Now, when in the market there is a fault tolerance equipment and fault tolerance operating systems - all attention should be directed on development of fault tolerance server applications.

Usually server applications which should provide continuous services of a hosting for the users, are created with a stress on fault tolerance. It is a question of such applications, as databases, systems of transactions processing, control systems of critical processes, the systems providing the interface with an external world (web, e-mail, ftp).

• Servers Internet are engaged in processing of transactions, work with databases, are "card" and, hence, are critical for reputation of their owner.
• FTP, telnet, etc. - these "old", leaning virtual connections (one connection exists in time as much, how many and a session) services and reports cannot be made failure-safe. Their dependence on through connection does not allow to carry out smooth transition from the process worked on a server before failure, to the process generated on a server after restoration. In most cases connection will be lost by the client, and the user should begin a session all over again.
• E-mail - though the most widespread protocol of e-mail SMTP (the simple protocol of mail transfer) also is focused on virtual connection, interaction with the end user is absent. Therefore the broken off connections can be restored without participation of the user. Protocols of acknowledgement of delivery help to support additional reliability of through data transmission.
• A Web-server - protocol HTTP does not lean on virtual connection. To restore the lost communications it is possible. However, if transfer of on-line transactions on Web interrupts, they should be restored by means of the methods applied in systems of transactions processing for databases.
• Network File Services (NFS) - NFS-most often applied network file system, does not lean on connections and, hence, is tolerant to failures in work. However, if for acceleration it is applied cashed data in a cache-memory will be lost. For architecture with the divided memory there are decisions of this problem (through record in a cache-memory).

First of all, it is the people, aspiring to use Internet for the business purposes, their number very quickly grows. Reliability of data transmission from the end in the end is especially important, when monetary transactions take place. Imagine a situation when someone does the order through Internet with on-line payments from its account on the account of the company receiving the order. If server Internet refuses during that critical moment when the first transaction is not completed yet, and the second already is completely processed, the buyer should pay that it and has not ordered. Use of fault tolerance technologies will lower similar risk practically to zero and, thus, will raise trust of people to orders through Internet.

The similar script is developed owing to development of commercial applications of small capacity. Users any more do not buy software on carriers, and if necessary download them for a payment from Internet. After use such applications automatically leave. Payment is made for each use of software product, instead of for each copy. Hence, users are very interested in that process of swapping did not interrupt and did not contain mistakes.

One more example: the number of hosting service providers a grows almost also quickly, as well as number of consumers. Hence, the nearest years they should collide with growth of a competition. Simple reduction of prices is not very favourable, as will lead to decrease in incomes. Therefore it is much more favourable to improve quality of services. Ability to offer uninterrupted and fault tolerance services without increase of service charges - convincing enough argument for perspective clients.

We have resulted only some typical examples of use of high availability services. Actually it is much more than them, and any we yet only do not know, but even we can not imagine.

We have considered some aspects of fault tolerance in the field of a hosting granting services. By and large, there are three levels, each of which should possess properties of fault tolerance that all system was fault tolerance: the hardware, operating system, the the server software and applications.

The hardware can be made fault tolerance in several ways, each of which, anyhow, means redundancy. Operating systems of type UNIX are under construction by a modular principle so they can be improved by addition of failure-safe elements at the levels close to the equipment, thus practically it is not necessary to touch the top layers. Software of Internet servers can be considered as the usual application and, hence, can be made fault tolerance by similar methods. The services which have been not based on virtual connections to modify much easier.

Today in the market already there is a need for high availability hosting services, and will appear eventually more increasing and more products for its satisfaction. Interaction and cooperation between the companies for a long time working in the field of fault tolerance computer facilities, and the companies, engaged networks and Internet, will help to develop and advance more quickly on the market such products.