Which QoS mechanism will prevent a decrease in TCP performance?

Question

Options

• AShaper
• BPolicer
• CWRED
• DRate-Limit
• ELLQ
• FFair-Queue

• C
  85% (11)
• D
  8% (1)
• E
  8% (1)

Explanation

WRED Prevents TCP Performance Degradation

WRED (Weighted Random Early Detection) is correct because it proactively drops packets before a queue becomes completely full, triggering individual TCP flows to slow down gradually through TCP's congestion control mechanism. This prevents tail-drop, where the queue fills completely and drops packets from all flows simultaneously - a phenomenon called TCP Global Synchronization, which causes all TCP flows to back off and restart at the same time, creating dramatic throughput oscillations.

Why the distractors are wrong:

• Shaper (A) smooths traffic bursts by buffering, but doesn't address TCP synchronization issues
• Policer (B) drops or re-marks traffic that exceeds a rate, which actually worsens TCP performance by causing abrupt drops
• Rate-Limit (D) is essentially policing - it hard-drops excess traffic, degrading TCP performance
• LLQ (E) prioritizes real-time traffic (like voice), but doesn't specifically protect TCP throughput
• Fair-Queue (F) distributes bandwidth fairly but doesn't prevent global synchronization

Memory Tip: Think of WRED as a "crowd controller" - instead of waiting for the venue to be completely packed before turning people away (causing chaos), it starts selectively turning people away early, preventing a stampede. "Early Detection = Early Prevention."

No community discussion yet for this question.