In recent years, Web services and applications are gradually shifting into cloud environments. These ever-increasing cloud services are being hosted on large scale computation and storage infrastructures called data centers (DC). A data center contains a large number of servers interconnected by a specifically designed network, named data center network (DCN). Today’s data centers host a variety of different applications that exhibit unique traffic characteristics. Some applications exhibit many-to-one communication pattern, some others follow ON-OFF pattern for flow arrivals resulting in bursty traffic. The many-to-one communication is a frequent pattern in data centers and may cause buffer overflow at the bottleneck link which ultimately results in throughput degradation of the flows; this problem is known as TCP Incast. In addition, data centers exhibit heavy tailed distribution for flow sizes; more than 80% of the data center flows have size less than 10 KB, called short flows, but around 90% of the traffic volume, in data center networks, is produced by long flows. Short flows are usually latency-sensitive and long flows demand high throughput. Short flows suffer to quickly increase their sending rates due to the existing long flows occupying most of the available capacity; this results in inefficient flow completion time (FCT) of short flows which ultimately results in degradation of the application’s performance. Furthermore, data centers employ scale-out model for growth to keep up with the rapid increase in the cloud services; thus, multiple paths exist between each pair of end-hosts. Load balancing of the traffic over the multiple path is required to improve network utilization. In addition, requirements like: i) high performance for applications, ii) scalability, and iii) robustness against topology changes, makes the load balancing task more challenging. This research thesis proposes queue length based solutions to the performance issues of TCP Incast and flow completion time of short flows. In addition, a queue length based traffic load balancing scheme is proposed to improve network utilization and applications’ performance. The proposed solutions are validated through implementation in network simulator and compared with the existing schemes. The results show that the proposed solutions significantly improve the performance of applications in the data centers.
Chapters
Title |
Author |
Supervisor |
Degree |
Institute |
Title |
Author |
Supervisor |
Degree |
Institute |
Title |
Author |
Supervisor |
Degree |
Institute |
Title |
Author |
Supervisor |
Degree |
Institute |
Book |
Author(s) |
Year |
Publisher |
Book |
Author(s) |
Year |
Publisher |
Chapter |
Author(s) |
Book |
Book Authors |
Year |
Publisher |
Chapter |
Author(s) |
Book |
Book Authors |
Year |
Publisher |
Similar News
Headline |
Date |
News Paper |
Country |
Headline |
Date |
News Paper |
Country |
Similar Articles
Article Title |
Authors |
Journal |
Vol Info |
Language |
Article Title |
Authors |
Journal |
Vol Info |
Language |
Similar Article Headings
Heading |
Article Title |
Authors |
Journal |
Vol Info |
Heading |
Article Title |
Authors |
Journal |
Vol Info |