Section 3.1. The Emergence of NAS

   

---

3.1 The Emergence of NAS

After the appearance of direct-attached storage, several factors resulted in a situation in which multiple servers with direct-attached storage were deployed. The storage was accessed via a network file system, typically NFS, at least in the mid-1980s. One relevant factor was that in the 1980s, the client/server era had begun, following the eras of the mainframe and the minis. Client/server computing was widely accepted in commercial enterprises .

In addition, storage was still relatively expensive, yet storage needs were growing. The bottleneck in deploying more storage was the fact that a server could have only a limited number of direct-attached storage devices (typically only seven or eight addresses were allowed, and the host took one address, before advances in SCSI allowed more). This constraint meant that more storage could easily be deployed only if more servers were deployed. Finally, the bottleneck with I/O operations was (correctly) perceived to be the storage I/O bus. Deploying more servers meant reducing the load on each server bus.

Vendors took advantage of the prime conditions and started pushing the concept of network-attached storage, or NAS. As Tom Clark points out in his book Designing Storage Area Networks , NAS is more a marketing term than a technical term. NAS devices were touted as easy to use, easy to manage, and easy to deploy. NAS devices were also described as specialized for optimum storage device I/O throughput. Although in some cases this was true ”that is, the NAS operating system had really been honed down to the bare minimum required ”in other cases NAS devices were simply thinly disguised general-purpose servers. Even the term NAS device has implications of a specialized device and not a general-purpose server that happened to have some large amounts of storage attached to it. ^[1]

^[1] Note that the storage local to a NAS device is typically of the direct-attached storage type, discussed in Chapter 2.

A NAS device consists of some server software that runs with the aid of an operating system. Further, as shown in Figure 3.1, the NAS software stack builds on the normal server file system plus I/O system stack by adding the network server code, network protocol stack, and network file system stack, which provides elements of distributed file systems such as caching and coherence . These three additions are shown shaded in Figure 3.1.

Figure 3.1. NAS Device I/O Stack

In general, the network-related portions of a NAS device are shown on the left-hand side of Figure 3.1, whereas the local file system and storage- related stack are shown on the right-hand side. In the interest of simplicity, only a TCP/IP network stack is shown and discussed here. In reality, several different network protocols, such as UDP/IP or Netware IPX/SPX, may be used.

In Figure 3.1 the NAS server software posts a TCP/IP listen request, waiting for a client to send it a request. When a client sends a request, the listen completes and a TCP/IP session is started. Once the TCP/IP session is established, the client can authenticate and then send requests to open files and read or write files via the SMB/CIFS or NFS network file system protocols (described later in this chapter). Once the NAS server software receives a file I/O request, the NAS server uses the services of the local file system to accomplish the I/O operation. The results of the I/O operation (read data or write status) are sent back to the client via the network file system and network device protocol stack.

NAS vendors have adopted various different strategies for developing the operating system and file system needed for the NAS devices:

• Using a standard operating system such as Windows NT or UNIX

• Developing their own operating system and file system ”for example, Network Appliance

• Buying an operating system and file system from another vendor

---

Top