How to Maximize User Satisfaction Degree in Multi-service IP Networks

Huy Anh Nguyen, Tam Van Nguyen, Deokjai Choi

TL;DR

Researchers have developed BASMIN, a bandwidth allocation scheme that prioritizes user "happiness" over raw throughput. By using a new metric called Network Worth, which weights application utility by priority, BASMIN outperforms traditional allocation methods in multi-service IP networks.

What Problem Does This Paper Address?

As we move toward the "Future Internet," network traffic is growing far faster than our ability to expand bandwidth. Traditionally, network management focused on raw metrics like throughput or packet loss. However, these don't always correlate with how satisfied a user actually is. For example, a slight drop in bandwidth might go unnoticed by someone downloading a file (Elastic traffic) but could completely ruin a VoIP call (Hard Real-Time traffic).

The challenge is twofold:

  1. How do we mathematically define "user satisfaction" across different types of applications?
  2. How do we allocate limited resources to maximize the total satisfaction of the entire network, especially when some services are more critical than others?

Methods

The authors propose the Bandwidth Allocation Scheme for Multi-service IP Networks (BASMIN). The core of their approach rests on three pillars:

  • Utility Functions: They use modified versions of Shenker’s utility curves to map allocated bandwidth to a satisfaction score (0 to 1). They categorize traffic into three types: Elastic (e.g., email), Hard Real-Time (e.g., telemedicine), and Real-Time (e.g., online gaming).
  • Network Worth ($W_k$): This is their novel metric. It isn't just the sum of utility; it’s utility weighted by priority. They use an exponential weighting system ($W_k = 2^{(i)}U_k$), where $i$ is the priority level. This ensures that satisfying one high-priority user is mathematically more valuable than satisfying several low-priority ones.
  • The BASMIN Algorithm: A three-step process consisting of:
  • Admission Control: Deciding whether to let a new flow in based on current congestion and the flow's priority.
  • Path Selection: Finding the best route to accommodate the new flow.
  • Load Balancing: An iterative algorithm that redistributes bandwidth among active flows based on their marginal utility—essentially asking, "Who would benefit the most from an extra kilobit of data right now?"

Datasets

The paper does not use a public dataset. Instead, the authors built a custom Java-based network simulator to evaluate their algorithm. They defined six specific traffic profiles (ranging from voice services to file transfers) with varying bandwidth needs and utility functions to create a realistic multi-service environment.

Key Insights and Ideas

  • Utility is Subjective: The authors argue that utility isn't fixed. A service's value depends on the user's perception and the application's specific requirements.
  • The Failure of Static Partitioning: One of the most striking insights is how poorly "Complete Partitioning" (giving each service type a fixed slice of the pie) performs. While common in mid-sized corporate networks, it is too rigid for the dynamic nature of modern traffic.
  • Total vs. Average Satisfaction: There is a trade-off between making everyone a little bit happy versus making the most important users very happy. BASMIN leans toward the latter by maximizing "Total Network Worth."

Key Results

The authors compared BASMIN against three industry standards: Best Effort, Complete Partitioning, and Trunk Reservation.

  • Superior "Worth": BASMIN consistently achieved higher Total Network Worth as bandwidth increased compared to Best Effort and Complete Partitioning.
  • Efficiency vs. Pickiness: While a method called "Trunk Reservation" showed higher average connection worth, it achieved this by being extremely "picky"—rejecting many users to keep a few very happy. BASMIN, by contrast, managed to keep more users satisfied while maintaining high total worth.
  • Link Utilization: Best Effort (predictably) had the highest link utilization because it has no admission control, but BASMIN remained competitive while providing much better Quality of Service (QoS) guarantees.

Limitations and Future Work

The primary limitation is the simulation environment. While the Java simulator is a good first step, the authors acknowledge that the network model used was relatively simple.

Future Work includes: * Expanding BASMIN to include a traffic rerouting mechanism to dynamically clear room for incoming high-priority requests. * Testing the algorithm in more complex MPLS (Multi-Protocol Label Switching) environments to see how it handles large-scale provider networks.

Summary updated

BibTeX
@inproceedings{nguyen2009how,
  author = {Huy Anh Nguyen and Tam Van Nguyen and Deokjai Choi},
  title = {{How to Maximize User Satisfaction Degree in Multi-service IP Networks}},
  booktitle = {ACIIDS},
  year = {2009},
  doi = {10.1109/aciids.2009.16},
}
Figure