Chapter16.New Directions in Network Filing: Clustered File Systems, Distributed File Systems, and Network Storage for Databases

Chapter 16. New Directions in Network Filing: Clustered File Systems, Distributed File Systems, and Network Storage for Databases

Upon completing this chapter, you will be able to

• Describe the architectures of data access in clustered systems

• Participate in product evaluations for distributed file system products

• Discuss the implementation of storage used with database systems

Traditional file systems are a single point of failure for data access. If a system becomes unavailable, access to the data managed by file systems that were running in it is blocked. Although systems can be built with redundant components and high-availability architectures, they can still failand their file systems along with them.

Obviously, there is a need to create redundancy for the file system itself. One way to do that is through the use of clustering technology. While clustering is often thought of as an advanced topic in systems, one can make a convincing argument that the primary purpose of clustering is to provide redundant file system access. A cluster cannot work properly if the systems in it do not have access to all the data.

Another shortcoming of traditional file system implementations is scalability. Modern file systems running on single systems can be very large, but past a certain point, their performance suffers. Spreading the file system work over multiple systems gives file systems the parallelism needed to grow in size without losing performance.

Because farm and grid computing are more commonly deployed, it only makes sense to think about the server architectures that provide the best match for them. If tens or hundreds or even thousands of distributed systems are working together in a single application environment, what is the optimal file system architecture? Not a single point of failure with capacity/performance limitations. It's clear that a different type of network filing system is the answer.

This chapter looks at ways to distribute the file system function in system clusters and distributed file system technologies. The chapter closes with a discussion of storage for database systems, first looking at traditional database storage architectures and then looking at storage designs for clustered and grid database architectures.