RT, such as domestic large site taobao, baidu, tencent, sina and other large web site must have a server in every region across the country, they each server data is how to achieve synchronization? What is the specific steps? And different areas access to different server computer room, such as jiangsu area access baidu is visiting Shanghai server, the beijing-tianjin wing user access is Beijing server, this is how to implement? It is better for concrete implementation method! Wish you eldest brother elder sister can tell, I really appreciate!

CodePudding user response:

Synchronization is on each server insert data line ah, because he has multiple copies, so each server to do the same action line)

Different regional access server is doing, there is a primary server to determine which will request to the server, such as judging the primary server is now working in Beijing, the less the next access server by Beijing to deal with

CodePudding user response:

Personal think, if is the cluster, can be done by Shared storage, if it is a long distance, is likely to be the CDN to accelerate the synchronization

CodePudding user response:

Server cluster implementation

CodePudding user response:

group machine implementation

CodePudding user response:

reference 4 floor u013129789 response:

group machine implementation.

hey hey

CodePudding user response:

The problem seems to be bigger, what first synchronous data is synchronized data, this must first determine if it is static data (such as images, DHTML file, etc.) the CDN to distribution is a common method (in fact the general also called data synchronization, but called data distribution), if it is a dynamic data (the most typical is the database data), the general is for the attention of the centralized server cluster (in order to ensure the consistency of the data, inventory data, for example, you may be out of sync overbought), but in order to ensure a high concurrent data availability, tend to do a lot of optimization
1. The data is horizontal segmentation, in front of a group of database server set a LB or Proxy, distributing the for different parts of the data access to different server subset group, so only need to do in subsets of internal synchronization, without the need for the large cluster synchronization,
2. To read and write database separation (main database processing write operations, do read with other databases), and, of course, how to do reading and writing database data synchronization can be a challenge,
3. Use a distributed Cache to access to data, because it is a distributed Shared Cache, and all the data access will be here, but the Cache and background database synchronization is a problem,

So, the data synchronization in different architectural requirements are different, and the corresponding solutions will be different, and these technologies are giants core technology,

CodePudding user response:

In addition, this article is very good at http://www.csdn.net/article/2014-01-20/2818197-distributed-system