Effectively overseeing data storage across the network is a problem that all network managers face. Some estimates peg annual storage growth at 60 percent to 100 percent; with growth numbers like this, even small and branch offices can quickly find themselves in a real storage crunch.
Microsoft's recent release of Windows Storage Server 2003 R2 aims to help admins avoid that crunch by providing centralized management, optimized file sharing, and more efficient file replication between servers. New to R2 is Single Instance Storage (SIS), a very effective method of reducing data bloat by replacing duplicate files with a pointer to an original copy.
Other upgrades include performance boosts for CIFS (common Internet file system) and NFS (network file system) file serving, and management is now available in a single user interface. But don't look for Windows Storage Server R2 at your favorite retailer -- this is an OEM bundle only.
Microsoft as an appliance?
Storage Server R2 is Microsoft's foray into NAS (Network Attached Storage), delivered through its OEM deals with vendors such as Dell and HP. The hardware-software bundle provides an optimized, ready-to-run server appliance that can quickly integrate into the existing infrastructure. (Check out our review for more on HP's StorageWorks All-in-One 600, which uses Storage Server R2.)
Because Microsoft knows that not every network is a Windows-only affair and it wants to be a good corporate citizen, this release features much improved NFS performance for Unix/Linux users. Although not the fastest NFS out there, Microsoft says properly tuned Storage Servers have been clocked at 22,416 ops/sec using the SpecSFS benchmark; the enhancements to CIFS traffic showed a 15 percent to 20 percent performance increase compared with non-optimized Windows servers. After my tests, I did have a gut feeling that Storage Server R2 did perform better than my other "fat" Windows 2003 Servers.
I tested Storage Server R2 during the course of a few weeks. I used two HP appliances -- a 1U ProLiant DL100 G2 with 450GB of RAID 5storage and a ProLiant ML310 with 750GB in RAID 5. I installed the ProLiant DL100 in a branch office setting, connecting back to the main office and the ML310 via IPSec VPN. Both appliances were also added to an existing Active Directory and DFS structure. Remember that admins must run the ADPREP utility and extend the schema on their domain before they can take advantage of the newest DFS-related features in R2.
No duplicates, please
One of the more interesting new features in Storage Server R2 is SIS (Single Instance Storage). SIS automatically locates duplicate copies of files on volumes and replaces them with a pointer back to the original file. Duplicate files are identified based on file name, contents, and checksums to ensure all possible duplicates are found. The sparse links used as placeholders look just like the original files when viewed through Explorer, down to date stamps and file size. I like that even though the duplicate file is replaced with a pointer, file and folder permissions are not disrupted.
When enabled for a volume, a background task "crawls" the system, indexing and identifying each file. I did not notice any decrease in server response while the system was performing the initial inspection, and recognition seemed instantaneous as I created duplicate files and folders.
Admins can monitor the space savings obtained by SIS through the MMC-based Storage Management UI. Using SIS on its own servers, Microsoft was able to gain a total space savings of more than 14TB across 269 servers -- my dataset was much smaller, but I saw immediate savings nonetheless. Admins can also filter out specific file types, such as audio, that they don't want to be indexed by SIS.
One feature previously available in Windows Server 2003 R2 is now a core feature to Storage Server R2: Distributed File System (DFS) replication. Previous versions of Windows Server relied on the file replication service to move files between servers. In R2, DFS replication also replicates data stored on distributed servers, but adds differential compression to reduce traffic. This means only changes to the files are sent across the wire, not the whole dataset. Another nice feature is that admins can schedule when this replication takes place and the amount of bandwidth that the replication is allowed to consume.
There are two types of DFS replication in Storage Server R2: multipurpose and data collection. Multipurpose DFS replicates data between two or more servers in order to share content and publish files. Admins define the member servers, the various connections between members, and the folders to replicate.