's shortcomings include media costs, capacity limits and limited shelf life, as well as the time required for backup and restore.

Improved data continuity often comes in the form of remote backup, which is often accomplished over a company's Internet connection. Although one of the most expensive forms of data protection--LiveVault's Online Backup and Recovery Service, for example, costs $1,250 per month for up to 50 GB of protected server data, making one month of service as costly as some backup appliances with equivalent capacity--remote storage replication overcomes most of the weaknesses associated with conventional storage-continuity solutions and is impervious to most threats to an organization's data store.

A RAID array is a collection of hard disks ganged together to create a storage space that leverages striping or mirroring technology to provide both capacity and redundancy. Conventional RAID arrays relied on SCSI drives for speed. But today's ATA (Advanced Technology Attachment) drives are fast enough for use in low- and midrange storage arrays, and the new SATA (serial ATA) drives are well-suited for use in RAID arrays because their point-to-point topology eliminates bus sharing and delivers 1.5 Gbps of bandwidth to each drive. A switch lets the SATA controller communicate simultaneously with all the drives, and the system scales up linearly as drives are added. Other RAID options, including Gigabit Ethernet and Fibre Channel, rapidly increase storage capacity and throughput, but also increase costs.

Built-in redundancy--duplicate power supplies and interfaces, for example--is critical to a quality RAID array. The array should be connected directly to a server and controlled by two separate HBAs (host bus adapters), so the storage system stays alive even if one HBA fails. The server also features built-in redundancy in the form of multiple power supplies, cooling fans, processors and memory banks.

Local mirroring--one RAID array duplicating the data on another--is also critical to successful storage, as is remote mirroring--duplicating the data on the RAID array off-site. Typically, a FCIP (Fibre Channel-to-IP) bridge is used to transmit data off the SAN and onto an IP network at the remote site.

Solid-state memory is another component of an ideal enterprise storage solution. An SSD (solid-state storage device) contains no spinning disks, just a lot of fast memory. An SSD can be connected directly to a server or installed in a SAN as a shared file-caching facility, providing the kind of ultrafast data retrieval regular hard drives could never match.

Today's high-end SANs are based on 2-Gbps Fibre Channel technology. But because Fibre Channel is not IP-based, mapping to WANs is difficult. SANs are usually implemented as separate islands, with FCIP bridges enabling SAN-to-SAN connectivity. The emerging iSCSI standard will eliminate the need for FCIP bridges. iSCSI converts a SCSI stream to IP packets, transmits those packets over a TCP/IP network, then converts them back to a SCSI stream at the receiving end. Moreover, you can install gateways to allow access to Fibre Channel fabrics via iFCP or iSCSI (Get more on iSCSI SANs).

If for any reason you can't or don't want to set up and manage your own remote storage facility, you can contract with a service company like LiveVault that will do the job for you for a fee. LiveVault partners with Iron Mountain data centers to offer online server backup, off-site electronic vaulting and data recovery. The data centers, which are housed in concrete, single-story, steel-reinforced, earthquake-resistant vaults protected by steel doors and state-of-the-art security systems, are guaranteed to be within three miles of police and fire stations and far away from flood zones, fault lines, airline flight paths and high-crime areas; they're also guaranteed to be nowhere near any area that has been flooded within the past 500 years.

Smaller companies that find online backup cost-prohibitive should consider the ever-economical DAT tape drive. DDS-4, the latest incarnation of DAT tape, offers 20 GB of native capacity and 40 GB compressed, with a compressed transfer rate of 6 Mbps. DDS-4 drives can read and write to DDS-2 and DDS-3 media. DDS-2 has a native capacity of 4 GB and compressed capacity of 8 GB, while DDS-3 has a native capacity of 12 GB and a compressed capacity of 24 GB.

SDLT (Super Digital Linear Tape) and AIT (Advanced Intelligent Tape) provide additional storage capacity as well as speed--for a price. SDLT tapes are available with native capacities of 40, 80, 110 and 160 GB that double with compression.

The maximum transfer rate of an SDLT 320 drive is 16 Mbps, and higher drives are backward compatible with lower-capacity units. AIT-3 drives have 100 GB of native capacity and offer transfer rates of 24 Mbps with compression. AIT-3 drives are backward compatible with AIT-1 and AIT-2 cartridges.

The ultimate backup unit, though, comes in the form of a jukebox that contains multiple drives and tape banks and has a mechanism that shuffles the tapes in and out of the drives as needed. Major players in the jukebox market include Hewlett-Packard, IBM and Quantum, which offer jukeboxes with terabytes of capacity for the enterprise.

Even with the optimal storage hardware in place, of course, you need the right software to make the most of a SAN. Storage software should be based on industry-standard architecture such as IP, SCSI, iSCSI or Fibre Channel. It must provide a complete, end-to-end network storage infrastructure with services for data and capacity management, data availability and data recovery, and it must be compatible with your existing storage hardware. Although Microsoft's Windows Server 2003 offers integrated storage-management utilities, most storage solution providers recommend using storage utilities from Computer Associates, EMC, FalconStor and other storage specialty companies.

Computer Associates' BrightStor SAN Manager serves as the central control for SAN devices from multiple vendors, enabling administrators to diagnose and solve most storage problems from a single console (and it topped all entries in Jon Toigo's storage-management RFI analysis, "Out of the Abyss," page 10). It automatically discovers and maps all SAN components and interconnections, creating a topology image of an enterprise's SAN assets and showing logical volumes, zones, zone sets, LUNs and routes within the SAN. It also provides drill downs to each device and displays device status.

EMC's ControlCenter provides an end-to-end solution for multivendor storage resource, network and device management. ControlCenter manages the host, SAN and arrays, regardless of manufacturer, and lets administrators improve performance and resource utilization, automate provisioning and increase productivity. It also enables the management of direct-attached, SAN-attached and NAS storage arrays, providing drill-down views into the subcomponents of supported arrays. EMC's software is compatible with hardware from Brocade, Hewlett-Packard, McData, Qlogic and other vendors.

FalconStor's IPStor manages complex storage environments from a single Java-based console, providing unified storage services regardless of vendor, connection protocol and disk class. It allows the unification of SAN and NAS systems under a single storage farm, and lets administrators migrate data smoothly with little or no downtime.

Of course, no matter how much you invest in hardware and software, the key to creating an efficient, comprehensive enterprise storage system is thinking it through from the start to ensure compatibility, manageability and scalability when your business grows. Do it right from the get-go and you'll be able expand your SAN while maintaining all the functionality and automation you built in from the beginning.

Frank Ohlhorst, CNE, is a technology editor at CRN. Write to him at [email protected].

Marc Spiwak is an associate technical editor at CRN. Write to him at [email protected].

• "Backup Plans Should Consider Local Data"

• "Ultimate White-Box NAS"

• "LiveVault, Iron Mountain Team Up for Fail-Safe Backup Service"

• "EMC DMX Technology Spurs Performance"