Chapter 11. Disk Subsystems

As mathematician and pioneering computer scientist John Von Neumann conceptualized it, a computer's architecture is built like a human brain. A server's processors correspond to intelligence or, more precisely, higher logic functions; your RAM and cache perform a function similar to short-term memory; and permanent storage such as disks functions similarly to long-term memory. We don't yet have computers that are massively parallel processed or compute with a fault-tolerant fabric, but one day, in the not-too-distant future, we most certainly will.

Access to massive amounts of storage is a feature of modern computers; and that is true to an even greater degree with servers. What once fit into a very large room now fits on your desktop. Shakespeare's life works is a 1MB text file, and Microsoft's Encarta encyclopedia, which would occupy several linear feet in printed book form, is now 5GB of data, images and multimedia components included. No wonder that EMC, the world's largest maker of enterprise-class shared storage servers has as its slogan "Where Information Lives." Here's a truly remarkable fact: Every year or two, information stored in digital form doubles the amount of all recorded information from the dawn of man on. Because of this exponential growth, a server is called upon to perform more and more tasks that require access to large amounts of storage. Therefore, it's really important for you to have a fundamental appreciation of what the different types of storage can offer and how to manipulate storage to suit your needs. This chapter focuses on the building blocks of storage systems: the different type of drives, connections (cables, backplanes), and host adaptors.

Servers are required to be much more reliable than workstations. Another way to describe reliability is the term availability, which is a statistic that can be quantified. If you are in a hospital and your computer provides a mission-critical service, you might require that it be down only five minutes a year (roughly two reboots). That level of fault tolerance translates into a 99.999% availability, also referred to as "five nines." For a departmental server, a more easily achieved level of availability might allow for nine hours of downtime over the course of a year; that availability would be 99.9%, or three nines.

Note

Because the focus of this chapter is on deployment of drives in server environments, it is assumed that you have some familiarity with basic disk drive technology. You should be somewhat familiar with terms such as platters, cylinders, heads, sectors, format, partitions, and addressing modes before you read this chapter. You should also have a sense for the metrics of disk performance: seek time, latency, transfer rates, and access times. For more information on these subjects, pick up a copy of Upgrading and Repairing PCs.

Because availability translates into fault tolerance, a central focus of server disk subsystems is data redundancy. The second half of this chapter examines redundant array of inexpensive disks (RAID). Servers use different types of RAID controllers and different configurations of drive arrays to create fault tolerance. This chapter looks at the definitions of the different types of RAID, as well as the advantages and disadvantages of each RAID level. From an operational standpoint, this chapter also describes procedures for adding drives to arrays, recovering from drive failures, doing maintenance on arrays, and protecting your data on arrays.

This chapter describes nonvolatile storage that is directly attached to a server, what is referred to in the storage industry as direct attached storage (DAS). In addition to disk-based technologies, you will find a variety of other useful nonvolatile storage technologies that are often directly attached to a server, including tape backups, libraries, and optical disk writers and readers (both singly and in the form of optical disk carousels). Because large libraries and carousels are mostly used for backup, hierarchical storage systems (HSS), and near online storage systems, we don't discuss them further in this chapter.

This chapter sets the stage for Chapter 12, "Storage Area Networks," where the focus is on shared storage solutions.