OPNET Technologies
7255 Woodmont Avenue 
Bethesda, MD 20814
Tel: 240-497-3000
Fax: 240-497-3001
E-mail: university@opnet.com
Web: www.opnet.com

© 2008 OPNET Technologies, Inc.
OPNET and OPNET Technologies Inc.
are registered trademarks of OPNET. 
All trademarks are the property of their
respective owners and are used herein 
for identification purposes only.

Title:  Internet Bandwidth  Distribution (2002 - 2008)
Despite numerous efforts, the problem of providing per-flow Quality of Service in a scalable manner still remains an active area of research. To solve the problem of providing scalable per-flow Quality of Service, a number of service differentiation models have been proposed. The Integrated and Differentiated Service models are among the most prominent approaches to providing Quality of Service in the Internet. The Integrated Services (IntServ) model requires each router in the network to reserve and manage resources for the flows that travel through it. In large networks, millions of flows may simultaneously travel through the same core routers. In such cases, managing resource reservations on a per-flow basis may cause enormous processing and storage overheads in the core routers. As a result, the IntServ is considered to be not scalable to large networks and thus is not widely deployed in the Internet. The Differentiated Services (DiffServ) model attempts to solve the scalability problem of the IntServ approach by combining flows that have similar quality of service requirements into the traffic aggregates or classes. The DiffServ core routers process incoming traffic based on the class the packets belong to and thus maintain and manage resource reservations only on a per-class/per-aggregate basis. Although the Differentiated Services approach provides a scalable solution to the QoS problem, it supports only coarse per-aggregate guarantees which in certain cases may not be adequate. 

We introduced an alternative approach for providing scalable per-flow bandwidth guarantees, called the Bandwidth Distribution Scheme (BDS). In this approach, the core routers do not maintain per-flow information (e.g. bandwidth requirements of individual flows); instead core routers keep aggregate flow requirements. The amount of information kept in the network core is proportional not to the number of flows but to the number of edge routers, which we believe does not raise scalability concerns. The edge nodes maintain per-flow information and fairly allocate network resources (e.g. bandwidth) among individual flows according to the flow requirements and resource availability. The BDS relies on the idea of pushing per-flow information to the network edges while keeping traffic aggregate information in the network core. The primary contribution of this research is a novel approach to aggregating flow requirements and a new distributed network feedback protocol that allows the edge nodes to discover network changes and compute fair per-flow bandwidth allocation that satisfies minimum bandwidth guarantees of individual flows.

We have designed and studied two different variations of BDS, named Estimation-Based BDS (S-BDS) and Exact Requested Bandwidth Range BDS (X-BDS). In X-BDS, the edge routers obtain the exact values of the aggregated flow requirements as contrasted with S-BDS, where the edge routers estimate the aggregate flow requirements. The X-BDS variation improves on the performance of the S-BDS approach in several categories. In particular, the X-BDS provides a more accurate computation of the flow fair shares, converges to the optimal rates faster, and extends the basic BDS architecture by including the admission control unit. 

The current study of the Bandwidth Distribution Scheme was conducted using the OPNET Modeler version 6.0 network simulation software. We implemented the Bandwidth Distribution Scheme in OPNET’s Proto-C and integrated it with the OPNET’s implementation of IP protocol. 

List of ongoing projects using OPNET Modeler for evaluation of BDS:

• BDS implementation (Summer 2006 -current): Update implementation of the Bandwidth Distribution Scheme for OPNET Modeler version 11.5 
• Designers: Matt Stiefel and Andrew Fabian
• Completed work:
• Updated BDS process model
• Modified ip_output_iface process model to include BDS processing
• Current stuatus:
• Debugging the BDS model

• Influence of TCP on BDS (Spring 2005 - current): Currently, we studied the influence of the BDS approach on the multimedia traffic performance. Multimedia traffic usually uses the UDP as its underlying transport protocol, which is unreliable and ignores any packet loss. This works well with the BDS approach where all out-of-profile packets (e.g. the packets that arrive at the rate higher than the rate allocated to the flow) are discarded. However, certain applications that can benefit from the BDS approach, for example FTP, use TCP as their transport protocol. TCP is reliable transport protocol and it treats packet loss as an indication of severe congestion. In which case TCP reduces the transmission window of a flow, effectively slowing down the rate of that flow. In this study we would like to examine what effects, if any, does the BDS out-of-profile packet treatment policy has on the TCP traffic. We would like examine the following two approaches for dealing with this issue:  (1) shaping (e.g. delaying) out-of-profile traffic until it becomes in-profile and (2) signaling the BDS rate adjustment events directly to the TCP process, effectively moving the BDS bandwidth policing mechanism from the edge routers into the source nodes. 
• Designers: J. Pucci, C. Clement, and J. Ogren
• Completed work:
• Developed a BDS process model
• Conducted a preliminary study of the TCP influence of traffic limiting mechanisms that rely on packet drops.
• Presented results at PDPTA'05 conference.
• Current status:
• Updating BDS model to continue study.

• Rowan University network (Spring 2006): Develop a simulataion model of the Rowan University network.
• Designers: Networking Club (James Metting, Andrew Fabian, Matt Stiefel, Pavel Bashkirtsev, Mike Simmons, Robert DeDomenico, Gregg Gramatges) 
• Completed work:
• Develope an outline of student profiles
• Developed a simulation model of Computer Lab
• Developed a model of AIM application
• Comparated AIM model with Ethereal trace
• To Present the result at International Conference on Telecommunication Systems – Modeling and Analysis,  October 5-8, 2006
• Current stuatus:
• Postponed until start of Fall '06 semester

• Provisioning of Differentiated Services networks using the BDS approach:One of the salient features of the Differentiated Services (DiffServ) approach is its static per-class network provisioning This may lead to violation of QoS requirements or waste of resources when a certain class is either over-subscribed or under-subscribed, respectively. Currently, the DiffServ approach relies on the Bandwidth Broker, a centralized node that maintains complete network’s information, to monitor network traffic and periodically adjust per-class resource allocation. We would like to study the possibility of improving the DiffServ provisioning mechanism by applying the BDS resource distribution protocol. The main advantages of using the BDS approach for DiffServ provisioning are: (1) absence of single point of failure, (2) robust and distributed BDS protocol dynamically adjusts per class resource reservation based on class usage and current network situation, (4) supports DiffServ scalability properties, and (3) provides framework for admission control in DiffServ networks. 
• Current status: postponed until BDS implementation is completed

• BDS for the inter-domain traffic: Currently, we examined the performance of the BDS approach only within the confines of a single network domain. However, we would like to investigate the venues for extending the BDS approach to the multi-domain environment. One of the main challenges of this problem is scalable propagation and aggregation of bandwidth requirements of individual flows that traverse multiple network domains. To solve this problem we must re-examine the data structures maintained at the edge routers of the BDS network domains.
• Current status: postponed until BDS implementation is completed

• BDS for Mobile environment: We would like to further investigate the possibility of applying the BDS approach in the mobile environment. The primary problems of extending the BDS approach to mobile environment are: (1) discovering available bandwidth (2) sharing available resource among individual nodes, and (3) properly updating the BDS information in the network upon topology changes.  We made the first steps towards resolving this problem by examining the influence of the topology changes on the BDS performance. We introduced a set of optimization to the BDS resource distribution protocol in the events of link failure and link being restored [5]. However, the other issues of bandwidth discovery and bandwidth distribution in mobile environment still remain unanswered. 
• Current status: postponed until BDS implementation is completed

• The BDS fairness issues: The BDS resource allocation mechanism relies on the “water-filling” techniques for distribution of excess bandwidth among the corresponding flows. The edge routers periodically probe the network and increase allocated rates of corresponding flows if the excess bandwidth was discovered. We would like to further examine the BDS mechanism for distribution of excess bandwidth as follows (1) create a mathematical model of the BDS resource distribution process, (2) conduct a simulation study of the BDS fairness properties, (3) conduct a study of the BDS fairness properties using a real network, and (4) compare obtained results. Based on the obtained results we would like to improve, if needed, the resource distribution mechanism of the Bandwidth Distribution Scheme.
• Current status: postponed until BDS implementation is completed