Abstract: Owing to excellent extensibility and usability，heterogeneous clusters have gradually become the focus of current parallel computing.In the heterogeneous clusters with real-time requirements，scheduling is the key to improve the system performance.It proposes two self-adaptive scheduling algorithms，SANOL and SAOL，which can adjust the task service level according to the burden of system on the premise of satisfying the requirements of real-time and fault tolerance，to improve system flexibility，schedulability and resource utilization.The two algorithms are compared with an effective scheduling algorithm DYFARS by simulation experiments.The experimental results show that the algorithm of SAOL has the superiority to others with higher performance quality.

Key words: real-time, heterogeneous clusters, fault tolerance, self-adaptive algorithm, service level, primary/backup task copy

摘要： 异构集群由于良好的扩展性和可用性，逐渐成为当前并行计算的热点。在具有实时性要求的异构集群中，调度是提高系统性能的关键所在。在此提出了两种自适应调度算法SANOL和SAOL，在保证异构集群中任务的实时性和容错性的前提下，自适应地根据系统的负载情况动态地调整任务的服务级别，从而提高整个系统的灵活性、可调度性和资源利用率。通过实验将这两种算法与另外一种有效率的调度算法DYFARS算法进行比较，结果表明所提出的SAOL算法具有更好的性能。

关键词: 实时, 异构集群, 容错, 自适应算法, 服务级别, primary/backup任务拷贝