Design And Implementation Of An OpenStack Scheduler

With the coming of the informationization, OpenStack, as one of the best OpenSource free software that can manage the IaaS cloud, has been widely adopt and researched in the industry. However, OpenStack has some weakpoints, especially the scheduling policy. When customer submits a request to the OpenStack system, the scheduler provides resources only once. If the request fails, no more chance that the request can get. Moreover, the schedule policy is a static one. From this Viewpoint, research OpenStack scheduler and improve its operating efficiency is significant. The main works of this paper as follows:1. Refrences the FairShare algorithm which has been successfully applied in batch system and grid computing. To modify the fairshare-scheduler according to the distribution of the cloud computing environment, I study the schedule strategy of cloud computing.2. By using MySQL to achieve a queue mechanism and design a retry mechanism, the article designs and implements a scheduler named FairShareScheduler that can be used in OpenStack. The FairShare part can ensure the customer a promising amount of resources with fairness, customer's requests are queued in a queue, request's priority is calculated by FairShare algorithm, and retry mechanism ensures an amount of retry chances, which can help improve the system efficiency.3. By using the FairShareScheduler can help improve the performance by automaticly re-schedule the previously failed request while resource is satisfied. And the FaiShare algorithm itself can ensure the customer get the resources which are promised.