12th Annual Well-Connected Awards: Storage and Servers

What a year for the storage market. Hurricane Katrina tragically spurred demand for disaster recovery, while several large data thefts put tape encryption on the front page. Large companies went virtual to manage explosive data growth, which was partially fueled by the increase in replication. Meanwhile, though iSCSI remained an industry favorite, we found the most promise in intelligent Fibre Channel switches.

Products in the intelligent switch category, such as Cisco Systems' MDS 9506 Multilayer Director, allow applications to run directly on the switch, rather than be tied to your storage arrays. Some of these products support iSCSI, and more generally their functionality is limited only by the number of applications--such as virtualization, replication and acceleration--that they run. No iSCSI or array vendor can adequately implement what these products do in mixed-vendor environments. Intelligent switches eliminate the need for agents for virtualization, as well as the dependence on your array vendor to support replication to a different brand of array. In some cases, these products can even increase your storage applications' performance. Although it has taken years of hype, intelligent Fibre Channel switches are finally here.

Raft of Replication

The amount of data we must manage and the ways we have to protect it are driving us to try anything that will reduce our man-hour investment. With smaller budgets, fewer staff members and more data, it sometimes feels as if we're being asked to fight a war with limited ammunition. Enter the raft of replication, CDP (continuous data protection), and virtualization tools that promise to help.

Replication and disk-to-disk-to-tape solutions are letting us back up more data in less time. Important for reducing the ever-growing backup window, these tools now give us the information necessary to control what gets backed up more sensibly. Many CDP vendors are even implementing self-service restores, letting end users return their systems to the point at which they deleted or modified important files.

Replication is being used in some instances to eliminate tape, and in others to keep a disaster-recovery backup that is complete and nearly up to date. It is revolutionizing the way we think of backups. In the past year, a growing number of organizations left the old-world thinking of "weeks and months" for tape rotations, realizing they could replicate how and when they wanted to. With tape, we were limited by physical tapes owned, downtime for backups, and the number of cartridges our tape changers could handle. With replication, our main limit is the amount of disk space on the target machine.

A significant number of us have started replicating to a backup server and then backing up to tape from there--significantly reducing or even eliminating the backup window. Typically aimed at the small-to-midsize business market, these disk-to-disk-to-tape systems let us set up a single solution that receives replication or backup-style data, saves it to disk where it is easily accessible, then sends it out to integrated tape devices after a set time period. These systems also give us tools and policy engines for configuration and management.

As an industry, our response to this technology has been "Fewer work hours? Easy to use? Tell me more!" We haven't seen a sales increase yet, but we expect to as soon as vendors reduce prices.CDP is rapidly becoming a misnomer: One vendor's continuous has become another vendor's hourly. And if you're only going to get hourly updates, then a good replication product that offers differential replication on your schedule is just as effective, and possibly more useful, than a CDP product that only updates hourly. As this market straightens itself out and definitions are formalized, we expect increased adoption at the SOHO and SMB levels. To get copies of your files that are accurate up to a given minute is a worthy goal, and we expect the market to hit that goal. As of now, CDP is still immature, but vendors such as Revivio, StorActive, StoreAge and Troika Networks are breaking new ground and creating the definitions that will give us usable products in the next year or two.

We are also trying to use virtualization to reduce man-hour investment and improve storage utilization. After all, we made the move to centralized storage (be it SAN, NAS, or iSCSI) to better utilize disk space. Virtualization helps make even better use of our storage pools. It also makes life easier, if we're willing to accept that we no longer control where our data is stored. With the exception of very large organizations, the original purpose of storage virtualization--moving data between different tiers of disk drives based on cost, performance and reliability--seems to be gone.

We started out thinking we could use fast Fibre Channel disks for data that needed fast access and secondary SCSI and SATA disks for data that didn't require such fast access. Over time, however, SAS and SATA II blurred the definition of "fast access" to the point that such benefits are questionable.

Instead, something new emerged from our love affair with virtualization: a workable simplification of storage. With higher-end virtualization products from vendors such as EMC and IBM, we can say, "Give me two terabytes of storage," and the system provides us that space as a virtualized disk. It could not possibly get any easier to manage storage: no need to manually move LUNs about to make space, no worries about having one terabyte on one array and the other terabyte on another. The virtualization engine handles all that for us.

In return, however, we have to be willing to cede some control over what is placed where--a reasonable trade-off. The reduction in work hours and downtime is worth the insecurity of not controlling where data is stored that's introduced by automated LUN management.

Storage & Servers Categories
• Intelligent Fibre Channel Switches
• Disk-to-disk-to-tape
• Replication
• Virtual Machines
• iSCSI Array

• Tape Encryption
• Disaster Recovery

Storage & Servers Winners

Intelligent Fibre Channel Switches
Winner: Cisco Systems MDS 9506 Multilayer Director
There's only one way to say it: Cisco's MDS 9506 Multilayer Director rocks! It's not just that Cisco found a way to leverage third parties in making its "intelligent" switch actually act intelligent. It's that Cisco extends this functionality to some iSCSI too. The marriage isn't perfect--iSCSI doesn't get the entire array of security that FC gets--but it's a great start. This device shows where storage is headed--to the robustness and centralized management of Fibre Channel and the simplified networking of iSCSI. Applications for the MDS 9506 are currently limited to replication, backup and virtualization, but the architecture Cisco created opens nearly endless possibilities.

Disk-to-disk-to-tape
Winner: EMC Dantz Retrospect
EMC's Dantz Retrospect handles disk-to-disk backups innovatively, by running incremental backups to the server after initial configuration. The software lets you configure jobs that will stream data out to tape; you can even copy disk-to-disk without stopping the software from updating the newest copy. Since Retrospect is software, the amount of disk available for replication is a function of how much you're willing to spend. And low maintenance? Once it's configured, you just have to change tapes. Priced so that most businesses can afford it, Retrospect gave us what we all need at a price we'd happily pay.

Replication
Winner: Symantec Backup Exec 10d
Built on Backup Exec and Windows Volume Shadow Copy Service, Symantec's Backup Exec 10d is affordable, easy to use and backs up admirably. Designed for relatively frequent replicas and easy restores, the system streams data continuously to the backup server and then "snapshots" a backup at regular, definable intervals. You get a constant stream of changes saved, but complete replicas built only at predefined times.

Though Backup Exec 10d requires agents on every machine it protects, Symantec stuck to its roots by making this time-consuming task simple with an autodeployment feature on the server. To protect a small group of servers without wasting valuable IT manhours, Symantec Backup Exec 10d is ideal.

Virtual Machines
Winner: VMware GSX Server 3.2
VMware GSX Server 3.2 is a virtual machine that allows you to run multiple operating systems as independent "systems" within a single physical server. For server consolidation, this is an administrator's dream. GSX has better management tools and support for more operating systems than its competitors, plus advanced virtual machine functionality, such as virtual CD-ROMs. Though the product lacks the virtual cloning capability of VMware Workstation, it's still the best data center virtual machine we've tested.

iSCSI Array
Winner: EqualLogic PS200E
In the world of iSCSI arrays, there are two vendor categories: EqualLogic and everyone else. We've beaten up every version of EqualLogic's iSCSI storage device, and they all just take it. With performance numbers that rival a Fibre Channel SAN, along with simple configuration, great expandability and rock-solid hardware, we find EqualLogic's products to be the perfect fit for the small-to-midsize enterprise market. In our Green Bay, Wis., Real World Labs®, we use the PS200E to support tests on other products because the machine performs so well that it's never the bottleneck. This device just keeps running, no matter how much we abuse it.

Tape Encryption
Winner: NeoScale Systems CryptoStor Tape FC 704
A vendor's ability to encrypt data has become mainstream. We've been watching how devices encrypt, what their architectural choices are and how reliable the solutions are. NeoScale has all the right answers. With the CryptoStor Tape FC 704, you don't need one appliance per port to maintain wire speeds. The architecture allows for a clustered configuration that offers redundancy and sharing of security information between appliances, with FIPS-standard encryption. Adding appliances to the cluster is simple, and doing so increases the volume of data that can be processed at a time. If your tape drives are block-based--Fibre Channel or iSCSI--and you need a way to encrypt them, NeoScale has the solution.

Disaster Recovery
Winner: EVault InfoStage

In our RFI for disaster-recovery products, EVault's remote replication solution excelled. Because Evault has agreements with Tier 1 hosting providers, your data is stored in a secure location of your choosing. EVault's ability to ship you a local restore solution in the event of a disaster is a huge benefit.

Support for near-real-time replication means the difference between the last backup and the point at which your data center goes dark is so small that Evault InfoStage could double as your backup solution.

This vendor has agreements with SunGard and other vendors to provide services such as consulting and redundant data centers. While the price is steep, it's reasonable compared with the cost of leasing space, filling it with servers and maintaining them.

Don MacVittie is a senior technology editor at Network Computing. Write to him at dmacvittie@ nwc.com.