The introduction of IBM's TotalStorage SAN File System renews some of the attention that has been placed on file...
systems the last several years and their central role to file management.
Architecturally, one could write a number of books, as well as Ph.D. dissertations (a number of folks in this business have), on the various ways file systems can be engineered to allow for open file management, file sharing and heterogeneous support regardless of file protocol (CIFS or NFS).
As you may already know, there are a few common problems facing file system management and NAS management. These have centered around:
- Where the file resides
- How the file can be accessed
- How many copies of the file are created
- How well the storage environment is utilized
- How well customers can implement common policies for file management across entire storage environments, instead of just one NAS or a single file system environment.
Certainly, not every product will solve all of these problems. IBM's TotalStorage File System is another product in a growing list of hardware and software products playing here, all of which use different technology architectures to do it.
Unfortunately, the significant number of ways to get to this nirvana of better file management can be complex and nearly impossible to explain without delving into technology elements.
Three types of file system management products
The products that focus on these series of problems today can still be classified into three basic categories:
- File management
- Distributed NAS platforms that incorporate advanced management functions
- SAN file systems that could be delivered as software-only platforms
Which vendors will you find in these three categories? First, file management includes the likes of DeepFile, which has built a policy-based management suite for files that improves the efficiency of management and enforcement of file quotas. Next, there are a number of vendors who provide distributed file systems as a high-performance enterprise NAS platform, such as Spinnaker Networks. Finally, SAN file systems take either pure software routes -- such as ADIC and Sistina -- or a hardware and software combination such as IBM's SAN File System.
Key areas to identify during your product evaluation
There are a handful of key technology nuances to consider when evaluating file system management products. (mileage varies depending on vendor):
- Global namespace: This allows clients to access files without knowing their location. Global namespace functionality also enables an administrator to aggregate file storage across heterogeneous, geographically distributed storage devices and to view and manage it as a single file system.
- Distributed Lock Manager: This is sometimes used to limit access to individual files with a mechanism that "locks" files. It's particularly useful when the file is "striped" across multiple storage subsystems so that the horsepower of multiple servers can be applied to simultaneous writes to a single file.
- Meta Data Servers: These servers centralize the file information in a common database that logs file information.
- Deployment: Some vendors provide a complete environment, both hardware and software, such as the distributed NAS platform from Spinnaker Networks and the SAN file system from IBM. Others provide just software that gets loaded onto hosts connected to the SAN environment, such as ADIC and Sistina.
- Performance: Customers need to determine their performance requirements before making the plunge on a purchase here, since vendors' approaches can vary greatly. The best method to measure performance is to examine the SPECfs performance benchmark, which gauges speed and request handling for NFS file systems. (This benchmark allows you to aggregate file system performance for reads and writes to disk across a spectrum of file sizes.) There are other more narrow tests, such as single file read/write, but these must be used cautiously since they are applicable to only a very narrow application profile.
- Availability: How will the file system handle host and storage system availability issues, and how does it handle file system journaling?
- Management: How are policies set on the files, how is block-level file access handled, can priorities be set on file and network bandwidth?
- Heterogeneous: Does this run over just the vendor's hardware, or is it heterogeneous across multiple storage systems and operating environments?
Let me make a few last comments: Determine the key problems you are trying to solve. Then, evaluate vendors based on their ability to meet your requirements. Make sure you know the limitations of each approach before making the investment.
About the author: Jamie Gruener is a senior analyst covering the storage and server markets for the Yankee Group's Enterprise Computing and Networking Planning Service. He is also a SearchStorage.com expert and contributor in the area of Storage Management: Trends & Buying Criteria. If you have a question for Gruener, ask him in our "Ask the Experts" area. Jamie will also be commenting and answering questions on IBM's TotalStorage SAN File System in a webcast on Thursday, Oct. 16, at 12pm EDT.