Unit
- Definition: Traffic engineering is the discipline of analyzing, modeling, and controlling network traffic to achieve performance objectives such as high throughput, low delay, and efficient resource use. It applies to both wired and wireless networks and spans planning, operation, and optimization phases.
- Primary goals: The main objectives include minimizing congestion, balancing load across links, ensuring that service level agreements (SLAs) are met, and maximizing the utilization of network infrastructure without degrading user experience.
- Scope: Traffic engineering covers macroscopic aspects like inter-domain routing policies and microscopic aspects like per-flow queuing. It operates at the network layer and above, often interacting with routing protocols and transport layer mechanisms.
- Key stakeholders: Network operators, service providers, enterprise IT teams, and content delivery networks all rely on traffic engineering to maintain reliable connectivity and cost-effective operations.
- Relation to QoS: While quality of service focuses on per-class treatment, traffic engineering addresses the overall traffic matrix and topology to ensure that resources are available to meet QoS requirements.
- Trade-offs: Optimizing for one metric (e.g., delay) may harm another (e.g., throughput). Traffic engineering involves balancing these trade-offs based on business priorities and application needs.
- Historical evolution: From simple link overprovisioning to sophisticated MPLS-TE and SDN-based approaches, traffic engineering has evolved alongside network complexity and traffic growth.
- Real-world example: A large cloud provider uses traffic engineering to distribute user requests across multiple data centers, rerouting traffic around failed links within seconds while maintaining low latency for video streaming.
- Traffic matrix estimation: Accurate knowledge of traffic demands is fundamental to traffic engineering. Methods include direct measurement, network tomography, and inference from SNMP counters. This matrix drives decisions for routing optimization, capacity planning, and load balancing across the network.
- MPLS-TE: Multiprotocol Label Switching with Traffic Engineering allows explicit path selection, fast reroute, and bandwidth reservation. Tunnels can be set up to bypass congested links and enforce traffic steering, improving overall network utilization and meeting SLA constraints.
- SDN and traffic engineering: Software-defined networking centralizes control, enabling dynamic adaptation to changing traffic patterns. Controllers such as OpenDaylight compute optimal paths using a global view, supporting fine-grained policies, rapid failure recovery, and seamless integration with legacy protocols.
Code Example
Python
import psutil
import time
def measure_link_utilization(interface='eth0'):
prev = psutil.net_io_counters(pernic=True)[interface]
time.sleep(1)
curr = psutil.net_io_counters(pernic=True)[interface]
sent_speed = (curr.bytes_sent - prev.bytes_sent) * 8 / 1e6
recv_speed = (curr.bytes_recv - prev.bytes_recv) * 8 / 1e6
return sent_speed, recv_speed # Mbps
sent, recv = measure_link_utilization('eth0')
print(f'Sent: {sent:.2f} Mbps, Recv: {recv:.2f} Mbps')This Python script uses the psutil library to measure link utilization on a specific interface over a one-second interval. Converting bytes to bits and scaling to Mbps gives a real-time view of traffic load, which is a basic input for traffic engineering decisions.
Quick Revision
- Traffic engineering is the systematic analysis and control of network traffic to meet performance objectives.
- Its primary goals are minimizing congestion, balancing load, and satisfying SLAs.
- It involves trade-offs between metrics such as delay, throughput, and jitter.
- Real-world application includes dynamic rerouting in cloud data centers.