Scalable design application server clusters
The application server is designed to be stateless, so that, each time a user requests can be sent to any server cluster up, and the results of any one server are the same.
Each request for each user may fall on different servers, the distribution means require an HTTP request, can be perceived or may be configured of a cluster number of servers, the new server cluster can be found on-line or off the assembly line in a timely manner, and to the new server on-line distribution request to the server to stop distribution request has been off the assembly line, then realized the scalability of the application server cluster.
This HTTP request is called the distribution server load balancing device.
Load balancing can not only achieve scalability site, while also improving the usability of the site. The following specifically addressed several basic technologies to achieve load balancing.
6.2.1 HTTP redirect load balancing
Calculated according to a user's HTTP requests a Web server's real address, and writes the address of the Web server returns the HTTP redirect response to the user's browser (302 in response to the state it).
Pros: Simple solutions
Cons: browser needs two requests to complete a server access, poor performance;
Redirect server itself may become a bottleneck in processing power, scalability limited size of the entire cluster;
Use HTTP302 response to redirect it, it has the potential to cause a search to determine SEO cheating, lower search rankings.
In practice this rarely practical solutions.
6.2.2 DNS DNS load balancing
A domain name server DNS configuration corresponding plurality of IP
We are each domain name resolution request according to a load balancing algorithm returns a different IP address.
advantage:
DNS load balancing to work, eliminating the need for cumbersome site management and maintenance of server load balancing, while many also supports location-based DNS domain name resolution, domain name will resolve to the user from a server geographically closest address, accelerate user access speeds and improve performance.
Disadvantages:
Multi-level resolution, each level may cache DNS A record current DNS, when a server is offline, even if modify the DNS A-record, to put it into effect also takes a long time, this time, DNS will still the domain name resolves to a server goes offline, causing the user to access failure;
DNS load balancing and control over the domain name resolution service provider where it can not do more to improve the site and more powerful management.
In fact, a large part of the site is always practical DNS name resolution, use DNS load balancing means as a first-class physical server, that is, a set of DNS servers have is not practical to provide Web services, but also provides load balancing internal server, which then set the internal load balancing server load balancing, distribute requests to the real Web server.
6.2.3 reverse proxy load balancing
Most reverse proxy server provides load balancing function, the load balancing algorithm according to the request forwarded to a different Web server. Web server processing is complete response also needs to be returned to the user through a reverse proxy server.
Because the Web server does not directly provide external access, the Web server does not need to use an external IP address, and reverse proxy server you need to configure dual-card sets of external and internal IP addresses.
Reverse proxy server in the HTTP protocol level work, also known as application layer load balancing, layer 7 load balancing.
advantage:
And reverse proxy server functionality integrated, simple to deploy
Disadvantages:
When reverse proxy server for all requests and responses of the relay station, which performance can be a bottleneck.
6.2.4 IP Load Balancing