Adaptive Lock Buffer Algorithm, On Onet,

This dissertation is about the adaptive call-back lock algorithm, a concurrency control strategy. It first gives out the basic theory of this algorithm, analyses it's advantage and weakness and resolves the problem of min-granularity lock by importing lock escalation mechanism. Then the article proposes an improved way to solve the problem of concurrently update in a page. With these measures, this algorithm turns into more practical and perfect, and we implement it for the concurrency control of ONet, an object-oriented database management system.As the development of object orient technology, object orient databases are widely used in many fields. They are GIS, IDSS, Internet database, content management and so on. This is the background of ONet system. Then, we introduce the figure of ONet in brief, including the history, architecture, used techniques and so on.In chapter 3, the basic things of concurrency control in database system are introduced. And we focus on the adaptive call-back lock algorithm, deeply discuss the difference between factual lock and cached lock, the granularity and distributing of the lock, and resolve the problem of min-granularity lock by importing lock escalation mechanism. Then the article proposes an improved way to solve the problem of concurrently update in a page.Chapter 5 is the implement of this algorithm in ONet. First, the architecture was changed from the page server into object-page server, which is the mix of page server and object server. The management of cached lock is the kernel of this algorithm, it's implementing on server sides and client sides were introduced in detail.At last, we concluded the advantages and disadvantages of the concurrency control in ONet, and something that can be improved in the future was put forward.