-
Scalability
- A distributed program is considered to be scalable if it remains effective when the quantities of users, data, and resources increase significantly.
- 从线性编程到分布式编程倍速公式:
- Ts: sequential programming running time
- Tp: parallel programming running time
- s: portion of program that is not parallelizable
- n: a cluster of n nodes
-
Communication
- Colocation: Distributing/partitioning work across machines attempts to place highly communicating entities together.
-
Heterogeneity
- In principle, if there is variety and difference in data center components, the cloud is referred to as a heterogeneous cloud
- Reasons:
- Cloud providers typically maintain multiple generations of IT resources, purchased over different time frames.
- Cloud providers are increasingly applying virtualization technology on their clouds to consolidate servers, enhance system utilization, and simplify management. Public clouds are primarily virtualized data centers. Even on private clouds, virtualized environments are expected to become the norm.
- Middleware: The part that can be incorporated in distributed programs to work out heterogeneity is commonly referred to as middleware.
-
Deadlock
- For instance, if task A succeeds in acquiring lock1 and, at about the same time, task B succeeds in acquiring lock2; then if task A attempts to acquire lock2, and task B attempts to acquire lock1, we have what is known as a deadlock.
云计算:核心概念
最新推荐文章于 2023-07-15 01:54:50 发布