An NFS Tutorial – Part 1: The Metadata Challenge By Andrew Zillman On March 21, 2011

Organizations struggle daily with the challenge of growing amounts of data and applications, which are placing a heavy burden on their existing storage infrastructures. Storage admins constantly fight to keep users happy with quick and efficient application and data delivery. Yet behind the curtain in many NAS environments, the fundamental read and write operations constitute a relatively small percentage of the NFS (Network File System) activity between clients and storage. Instead, NFS metadata makes up most of the mix, running rampant, often times stored on disparate servers across the organization or around the globe, consuming a great deal of system resources and negatively impacting the user experience. So, what exactly is this NFS metadata? And more importantly, what does it mean to you and how can you manage it rather than letting it manage you?

Metadata can be best described as data about data. When a user requests a file or browses a directory, many operations occur that are transparent to the user. These operations give NFS information about the file, such as: “When was the last recorded time an element of this file was modified?”, “What is the file handle, or identifier for a given file?”, “Provide me the pathname stored in a symbolic link.”. You might think of these in terms of “What’s my address?”, “Who do I belong to?”, “Who can access me?”, “When was I last accessed, by whom and how?”, etc. This asking and responding can be an endless, time-consuming process. Now imagine this scenario being played out for each and every file accessed by each and every user on the network. Literally hundreds of thousands of NFS metadata operations per second can be executed. The NAS can become so overwhelmed in processing metadata requests that its ability to provide its most critical function – serving payload data to users – is jeopardized. The associated symptoms include oversaturation of storage resources and less-than-pleasant user experiences.

Let’s now consider the significance of metadata and the NFSv3 RFC protocol which has paved the way for countless new technologies to be built upon its ability to efficiently move data. NFSv3 makes extensive use of NFS metadata. Despite its great simplicity and performance, admins need to be concerned about the effect NFS metadata can have on NAS performance scalability. While NFSv4 was designed in part to help mitigate the NFS metadata processing burden, its adoption remains relatively small. Customers are not finding a compelling reason to move to NFSv4, especially in the Local Area. NFSv3 continues to be used extensively and will be for a number of years, causing many enterprises to look for simple solutions that reduce the metadata processing burden, allow for far greater scalability and higher performance, all without migrating to new hardware or protocols.

One such solution happens to be Alacritech’s ANX 1500. Among its many advantages, the ANX is remarkable at handling these pesky bits of metadata, essentially acting as a vacuum. With the ANX 1500 scrubbing metadata, the existing storage infrastructure is better able to do what it was designed to do: serve data.

Next on tap: NFS READs, WRITEs and why caching matters.