Analysis of high performance cluster system technology based on IA architecture

Using the latest 4, 8-way IA server architecture of High-performance server cluster system, the use of the leading VI (Visual Interface) technology, effectively eliminates the communication bottleneck between nodes in the system, and the system uses load balancing technology so that user equipment can be fully utilized, and achieves 4 9 reliability, has the extremely high product price-performance ratio, provides a strong database server platform for the Internet application.

Cluster simply put, a cluster (cluster) is a group of computers that provide users with a set of network resources as a whole. These individual computer systems are the nodes (node) of the cluster. An ideal cluster is that the user never realizes the nodes at the bottom of the cluster system, in their view, the cluster is a system, not multiple computer systems. And the administrators of the cluster system can add and revise the nodes of the cluster system at will.

1. System Overview

The data center domain has been the world of high-end RISC server, for many years people only choose AS400, E10000, HP9000 and so on, the price is expensive, maintenance cost is high. While the IA architecture server is inexpensive, easy to use and maintain, supercomputers can be constructed via cluster technology, which can replace expensive medium mainframes and open up new directions for high-end applications in the industry.

For growing large users, the data center or data warehouse in the preservation of enterprise management, management data growth is very alarming, and the role of these data for large users is very important. The development of a number of years in the accumulation of data is a valuable wealth, through these Hao if the Smoke Sea data analysis, operators can draw intuitive business chart, curve, for the future development of large users to provide a strong decision support. But since such data has been expanding over time, it has put a lot of pressure on IT system managers for large users.

What kind of server do users need to meet current and future development needs?

First of all, we should have super computing ability, and can withstand the concurrent access of long time and large amount of users.

Second, the server system's high availability, easy to use management for users is also extremely important, if the system failure, resulting in service interruption, or the loss of important information, users will be caused to recover the loss. Therefore, the user must consider the highly available system solution when selecting the server system.

Third, with the continuous accumulation of data, query and statistics on the data will make the system become more and more slow, the replacement of hardware equipment can be said to be a large developing user indispensable work.

Using the latest 4, 8-way IA server architecture of High-performance server cluster system [1], the use of the leading VI (Visual Interface) technology, effectively eliminates the communication bottleneck between nodes in the system, and the system uses the load balancing technology so that user equipment can be fully utilized, and reached 4 9 reliability, with a very high product price, 1999 launched since the user's extensive concern for the vast number of users to provide a strong database server platform.

2. System principle

High-performance server cluster system is based on 2-node or 4-node cluster, up to 32 CPU, memory maximum support 32GB. The 4 node constitutes a unit of work that can be cascaded up to 16 units of work.

Each node is a IA server that supports parallel processing of the 4-or 8-channel Pentium III Xeon CPU. Each server is plugged in a gigabit NIC or VI structure of a high-speed switching card, connected to a high-speed switch (can be Gigabit Ethernet card switch or some special high-speed switches, such as the VI structure of the high-speed switch) as the server to each other data exchange, called the San (server area receptacle) switch.

Each server is plugged into a single gigabit or Gigabit Ethernet card, connected to a switch or hub on a local area network, providing connectivity services for client access.

Four servers share a single Fibre Channel disk array cabinet. Each server has two Fibre Channel cards, connected to two fibre channel hubs, each connected to the two controllers of a Fibre Channel disk array cabinet. For Fibre Channel disk array cabinets, the entire array cabinet works as long as one of the controllers works properly, so this configuration is redundant and prevents single point failure. For the most important data stored in a disk array cabinet, you can also ensure data security by storing the cluster system separately from the disk array cabinets, which allows for a maximum distance of 10 km between the cluster system and the disk array cabinet.

Each server has a hard drive area for installing the management portion of the native boot system and database system. User data is stored in a shared disk array cabinet.

In the LAN, there is a client as the management console, installed on its database management console, for the management of parallel databases, you can monitor the four nodes at the same time database instances, to achieve startup, stop, monitor performance functions.

In addition, the client has a network management system, a management console for the SAN, a management console for the disk cabinet, a UPS management console, etc. Some management functions are implemented with only the TCP/IP protocol installed, while others need to install the SNMP protocol to work properly.

A good cluster system, in addition to excellent performance metrics, also need to have the corresponding operating system and database support, our current cluster system support Windows NT 4.0 and Windows 2000 operating system [2], in the database support Oracle and DB2. It does not run on a single computer, only when multiple nodes work at the same time, can show its performance, and enable the system to truly achieve load balance (load Balance).

2.1 Two-node cluster system

From the configuration of the scheme, users can make different choices according to their needs, and can use two high-end servers to implement a virtual host. At this time, the use of the VI structure of high-speed switching equipment more advantages, do not need to use the VI switch, can be used on each server VI structure of high-speed switching card directly connected to achieve high-speed data exchange between servers. If you use Gigabit Ethernet cards, you need a gigabit switch, which is expensive.

2.2 Four node cluster system

The design goal of the cluster system scheme is to support the 4-node cluster, which consists of 4 nodes plus the corresponding storage devices and the switching equipment to form a work unit. Each unit of work consists of four independent 4-way or 8-way servers serving as a virtual fault-tolerant host, with four servers sharing a single Fibre Channel disk array cabinet. Each server has two Fibre Channel cards, connected to two fibre channel hubs, each connected to the two controllers of a Fibre Channel disk array cabinet. For Fibre Channel disk array cabinets, the entire array cabinet works as long as one of the controllers works properly, so this configuration is redundant and prevents single point failure.