Windows Azure Combat: Building Highly scalable applications

2010 Beijing TechEd on December 1, at the Beijing National Convention Center as scheduled, until December 3, the many technical people's haunting ceremony near the end, but it seems the enthusiasm of the whole venue is not the slightest weakening trend, IT168 will take you to continue tracking TechEd the latest technical information.

Masashi Narumoto, a lecturer for Microsoft, introduced the Architecting applications for high scalability course, which he showed in an example of how to use Microsoft's cloud computing platform to build highly scalable applications.

▲

▲windows Azure Architecture

Masashi that cloud computing is the ultimate distributed computing, to achieve the high scalability of cloud computing, the most important is six aspects: data fragmentation, anti-normalization, base rules, stateless, asynchronous, parallel.

Fragmentation: The so-called fragmentation is to divide the storage into several pieces of data, and the partition key using hashing algorithm, partition a typical application scenario is Weibo. The data fragment can be flexibly load-balanced and parallel query, which improves the query speed. There are also challenges, such as large query issues across partitions, connectivity issues across partitions, referential integrity questions, rebalancing, and more.

Anti-normalization: When designing relational databases, we are told that we should try to follow the principles of the Paradigm (NF) (paradigm, sometimes called planning), and the main goal of the paradigm is to reduce redundancy. The main problem with redundancy is duplication of data and increased complexity of operations. But as with many things in the world, database design cannot be blunt. Paradigm does avoid redundancy, but it also poses other problems, with two main problems: 1. Performance issues. As the pattern is designed, the data is split into different tables as much as possible. In this case, to find a complete piece of data, you need to join (join) A number of tables, which undoubtedly lowers the speed. 2. Historical data issues. For example, employees will have the title of information, but to consider the staff title will change. If you keep only one title information, there will be problems in some systems. So in this case, we have to consider recording the change in employee title. So, if we want to solve the above two kinds of problems, we may have to consider "anti-normalization", that is, to allow partial redundancy.