IN-DEPTH ANALYSIS OF THE MULTI-MODAL DISTRIBUTIONS OF MEASURED BANDWIDTHS

ABSTRACT

In this dissertation, we tackle the two virtualization challenges to design new bandwidth estimation methodologies for clouds. First, we design bandwidth estimation methods for networks with rate limiting, which is widely used in cloud networks. Bandwidth estimation for networks with token bucket shapers has been studied before, and the conclusion is that “both capacity and available bandwidth measurement are challenging because of the dichotomy between the raw link bandwidth and the token bucket rate”. Our methods are based on in-depth analysis of the multi-modal distributions of measured bandwidths.

The second virtualization technology is network virtualization. Network virtualization splits a physical network into multiple small virtual networks. Different network users can have different virtual networks over the same shared physical network. Network virtualization has many advantages. First, it isolates the virtual network users from one another, so that it is more convenient and secure for them to use the physical network. Second, it is easier for a network provider to manage the whole physical network.

To formally describe bandwidth, we usually use the term capacity. Capacity is defined as the maximum rate of data transmission that a network path allows a user to send. A links capacity is usually determined by the network cards and the cable connecting the two network cards. A network path includes more than one link. The whole paths capacity is equal to the minimum capacity of the path, which is called a bottleneck.

2.1 HARDWARE CONFIGURATION

PROCESSOR                         :           Pentium Dual Core 1.9 GHz

RAM                                       :           512mb RAM

MONITOR                             :           17” Color

HARD DISK                          :           128 GB

2.2 SOFTWARE CONFIGURATION

OPERATING SYSTEM        :           Windows XP

LANGUAGE                         :           VB.NET

BACKEND                            :           MSSQL