| With the computer and network spread quickly, more and more traditional applications are shifted to computer and network and this brings much convenience in the work and life. At the other hand, there is much thing to threat the security of data in the network such as virus, hacker, hardware fault and human or nature disaster. So we must create perfect system to protect the security and high-availability of data in the network. Availability is defined as whatever is happened, the system can be online and accessed. There are many facts that can make the system shutdown, including the shutdown in plan and unexpected faulty and so on. The target of high-availability is to make the down time and recover time least and to describe clearly scheme's maturity, complex and cost.Recently high-availability gets more attention from customer. And which tech will be used is a hot issue. By now the main technologies include clustering, firewall, intrusion detection, network anti-virus, data backup, UPS, and disaster tolerance in different area. In them, the type of cluster will affect the other technologies. There are many different method and technologies about clustering. Just from the infrastructure there are three modes including standby model with double nodes (master and slaver), ready model with double nodes (equal model) and multi-machines model.Master and slaver model: When a node is working as master server, the other is working as slaver server for backup. The backup server is monitoring master server. When the master is down, such as machine fault, network faulty and software fault, the slaver will take over the services of the master in order to make the whole system work continuously.In-ready model: the two nodes backup for each other and provide different services. When one node is down, the other will take over the services provided by the dead node. Then the normal node will provide all the services. Multi-node model: As the simplest model there are three nodes in the system. Two nodes can compose as standby or in-ready model. The third node will monitor the two nodes. When one of two nodes gets down, the third will take it over.The advantage of master-slaver model is simplifying the implement. And the shortage is wasting a haft of the resource. The both-in-ready model can take full advantage of the resource, but the load of normal server will grow up and affect the performance of the server. Multi-node can avoid the entire shortcoming, but it would lead the system more complex and increase the number of machine.At the other hand, there are two way to backup data. One is share the data through SCSI disk array. The other way is to make mirror disk by software. The first is depended on hardware deeply and function is least and design couldn't be flexible and need more cost and has limit of distance. So it is as the transitional technology. The last has more advantage than the first. Little cost, simple implement, good scalability and the limit broken in distance fulfill the customer's need of data security and disaster-tolerance in distance.Considering all fact and what we need actually, we design our system with master-slaver model and software disk-mirror. Linxu8.0 is the implement platform. There are four parts in the system.Central management: the main job of it is to initialize the system, and reset armaments including the IP address of two net interface cards and the other node's IP addresses and the role it will play. At the same time it invokes the other parts of the system. So it is the center of the system, And it select according role in the system Monitor model: the main work of the model is divided into four functions: First, it sends its own status to the other in a interval time. Second, it can receive the status report from the other node. Third, it checks the status of the key services in a interval time and will change the role of itself when some services are in exception. Forth, it checks the status of the lines and will deal with the exception. The lines involved in the...
|
PDF Full Text Request