Why US:

---

You can afford it. We can build for you the system with fault probability 0.9999 of two systems with fault probability 0.99 that will increase the fail-safety of your solutions in degrees. Our solution allows to create practically computer RAID massive of the inexpensive/ independent computers. We can build either solution for big business on the SUN, IBM servers base, or for small-scale business in Aquarius computers base. Program platform can be Windows as well as Unix/Linux.

Cluster - is two or more indeoendent systems joined into one unifies system of high availability level with a special software and hardware. There are four main benefits of using cluster systems:

• High availability
• Scalability
• Flexibility
• Management simplicity

Cluster - is also an opportunity to use the calculating resources of your system in such a way that the obtained system exceeds the facilities of its’ parts.

Cluster systems of high availability

Reliability of corporate data processing systems, their capability of incessant operative and proved information supply is one of the most important conditions of effective work and, as a result, competitiveness of modern companies. At the moment there are a lot of technical solutions ensuring the necessary level of reliability and fail-safety of information systems. One of the corner stones of such solutions is clustering of server systems, which supports the high availability. Saying about cluster systems of high availability it is necessary to realize properly that the fundamental element of such systems deployment at the enterprise is far from hardware and software used there but the set of competent expertise and consulting services, assignable either by the clusters supplier or by his partner-integrator aimed for adaptation of the typical hardware-software solutions available for the real requires and conditions of the customer. Building of the effective and qualitative cluster system of high availability is rather complicated problem solution of which requires deep knowledge and experience although the whole process is guided by using standard software and hardware facilities. To a great extent, every further realization of cluster system of high availability should be considered exclusive and in a certain sense unique problem, solution of which does not stand light-minded attitude neither from the supplier’s side nor from the customer’s.

Another important question often emerging when building cluster systems of high availability is adequacy of hardware in use of solutions solving with their help. Building of fail-safe clusters supposes using as central machines trusted servers with double and sometimes even multiple duplication of all the blocks and components. The opposite approach brought from the calculating cluster systems sphere and propagated by some suppliers is that the cluster architecture itself is acknowledged the only necessary and sufficient condition for securing the high availability of data processing systems and hardware base is given to cheap, lightly changeable and unreliable components. This approach is not only wrong but even dangerous as it gives customers just an illusion of high availability and reliability of their data processing systems.

And finally, the last digression: before immediate examining of the typical solutions for cluster systems of high availability building from three leading suppliers in this area – SUN, IBM and HP, it will be useful to remind in outline of the theory and technology sides of clustering. Typical scheme of cluster system of high availability depicted in the picture implies union of two or more (depending on the certain solving problem) servers into integral system, seen outside as a single server executing this or that corporate application, e.g., database or ERP-system.

For outside communication and data transferring from the consumers and towards them in most cases the standard connection Fast Ethernet or more and more extending standard Gigabit Ethernet, rarely ATM are used. Servers communicate among each other with so called interconnect – special fixed network for synchronization of states and sending job reset commands in the standby node in the case of failures or denials in the central one. Variants range for interconnect organization is quite large – from the well-known and widespread Myrinet, InfiniBand or Scalable Coherent Interface (SCI) to highly tailored firm developments – so suppliers may apply any of them at their discretion provided that it would have enough speed. It is important to mention here that unlike the calculating clusters, where a big volume of information is being transferred among the nodes, and accordingly high data throughput is necessary, in the clusters of high availability there is practically no traffic exchange among the nodes excluding the housekeeping commands, however the commands themselves must be transferred quickly to the limit, that is why here the standards with the least delay period are used.

Keeping systems interaction, as well as in the case with the highly productive servers, is organized in the most flexible way: it’s possible to use any schemes and logs – from the direct connection to NAS and SAN solutions, connected with optical fibre. Customer’s needs and real conditions of his system play the main role here, starting from which a certain decision is made. Naturally, in all these cases network interfaces fore more reliability are duplicated as minimum.

In the end, one more significant point is worth a bit more detailed examination. Real cluster systems of high reliability are built in one of two basic architectures – with share something or with share nothing. In the first case (bright example - Oracle Real Application Clusters) due to internal logic of intermediate software, customer’s task is “smeared” among servers of the cluster (that is where a real profit of several servers including is seen) and each server executes some part of it. Such an approach allows to increase the efficiency of funds investment, combining successfully the high fail-safety and productivity on relatively not expensive hardware means. In the second case customer’s problem is completely solved in the central system node while the second server plays the role of the hot reserve able to catch in any moment the “fallen” application. What variant of cluster system organization to choose depends on the company’s IT-infrastructure and is defined in every certain case starting from the customer’s tasks and software environment for their solving.


Order our services by filling in this form.