To the webmaster of WEB2.0

When the Internet noisy into the 2.0 era, when the Internet technology is no longer so unattainable, when replication becomes commonplace, the internet lively up

MySpace is on fire, and more myspace is springing up in China
YouTube has just got up and China's video sites are everywhere
51 springing up, China has a myriad of SNS
Facebook changed the way the Chinese webmaster copied, no longer learn chianren, school fire

When plagiarism becomes a habit, what I want to say is, imitate, stationmaster, are you ready?

If you are planning to do a garbage station, or to earn some advertising fees, please do not click this article, I from the technical point of view of the WEB2.0 site to imitate the problem.

When investment and traffic are not the problem, I want to say is that you really smooth sailing?

Take SNS website, when rush on the line of 2.0, when a sum of investment hit in, when the flow up, your confusion in what place?

I have worked as a technical advisor to more than 2.0 companies, simply to talk about 2.0 of the company's problems (involving privacy, I use a B C D instead), here is no longer to repeat the well-known page static, caching and code security issues, Some technical CTO of the 2.0 companies are aware of these things and we talk about the problems after development

A company

A company is to do is SNS website, the procedure is two hair head lad to do, the goal pointed at 51, the procedure development is smooth sailing, the function also is more than 51 cattle, the promotion is also plain sailing (a company has its own unique promotion way. But when the Alexa to 2W when the problem out, every day around 4 o'clock in the afternoon, the site is slow and amazing, basically not open, the company three servers cpu100%, people are depressed is the company's network configuration way, incredibly is a dual-web cluster, and a single DB database. The entire bottleneck in the database, so I suggest to do the DB cluster, analyzed the data structure, MD, typical web programmer works, not a little database design specifications, functional realization is can, if you want to expand, impossible, cluster is basically impossible, how to do? Can not do, so, one months of time to modify the program, the data structure basically changed over the earlier hit the hundreds of thousands of dozen water drift, the user has gone.

Conclusion: WEB2.0 design should not only consider the function, should seriously consider the underlying and data structure.

B Company

B Company is also doing SNS website, the program is 3 people developed, the CEO is a famous University of Economics master, a little confidant of the taste of the net, there are some features out, to tell the truth, the company's potential is good, the CEO has a strong operational ability, the prospect is good. The system architecture is OK, but---but the system crashed, why? The system does not take into account the user has a mass of the argument, the file also has a mass of accounts, the user's albums, pictures all stored in a Web server partition, each user a directory, and open Performance Monitor, disk IO high staggering, Almost flawless response. As we all know, file system is also a database, a single large file does not matter, the key is the whole is more than 300 g of fragmented files, a large number of read and write operations, system crashes, data loss, a chain of the file system is broken, the user data all lost!!! This is a very heavy problem, the system stopped for one months to do data recovery (separate file is easy, but the mass of files currently do not have a software to organize the software architecture). Solution: Modify the program architecture, do Distributed file storage (program modification took 8 days, but the file transfer has been used for nearly one months), 200,000 users lost

Conclusion: The design of WEB2.0 should have to deal with the mass storage, the whole involved the revision of the program structure, the early planning is not good words basically a way.

C Company

C Company is a respectable company, the CEO of Technical origin, and Bill Gates, like, the university did not graduate out of the network, 01-03 years to do a text message ruthless earn a sum, and then do small projects have become, to tell the truth, I admire. Companies do is alumni, but more emphasis on the MySpace style, focus on the personal homepage, promotion also under the big. The cause of the system crash is actually very simple, because the use of Microsoft's SQL Server, and Microsoft directly told us that SQL Server does not support the cluster, their database overload, 100% did not go down, only lateral increase configuration, using the 4-way 4 core CPU system, But the system still crashes ... High interaction destined for high load. Solution: Now from the basic start, to solve a few programs to consume large energy users, the database using transverse cutting, the user will be grouped every 100,000, at the same time the database system for hashing, a number of tables vertical segmentation, at the same time to group files, to solve the problem. Because the data structure is modified, the program basically moves a bit. Fortunately, the system did not make a big mistake, the loss is not very large, but the user experience caused a bad impact.

Conclusion: The design of WEB2.0 should have good hashing consideration, the program should be able to fit the expansion of the database.

D Company

D Company is a better company in all aspects, do a CDN acceleration, the picture is independent of the N server, a good database, (CTO is a database expert), the system crashes because of the web, according to the truth that the Web is easy to do cluster, but found that the cluster and solve the problem, Their cluster allows only 4 web clusters to be made, but 4 are lost. Careful analysis, to find out why, I estimate that the whole is also most of the CTO's most likely to make a mistake, or they simply can not think of the problem, that is, the Web upload problem, upload time due to the reason, the thread is kept linked, 300 of threads can be a Web server off. Solution: This is the simplest, separate uploads and other energy-consuming large consumers. The program changes are not very large, but the previous half month speed full of user experience loss can not be belittled.

Conclusion: There is no conclusion, after all, there are a lot of experience in the CTO not many, that is, the number of large stations