Jul 07 2008
Data Center

FAN Club

State and local agencies look to file area networks to ease storage management.

State and local agencies look to file area networks to ease storage management.

Anyone managing servers or storage knows all about file-system sprawl. Because each network-attached storage box (NAS) and file server contains its own file system, file management and data movement can become a real hassle — especially when you get beyond a handful of boxes.

“NAS and file servers are too rigid, as you have to spend time breaking and rebuilding all your connections and configurations whenever you have to move data,” says Paul Lubold, infrastructure and operations manager at the Pennsylvania Office of Attorney General in Harrisburg. “With a file area network, I can move data transparently in seconds — no more coming in weekends to avoid user disruption.”

A file area network (FAN), which is also loosely termed file virtualization, consists of a collection of network-attached storage appliances and file servers that are virtualized to enable the data on them to be managed under a single file system (see sidebar).

“As NAS appliances proliferate, the management of them becomes more complex,” says Deni Connor, an analyst at Storage Strategies Now in Austin, Texas. “Rather than managing each NAS device by itself, combining them into a FAN allows everything to be administered collectively.”

The Pennsylvania Attorney General’s office relies on a main data center in Harrisburg and 22 smaller remote sites to serve about 1,000 employees statewide. In the past, when an employee moved from one office to another, it caused a headache for IT as user files had to be taken off one server, transferred manually over the WAN to the new location and reassigned to a specific storage location. When servers ran out of room, IT had to physically move user storage onto a larger box during off-peak hours in order to avoid interference with attorney operations.

The change to an IT architecture was driven by stricter data-retention requirements and the mushrooming demands of electronic discovery. IT added two Network Appliance FAS3050 NAS boxes at its headquarters to store about 16 terabytes of data, and a Network Appliance FAS3020 NAS box at a disaster-recovery site. As part of this, it set up a global namespace using Brocade StorageX software.

“We had already implemented VMware to virtualize our servers and saw we could do the same with storage,” says Lubold. “Once we get it fully deployed, there will be no such thing as a maintenance window anymore — at least for file management.”

His IT group plans to pilot a FAN before deploying it throughout the state. The architecture consists of Microsoft Distributed File System (DFS) software, which maps logical physical devices to logical storage and replicates data across WAN links. StorageX software extends the functionality of DFS by adding global namespace capabilities and simplifying data management at remote locations. The network operations center will have DFS and StorageX running on an existing domain controller, while smaller offices will require only client software running on desktops.

“The main thing we have to watch for is network performance,” says Lubold. “We have plenty of bandwidth, but we don’t know how the network will react to the load.”

FAN Value

Those implementing FANs report high value. According to analyst firm IDC, file virtualization allows organizations to reduce storage costs by 50 percent to 80 percent, and improve management efficiency by up to 90 percent.

“Any state or local government agency intent on organizing their file data should be interested in FANs,” says Connor of Storage Strategies Now. In addition to making it easier to move data around, FANs help to balance workloads. Instead of one box sitting idle while another is overloaded, file virtualization makes it a simple matter to spread loads across physical machines. This is a good way to avoid server sprawl due to exploding storage demands.

The city of Encinitas in Southern California, for example, has about 18 servers and, until recently, had been adding more as needs dictated. Soaring storage demands were addressed by buying servers with bigger hard drives than were actually required, just to offset anticipated expansion.

“We have a number of individual servers that we believe would be better utilized if we consolidate their storage capacity via file virtualization,” says Rainer Mueller, IT analyst for Encinitas. “File virtualization would help us to dynamically allocate storage and prevent underutilization of storage capacity.”

The one negative he raises, however, is cost. One bid for FAN implementation came to $180,000 — more money than the city had previously spent on all its servers. Mueller thinks that amount will drop significantly before the dust settles. Before too long, he believes file virtualization will be used to consolidate server storage and make it more manageable.

“That’s probably the direction we’ll be heading in the near future,” says Mueller. “But right now we are at the early stages — we’re talking to vendors and speaking with consultants before any final decisions are made.”

Mueller’s numbers tie in with those cited by Rick Gillett, vice president of data systems architecture at F5. F5 Acopia ARX appliances are generally clustered; pricing for two boxes starts at just under $50,000. For large-scale deployments encompassing multiple locations, the price tag can exceed the million-dollar mark.

One large city is adopting F5 Acopia’s FAN on a trial basis, according to the manufacturer. Unlike the Pennsylvania Attorney General’s office, it is focusing on internal file-system management and storage tiering within one data center. By using a FAN to simplify the process of separating active from inactive data, it significantly reduces the amount of data to back up. In the future, the city is looking at adding global namespace in order to share files among different offices and as replication for disaster recovery purposes.

Correct Sequence

Lubold from the Pennsylvania Attorney General’s office advises server and storage administrators to use file virtualization to save themselves time and avoid laborious storage administration. Instead of taking down an aging file server, moving the files over to another box and going through the process of re-establishing the data paths to the various users, he says it’s easier to implement a FAN.

“Once you create your global namespace and virtual environment, you don’t have to move anything manually again — and that saves you an awful lot of time,” says Lubold. “Now that we have it set up, we will be able to harness it as part of our information lifecycle management strategy to move older or infrequently used files to near-line storage.”

FAN Explained

A file area network (FAN) consists of a collection of network-attached storage appliances and file servers that are virtualized to enable the data on them to be managed under a single file system. F5, maker of the F5 Acopia ARX devices, describes a FAN as a way to improve management of unstructured data by decoupling that data from specific servers or NAS filers to offer services such as data migration, storage tiering and replication.

Similar to server virtualization, a FAN adds a virtualization layer that sits between users and the storage devices to simplify file management and access. Vendors such as F5, Brocade, EMC and Microsoft implement this either through an appliance or through software. While each implementation has its own approach, there are two basic elements: decoupling and FAN services.

Decoupling is generally accomplished using what is known as a global namespace, which is a layer that sits on top of the physical equipment it is stored on. FAN services include such things as WAN optimization, file reorganization and consolidation, information lifecycle management, disaster recovery and replication.

Virtualizing Disaster Recovery

A FAN is useful in making files available elsewhere for disaster recovery, which is why New York City’s Department of Consumer Affairs deployed Network Appliance Virtual File Manager.

Its main office in Manhattan has about 15 servers that store data on an EMC Celerra NS500 NAS box, which holds about 6 terabytes of data. The organization’s other major site in Queens has a NetApp FAS270, as well as two servers. Users at Queens also write to the EMC box.

Everything in the Manhattan office is replicated to Queens via Microsoft Distributed File System. Network Appliance Virtual File Manager (VFM) aggregates the distributed file data across this mixed-vendor environment into one logical file system.

“Everything is replicated to Queens for [disaster-recovery] purposes,” says Matthew Miller, LAN administrator for the Department of Consumer Affairs. “If Manhattan goes down, users are directed to Queens automatically.”

Miller was so confident in this system that he conducted the ultimate test — he unplugged the EMC box right in the middle of the workday. In this scenario, VFM and DFS work closely together: VFM directs the saving process and invokes DFS for replication.

“When we tested it, everything worked beautifully,” says Miller. “The system failed-over to the secondary location and users didn’t notice a thing.”

Close

Become an Insider

Unlock white papers, personalized recommendations and other premium content for an in-depth look at evolving IT