Foreword
I had a reading notes before this chicken dish, arrange a "large-scale Web Site Technology Framework," a book five architectural elements described Zhihui teacher's book. These five factors are performance, availability, flexibility, scalability, security . Aiming availability make this simple discussion of factors, content is the main reference from "large-scale Web Site Technology Framework" This book (a very recommend this book, personally think that this book can be said to exist as the introduction of technology architecture) .
Availability is an important element of the reference site architecture design, site service within a short period of time can not be brought about by the loss are likely to be very great.
Ensure system availability we have the following main means.
redundancy
The main means to ensure high availability is to use redundancy, when a server failure to request other servers.
First, the application server redundancy
Redundant application server is relatively easy to achieve, as long as the application server does not have a state, when an application server failure, the load balancer original user application server forwards the request to another application server, the user will not have any influences.
Refers to a so-called stateless server does not save the context information of the application services ( i.e. Session try not stored on the application server ), only the corresponding service logic processing the data for each request submitted.
The application does not save the state of a high-availability architecture with great convenience, since the status of the request server does not hold , then all servers complete reciprocity , When either or more than one server goes down, the request to the other cluster Renyiyitai available machine processing, so that the end user, the request always successful, the entire system is still available.
Second, the storage server redundancy
Requires the use of redundant storage server master copy from achieved, if the primary server fails, the need to upgrade the server as the master, a process called handover.
monitor
CPU, memory, disk, network load information systems for monitoring, operation and maintenance personnel notification when a message reaches a certain threshold, to detect problems before a system failure.
Service downgraded
Service downgraded the system to cope with a large number of requests, take the initiative to shut down some functions, thus ensuring core functionality is available.
There are two means of demotion:
- Denial of Service
- Close Service
Denial of service : the call is rejected low priority applications, reducing the number of concurrent service call to ensure that the core application of normal use; random or reject the request calls, saving resources, so the other part of the request is successful, avoid the tragedy to die we die together.
Close Services : closure of some nonessential services, internal services or closure of some important features to save overhead, for the important services and functions to make the resource.