Free tutor on Basic principle NETWORK FILE SYSTEMS AND FILE SERVERS and Network Attached Storage (NAS)

Free tutor on Basic principle NETWORK FILE SYSTEMS AND FILE SERVERS Network Attached Storage (NAS)

Network file systems are the natural extension of local file systems. End users and applications can access directories and files that are physically located on a different computer – the file server – over a network file system (Section 4.2.1). File servers are so important in modern IT environments that preconfigured file servers, called Network Attached Storage (NAS), have emerged as a separate product category We highlight the performance bottlenecks of file servers and discuss the possibilities for the acceleration of network file systems. Finally, we introduce the Direct Access File System (DAFS), a new network file system that relies upon RDMA and VI instead of TCP/IP.

Basic principle

The metaphor of directories and files for the management of data is so easy to understand that it was for a long time the prevailing model for the access of data over networks. So-called network file systems give end users and applications access to data stored on a different computer (Figure 4.5).

The first widespread network file system was the Network File System (NFS) developed by Sun Microsystems, which is now the standard network file system on all Unix systems. Microsoft developed its own network file system – the Common Internet File System (CIFS) – for its Windows operating system and this is incompatible with NFS. Today, various software solutions exist that permit the exchange of data between Unix and Windows over a network file system.

With the aid of network file systems, end users and applications can work on a common data set from various computers. In order to do this on Unix computers the system administrator must link a file system exported from an NFS server into the local directory structure using the mount command. On Windows computers, any end user can do this himself using the Map Network Drive command. Then, both in Unix and in Windows, the fact that data is being accessed from a network file system, rather than a local file system, is completely hidden apart from performance differences. Long before theWorldWideWeb, the File Transfer Protocol (FTP) provided amechanism by means of which users could exchange files over the Internet. Even today, FTP servers remain an important means of distributing freely available software and freely available documents. Unlike network file systems, access to FTP servers is clearly visible to the end user. Users require a special FTP client with which they can copy back and forwards between the FTP server and their local computer. The Hyper Text Markup Language (HTML) and the Hyper Text Transfer Protocol (HTTP) radically changed the usage model of the Internet. In contrast to FTP, the data on the Internet is linked together by means of HTML documents. The user on the Internet no longer accesses individual files, instead he 'surfs' the World Wide Web (WWW). He views HTML documents on his browser that are sometimes statically available on a HTTP server in the form of files or today are increasingly dynamically

generated. Currently, graphic HTTP clients – the browsers – without exception have an integrated FTP client, with which they can easily 'download' files.

Network Attached Storage (NAS)

File servers are so important in current IT environments that they have developed into an independent product group in recent years. Network Attached Storage (NAS) is the name for preconfigured file servers. They consist of one or more internal servers, preconfigured disk capacity and usually a stripped-down or special operating system (Figure 4.6). NAS servers are usually connected via Ethernet to the LAN, where they provide their disk space as file servers. Web servers represent a further important field of application for NAS servers. By definition, the clients are located at the other end of the WAN so there is no alternative to communication over IP. Large NAS servers offer additional functions such as snapshots, remote mirroring and back-up over Fibre Channel SAN.

NAS servers were specially developed for file sharing. This has two advantages: since, by definition, the purpose of NAS servers is known, NAS operating systems can be significantly better optimized than generic operating systems. This means that NAS servers can operate more quickly than file servers on comparable hardware that are based upon a generic operating system. The second advantage of NAS is that NAS servers provide Plug&Play file systems,

i.e. connect – power up – use. In contrast to a generic operating system all functions can be removed that are not necessary for the file serving. NAS storage can therefore excel due to low installation and maintenance costs, which takes the pressure off system administrators. NAS servers are very scalable. For example the system administrator can attach a dedicated NAS server for every project or for every department. In this manner it is simple to expand large websites. E-ail file system full? No problem, I simply provide another NAS server for the next 10,000 users in my Ethernet. However, this approach can become a management nightmare if the storage requirement is very large, thus dozens of NAS servers are required. One disadvantage of NAS servers is the unclear upgrade path. For example, the internal server cannot simply be replaced by a more powerful server because this goes against the principle of the preconfigured file server. The upgrade options available in this situation are those offered by the manufacturer of the NAS server in question. Performance bottlenecks for more I/O-intensive applications such as databases, back-up, batch processes or multimedia applications represent a further important disadvantage of NAS servers. These are described in the following subsection.