Vous êtes sur la page 1sur 3
Research paper of lottery scheduling #DB22883 Xu Hongsi Questions Answering 1) What problem does the paper solve and why is it important? The purpose is to control the number of resource users, resource users select the license footprint. It solves starvation. To guarantee a non-zero probability of being selected at the scheduling operation, the algorithm gives each process at least one lottery ticket. Because lottery scheduling algorithm is a probabilistic scheduling algorithm, processing "a quantity of lottery" to give different processes different calling time. It is important because it can be generalized to manage many diverse resources, such as I/O bandwidth, memory, and access to locks. 2) What is the hypothesis of the work? The hypothesis is there is a algorithm that has major impact on throughput and response time, it’s fair, it also adjust rapidly. 3) What solution does the paper propose and what is the key insight? It can give additional lottery process in order to increase the winning chances of a more important process. If we sold 100 tickets, and there is a process in which holds 20 of them, then the process has a 20% chance of winning in every lottery pick. In the long run, this process will get 20% of the CPU. Lottery scheduling has several interesting properties. For example, if there is a new process and get some lottery tickets appeared, then the next time the lottery, the lottery process will be in proportion with the number it holds opportunities to win prizes. In other words, lottery scheduling is responsive. 4) Does the idea work, and if so, how well? 1.Fairness Although there is clearly some variation, the two tasks remain close to their allocated ratios throughout the experiment. Note that if a scheduling quantum of 10 milliseconds were used instead of the 100 millisecond Mach quantum, the same degree of fairness would be observed over a series of subsecond time windows. 2.Flexible control It’s allocated a share of time that is proportional to the square of its relative error. When a new task is started, it initially receives a large share of the processor. This share diminishes as the task reduces its error to a value closer to that of the other executing tasks. 3.Client-Server Computation The ticket allocations affect both response time and throughput. After the high-priority client terminates, the response times are 44.17 and 15.18 seconds,for a 2.91 : 1 ratio. For all average response times, the standard deviation is less than 7% of the average. 4.Multimedia Applications Lottery scheduling enables the desired control at the operatingsystem level, eliminating the need for mutually trusting or wellbehaved applications. 5.Load Insulation Support for multiple ticket currencies facilitates modular resource management. A currency defines a resource management abstraction barrier that locally contains intracurrency fluctuations such as inflation. The currency abstraction can be used to flexibly isolate or group users, tasks, and threads. 5) What is a drawback or limitation of the system and how you might improve it? It’s not as fair as we thought. 6) What are the paper's contributions (author's and your opinion)? Ideas, methods, software, experimental results, experimental techniques...? It presented lottery scheduling, a novel mechanism that provides efficient and responsive control over the relative execution rates of computations. Summary Lottery scheduling facilitates modular resource management, can be generalized to manage diverse resources. Since lottery scheduling is conceptually simple and easily implemented, it can be added to existing operating systems to provide greatly improved control over resource consumption rates.