Combining Hardware Fault Tolerance with Windows Server 2003 Technologies

Windows Server 2003 possesses several technologies that improve system availability and recoverability with Direct Attached Storage. These same technologies work very well with NAS and SAN devices as well.

Distributed File System with NAS or SAN

Distributed File System has come a long way with Windows Server 2003. The ability to pre-stage data, the ability to mirror the DFS root, and improvements in control of replication topologies have all helped companies to adopt and take advantage of Distributed File System. Although things have gotten better, many administrators still find that there are too many limitations to the functionality of File Replication Services. FRS can't deal properly with roaming profiles and it often replicates data unnecessarily. For example, most antivirus products flag files that they have recently scanned. FRS will misinterpret this as a modification to the file and will queue it up for replication to all of its partners .

Many administrators are finding that the native replication functions of NAS and SAN devices are a much better way to maintain replication for DFS replicas. By having Windows Server 2003 servers mount those replicas and present them as nodes in the DFS tree, they are able to achieve enhanced usability in their DFS structures. This adds an additional layer of redundancy because NAS or SAN mirrored replicas are automatically presented to users if a local copy were to fail.

Leveraging Logical Disk Manager

Most administrators are very familiar with the Logical Disk Manager that comes built into Windows. This is the interface that is regularly used to mount new disks, create partitions, and format drives . Administrators who are a bit more advanced have no doubt learned the benefits of being able to mount disks to folders rather than drive letters. Very large powerful servers that serve multiple tasks no longer have to worry about running out of drive letters .

Most SAN and NAS devices on the market offer management plug-ins that enable them to be managed through the existing Logical Disk Manager. Through hardware abstraction the remote disk is made to look like a local disk. This enables you to perform much of the management of the remote resources through an interface that they already know. This reduces the requirements to retrain administrative staff on a new technology.

Remote Storage Management

Remote Storage, the Windows version of Hierarchical Storage Management (HSM), enables you to increase disk space on a server without physically adding more hard disks. Remote Storage automatically moves data back and forth between high-cost, faster disk drives and low-cost, high-capacity storage media. This traditionally meant moving data from Direct Attached Storage to either a lower performance server or a very low cost but high capacity storage media such as a tape library or rewritable DVD or DC media. Remote Storage monitors the amount of space available on local NTFS volumes , and when the amount of free space dips below the low water line, it transfers eligible files from the primary storage to the secondary storage. Users can still see and access archived files but they are actually seeing links to the files stored on the lower cost media. This frees up storage on the file server without requiring the purchase and installation of additional hard disks. The Remote Storage service also works with the Removable Storage service to access any removable media used for remote storage.

In an environment that uses NAS or SAN for storage, this concept can be moved up one tier . In most cases, the NAS or SAN device will replace a number of servers. These are servers that would normally be retired . Rather than scrap all the servers, you can take a subset of the servers and salvage the disks from all the systems and distribute them across the remaining servers. Those servers would be relegated to storing low demand data only. This would enable you to still lower the overall number of servers in the enterprise that require management.

Optionally these "lower tier" servers could be low-end servers with IDE RAID devices attached to them. Given the current prices and sizes of IDE drives, a 1-TB server could be built for well less than $5,000. Because it wouldn't be holding "primary" data, it would be acceptable to take the hit in performance and reliability that comes with IDE drives. This could allow for an environment where high demand and highly important data could live on redundant and high performance SAN/NAS devices. Data that hadn't been accessed in some time could be moved to the IDE RAID systems where the data would still be retrieved quickly. Data that hadn't been accessed in long periods of time would be moved to tape or rewritable media for long- term storage.

Integrating Backups with NAS and SAN

Many administrators have faced the challenge of dealing with narrow windows of opportunity for performing backups. Given a choice, every department would request that their systems only be backed up between 2 a.m. and 3 a.m. so that they wouldn't suffer a performance hit while users might be accessing them. This is especially difficult in environments where systems are used globally. The challenge with narrow backup windows is that data cannot be spooled to tape fast enough to meet the time commitments. One of the easiest ways to improve this performance is to perform the remote backups to disk.

By using a NAS or SAN device to buffer the backups, you create an environment where the NAS or SAN becomes the backup device. Backups are written directly to the NAS/SAN at impressive speeds. Outside the backup windows, the data is backed up from the NAS/SAN to tape for long term storage. The added benefit of this model is that 90% of restore requests are for data from the previous evening. By having that data on disk on the SAN/NAS, the restore process is also much faster than it would have been from tape. SAN devices often support what is known as serverless backups. This means that the SAN itself controls the tape device and writes the data to it without the need for a traditional backup server.

SAN-attached Tape Devices

SAN-attached tape devices range from single drive single tape devices all the way up to office- sized backup systems with dozens of drives and literally thousand of available tapes. Robotic arms drive on tracks around the device scanning and moving tapes back and forth between drives, storage slots, and mail slots.

Leveraging Disk Quotas on NAS or SAN Devices

Disk quotas were introduced as a long overdue feature in Windows. For a long time, you had to worry about users storing inordinate amounts of data in their user directory and filling up the entire volume. Any other users whose home directories were on the same volume would suffer through no fault of their own. In an environment that makes user directory space available on a NAS or SAN will appreciate the safety that comes with being able to enforce quotas through Windows Server 2003. Imaging having to explain to your management that your brand new 4-TB SAN is already full because your users have filled it with "personal" data.

Using Encrypted File System to Protect Files on the SAN or NAS

One of the greatest concepts ever put into use is the OSI Model. In a nutshell , the concept is that if you have a series of layers that work together than anything that can be abstracted to look like one of those layers will automatically work with the rest of the layers . In the case of SAN or NAS, a driver is used to make the remote storage look just like regular attached storage. After that is done, the rest of the system will treat the abstracted drive like another locally attached drive. This means that technologies like EFS will automatically work with the remote disk. Users can continue to encrypt data on the SAN or NAS and administrators can continue to manage EFS keys and agents the way that they always have.