Question

With the appropriate use of diagrams and clear descriptive text:
•Explain briefly what is meant by tail-drop FIFO queuing.
•Explain what is meant by BufferBloat, and what are the consequences for a VoIP flow's latency and throughput when a VoIP flow and bulk TCP flow share and congest a FIFO bottleneck queue.
•Explain the role and operation of the active queue management (AQM) schemes known as PIE and FQ-CoDel. For both PIE and FQ-CoDel separately, explain how they impact on latency and bandwidth experienced by individual flows when multiple flows compete at a bottleneck router.

Answer 1

• FIFO-First Come First Served queuing.
• Given that the amount of buffer space at each router is finite,if a packet arrives and the buffer space is full,the the router discards that packet.
• This is also called tail-drop,since packets arrive at the tail end of the FIFO are dropped.
• FIFO: It determines the order in which packets are transmitted(scheduling discipline)
• tail-drop:It determines what are the packets to be dropped(drop policy)
• It is most widely used in Internet routers at the time of writing.
• This simplest approach pushes all responsibility for congestion control and resource allocation out of the edges of the network.
• BUFFERBLOAT: Bufferbloat is a huge drag on Internet performance created,ironically,by the previous attempts to make it work better .
• It is simply "Bloated buffers lead to network-crippling latency spikes".
• CONSEQUENCES OF VoIP LATENCY:Latency in VoIP(voice over internet protocal) is the time between the moment a voice packet is transmitted and the moment it reaches its destination.
• High latency translates to delay and echo caused by slow network links.
• It's effects are:
• Overlapping noises,echo,slow and interrupted phone conversations,disturbed synchronisation during video conferencing.
• This problem can be get ridden by using Sufficient Bandwith,Using New Hardware,Clearing Audio Latency,Checking the Wrong Codecs(programs used to encode voice signals).
• THROUGHPUT:Lets see how the throughput of indiviual long-lived TCP flow changes as the no.of VoIP increse.
• We use the same bottleneck link setting and assume that there are only two types of flows in the network:Fixed 10 long-lived TCP flows and g.711 VoIP flows withduration 120 sec and a variable arrival rate.
• Here,TCP throughput drop slows down as we further increse the VoIP arrival rate.
• The curve becomes almost flat as we increse the arrival rate beyond 20 times.
• Some changes can occur in TCP throughput when FEC is used by G.711 flows.
• Therefore they can grab more bandwidth from TCP flows,because with the same arrival rate,g.711 flows with FEC will have better quality and less likely to drop out than G.711 flows without FEC.
• AQM:Active Queue Management schemes controls queue length by dropping or marking packets from bottleneck buffer when it becomes full or queue delay becomes over a threshold value.
• PIE:Proportional Integral Controller and FQ-CoDel are examples of a modern AQM that are designed to remedy these limitations by using queue delay,instead of queue length.
• The bandwidth of PIE controller is about 1.3 db(from 186 Hz to 206 Hz).
• There is no bandwidth separetely for FQ-CoDel(around 3 to 100 Mbit/s).