学生姓名(Name) ：

实习目的及要求Practice purpose and requirements

1)Purpose:
There are different ways to implement concurrency for Internet service. In the lowest level there is threading, which uses either system threads or some kind of light weight threading scheme. Next level is using operating system processes. When a single computer does not have enough capacity, then the concurrency can be implemented using multiple computers. In practice modern Internet services use a combination of these strategies to implement concurrency.
2) Segmentation:
There are usually multiple computers and/or virtual machines serving the clients. These computers can have multiple processes running, and finally a single process can use multiple threads internally. What are then the benefits of using the different concurrency implementations? For multiple servers benefits come from added redundancy and the possibility to bring the servers closer to the clients. Both of these benefits can be expanded further by using the servers located in different data-centers around the world. Of course this decentralization of servers comes with the cost of more complicated implementation and problems, such as consistency and timely access of data. Then the next question with all of these concurrency options is, who needs this increased concurrency? The simple answer is; all internet services. To a certain degree this is probably true. But one segment which will benefit this increased concurrency is large scale internet services like Google or Amazon. According to the definition in a large scale internet service is one with millions hits per day and hundreds of servers.
3) Classification:
The main difference between the operating system threads and lightweight threads is the way they are scheduled. Operating system thread scheduling is closer to scheduling processes and needs multiple operating system and CPU level operations. Lightweight threads can have more lightweight scheduling operations which take less time and resources. One way of implementing lightweight threads is cooperative multitasking. In cooperative multitasking thread or process executes code for a certain number of steps before yielding the execution turn to the next thread or process. Classic well known examples of co-operative multitasking are Windows versions before windows NT and 95. Event based concurrency has been an important part of computing for a long time as evident in a 1975 paper by Leon Presser. Early concurrency related events were often related to the peripherals connected to the computers. Most common event based task even in most modern operating systems is handling of clock interrupt which is used for scheduling different processes to run. User interfaces often have this kind of event based model where the user presses a button or inputs text which is then processed by the event processor.