How do I implement a server cluster? What is a server cluster?

It's not difficult, the hardware uses a router, the software, and the operating system uses win2003

server

enterprise

Enterprise version, it is recommended to install the R2 upgrade package together, all machine group LAN, use a gigabit network card for domain control, set up a stream ** server, other machines as domain members to join in, intranet IP with their own, external network with port mapping to an IP, use domain control to do network traffic load balancing, domain control machine configuration to be strong, if you have a large network traffic, it is recommended to use a professional-grade server, Xeon + 2GB + SCSI hard disk and the like, depending on your environmental requirements, if necessary, you can go to double Xeon, Then use a P4 with 512MB RAM

The above machine is used for backup domain control, so that the main domain control is online and offline or restarted or malfunctioned, and the normal work of the members in the domain will not be affected, and the backup domain control can be improvised, according to my above requirements, of course, there is money to use well.

If you have high security requirements, it is recommended to use an ordinary P4+512MB memory machine rack for the front-end of the routerISA2004

The server group firewall has a good effect than the general hardware firewall, which does not affect the operation of the network environment at all, and the members of the domain can run naked without fear of poison and black.

As for the member machines in the domain, if you only go all out to ** the film source, the current mainstream home models are enough.

HP is recommended for the server

360G series, the current price is not high, the cost performance is not bad, the after-sales service is very good, if you don't know much about network building, you can let them help you install it, buy their server is to use their human resources.

The router can be selected with Flying Fish 4200 or above models, and the dual WAN ports of telecom and network communication are dedicated to large Internet cafes that can provide 150 250 machines, and the built-in parameters are very rich.

In addition, a few more wordy words, don't buy the board card, I was fooled, the rainbow itself is a consulting company, there is no board production capacity, are the same German OEM, thought it shipped a lot, it was selected, the result of the advertising on the index parameters and the actual product is fundamentally different, too much water, after-sales is also very bad, hereby recommend ......

Downstairs, don't plagiarize me anymore, I'm copied every day with several 200 points best, it's really depressing!

Server Clusters:

A server cluster is a collection of servers that work together to provide the same service, as if there was only one server for the client. Clusters can use multiple computers for parallel computing to achieve high computing speeds, and can also use multiple computers for backup, so that the whole system can still run normally if any machine is broken.

Server Load Balancing:

Load balancing. （load

balancing）

Building on the existing network structure, it provides a cheap, effective, and transparent way to expand the bandwidth of network devices and servers, increase throughput, enhance network data processing capabilities, and improve network flexibility and availability.

Distributed Servers:

The so-called distributed resource sharing server refers to a form of theoretical computing model server in which data and programs can be distributed to multiple servers instead of being located on one server, and the distributed geographic information data on the network and the database operations affected by it are the research objects. Distributed is conducive to the distribution and optimization of tasks on the entire computer system, which overcomes the shortcomings of the traditional centralized system that will lead to the shortage of central host resources and response bottlenecks, and solves the problem of network GIS

The problems of data heterogeneity, data sharing, and complex operation are a major progress in GIS technology.

These three architectures are common server architectures, and the cluster is mainly done by IT companies, which can ensure the security of important data; Load balancing is mainly used to share the amount of access and avoid temporary network congestion, which is mainly used for e-commerce type. Distributed servers are mainly used to solve the problem of cross-regional, multiple single nodes to achieve high-speed access, and generally use distributed servers for the purpose of CDN.

Pure hand-typed, I hope it can help you!

1) Strong scalability.

Other scaling technologies usually only support the expansion of 10 CPUs, and the scaling capacity is limited. The cluster system using cluster technology can be extended to multiple service spikes including hundreds or thousands of CPUs, and the expansion capability has obvious advantages. Cluster services can also be continuously adjusted to meet the needs of growing applications.

When the overall load of the cluster exceeds the actual capacity of the cluster, additional nodes can also be added.

2) Easy to implement.

Server clustering technology is easier to implement than other scaling technologies and is mainly done through software. In terms of hardware, a number of low-performance, low-cost servers can be connected together through cluster services to multiply or even multiply the entire server system.

10. Hundreds of times more. Both in terms of software and hardware composition costs, as well as in terms of technical implementation costs, they are lower than other expansion methods.

3) High availability.

Use the cluster service to take ownership of the entire cluster system resources. Such as disk drives and IP addresses will be automatically transferred from the faulty server to the available server. When a system or application in a cluster fails, the cluster software restarts the failed application on an available server, or distributes the work on the failed nodes to the remaining nodes.

During the switchover, the user simply felt that the service took a momentary pause.

4) Manageability.

You can use the cluster manager to manage all the server resources and applications of the cluster system as if they were all running on the same server. Applications can be moved between different servers in the cluster by dragging and dropping cluster objects, data can be moved in the same way, and server load balancing and offloading can be manually done in this way for easy maintenance. At the same time, the status of the cluster can be monitored from nodes and resources anywhere in the network.

When a failed server is reconnected, it will automatically return to working status, and the clustering technology will automatically balance the load in the cluster without the need for on-worker intervention.

Server clustering is a complex feature, and it depends on the needs of your business to determine what architecture to use.

If you are using MySQL clusters, you can use MyCAT middleware for read/write splitting, or you can use MHA to implement MySQL clusters.

If you want to do a web project, you can use lvs+keepalived to do it. You can also use nginx for the reverse**.

Now the more popular may be virtualization, that is, to configure a high-configuration server, in which running docker or openstack and other virtual core technology can also achieve the function of collapsing the blind cluster, one disadvantage is that once the host fails, the entire business is paralyzed, of course, this is more cost-saving.

Essentially, cluster deployments don't improve performance, and performance looks like improvements are just additional attributes.

The main purpose of a cluster is twofold:

Avoid single point problems.

Expansion. A single point of question.

Assuming that there is no cluster deployment, then it means that only one server provides services to the outside world, and if this server goes down, then all clients will not be able to access it.

If a cluster is deployed, let's say two servers are load balanced by iphash, then when one of the servers goes down, only about half of the clients are affected (if all the clients happen to hash the server that hangs down, then there is no way).

If the route has a failover mechanism, it can basically ensure that a server is down, and the client is not affected.

Expansion. In terms of scaling, the maximum load that a server can provide to the outside world is a certain amount. When the number of clients exceeds the maximum load of the server, the server may go down, or the client may respond slowly or even at all.

With cluster deployment, you can increase the maximum number of clients. Assuming that one server can support 1,000 clients at the same time, then two servers can theoretically support 2,000 clients at the same time.

Performance. As mentioned above, cluster deployment does not improve performance. The cluster only has more channels.

For example, if there is a pool, and one pipe fills the pool, then the cluster is deployed with multiple pipes filling the pool. The speed of the individual pipes has not changed, but the flow rate has increased.

It seems that the reason for the performance improvement is that performance is generally defined in terms of TPS QPS, which is the number of requests completed in one second.

After clustering, because the traffic is large, then the TPS must be high.

However, TPS QPS calculates the maximum runnable load of the system. That is, the system can serve externally for a long time under this load, and if the load is exceeded, it may cause the response to be slow or unresponsive, or even cause the server to hang down directly. In general, redundancy is done, so the system will not run for a long time under the maximum operating load.

Therefore, in general, the impact of clustering on performance is limited, and in extreme cases, performance may be degraded. Suppose a server can support 1000 clients, and you deploy two servers, but only 10 clients actually access them, which is far less than the load of a single server, and the cluster needs a routing algorithm, which may be a little slower than a single server.

In the event of any failure, such as hard disk, memory, CPU, motherboard, IO board, or power failure, applications running on this server will be switched to other servers.

Second, the cluster system can solve the problem of software system, we know that in the computer system, the user is using applications and data, and the application system runs on the operating system, and the operating system runs on the server. In this way, as long as any of the application system, operating system, and server fails, the system actually stops providing services to the client, such as our common software crash, which is one of the cases, although the server hardware is intact, the server still cannot provide services to the client. The biggest advantage of clustering is that the monitoring of failed servers is application-based, that is, as soon as the server's application stops running, other related servers will take over the application, regardless of the reason why the application stopped running.

3. The cluster system can solve the situation that the application system stops working due to human error, for example, when the administrator improperly operates a server and the server stops running, the application system running on this server also stops running. Since the cluster monitors the application, other related servers take over the application.

The disadvantages of the cluster system are:

We know that an application in a cluster runs on only one server, and if that application fails, one of the other servers will restart the application, take over the data area on the shared disk enclosure, and get the application up and running again. We know that the entire application takeover process generally requires three steps: detecting and acknowledging the failure, restarting the application on the standby server, and taking over the shared data area.

Therefore, it takes a certain amount of time to switch during the switchover process, and in principle, the switching time will vary depending on the size of the application, and the larger the application, the longer the switching time.

The server cluster is mainly to solve the large-scale application platform, high concurrent load, the cluster can distribute the access pressure of each server, and can also realize server failover, a hardware problem, will quickly transfer to a good server to continue running, business will not be interrupted. This avoids the failure of a single server and the interruption of services due to high access load.