2017 Exam Flashcards
Where in the protocol stack should network QoS be evaluated?
Between the i/p of source appl and o/p to dest appl
Metrics used to evaluate network QoS (3)
- Data rate / throughput
- Delay (mean and jitter)
- Loss/Error rate (PLR and PER)
PHY Parameters that a system can change + effects upon PHY performance (2)
- System can adapt modulation in terms of bits per psymbol + its coding rate
- Both of these will increase/decrease the raw rate of data tranmission at PHY whilst incr/decr BER + PER
Implications of changing the bits per symbol + coding rate on network QoS (3)
- Changes in PHY affect the throughput + delay
- Increase PHY rate will increase throughput, as long as not too many errors require retransmission
- Increased PER results in more delay + jitter
Definition of CSMA (2)
- ‘listen before talk’
- A node must first check whether the channel is in use. If so, it must wait until its free
Definition of ARQ
- all nodes required to ACK receipt of packets addressed to them, otherwise packet will be retransmitted
6What are the implications of the compression efect on the QoS requirements? (3)
- Compression reduces throughput requirement
- However channel errors usually have a greater impact upon the QoS perceived by the user after the src
- Compression causes a requirement for a lower PER/PLR
Which layers of the ISO protocol are covered by 802 standards + how is functionaility of other layers of the protocol stack required? (2)
- Controls functionality at layers 1 (PHY) + 2 (Data Link Control)
- A convergence layer is also specified, enabling generic higher layer protocols (3+4) to operate on top of the specific 802 standard.
Why does compression effect in 802.11 occur?
- Various overheads in the MAC protocol are of fixed time duration
- Increasing PHY rate shortens the data frame but not the overheads. As PHY rate increases, the total time to transmit reduces but less is used for actual data.
Implications of the “compression effect” for effective throughput? (2)
- Overall slower data rate meaning that throughput converges to a maximum irrespective of the PHY rate.
- Compressive MAC eff. results in diminishing returns
Definition of ‘backhauled’
links between core and subnetworks
Impact of 802.11 BBS being backhauled (2)
- Backhaul tech will limit the packet sizes arriving at the 802.11 AP
- Smaller packets results in a lower MAC eff thererfore limit the throughput
Merits of operating 2x 20MHz BSS vs 1x 40MHz when providing service to multiple user (3)
- Better to operate 2x 20MHz
- Sum of lower rates is greater than rate achieved by 40MHz BSS
- Only works for multiple users as a single user cannot connect to multiple networks
What is BSS (2)
- Basis Service Set
- Deinfe a network of nodes communicating ONLY via 802.11 protocols
Traffic Class Scheduling: UGS
(Unsolicited Grant Services)
BS automatically schedues a fixed fraction of a resource for this service without service having to continually request it, only requests once.
Traffic Class Scheduling: RTPS (2)
(Real Time Priority Service)
- BS does not auto schedule resource, it schedules regular oopportunities in the UL phase for the SS to request resource for their RTPS service.
- Creates more resource request overheads than UGS but is more adpatable to a services’ need.
Traffic Class Scheduling: NRTPS (2)
(Non-real Time Priority Scheduling)
- Similar method to RTPS but also allows nodes to contend to use TDMA slots for resource requests.
- Enables BS to schedule fewer resources for the resource request but makes the availability of resource to NRTPS less reliable
Traffic Class Scheduling: BES (3)
(Best Effort Service)
- No resource requests are scheduled ny BS
- SS may only obtain resource by contending to transmit resource in spare UL data frames
- BES gets worst deal
Order of Priority for Traffic Scheduling
UGS
RTPS
NRTPS
BES
Implication of contention based MA srategies for network QoS as well as fixed/variable packet size with contention based MA?
How does this affect when scaled?
(5)
- Contention based medium access requires nodes to compete for medium access
- No guarantee which node will win
- An ‘unlucky’ node may have to wait longer to access the medium, increasing delays
- Variable packet sizes result in variable packet time occupation of the channel therefore exacerbates delays if node keeps loosing contention to nodes transmitting large packets
- Exacerbated further when there are more nodes contending for access to the medium
Fragmentation Threshold - 1500 bytes
RTS / CTS threshold - 700 bytes
Src node commences contention. Channel remains free until the src’s contention window reaches zero.
Next packet = 800 bytes. No errors occur during the transmission of any frames (5)
- Src win contention, next packet exceeds the RTS/CTS threshold so src transmits an RTS
- Dest responds w/ a CTS
- Src transmits a data frame
- Dest responds w/ ACK
- Upon receiving ACK, src resets CW back to starting value + generates a new value
Fragmentation Threshold - 1500 bytes
RTS / CTS threshold - 700 bytes
Src node commences contention. Channel remains free until the src’s contention window reaches zero.
Next packet = 800 bytes. Error occurs in first frame transmission (4)
- Next packet exceeds threshold, src transmits RTS
- Error occurs, dts CANNOT send CTS
- Src fails to recieve CTS, src infers a collision
- Src doubles CW + generates new value
Sr node commences contention, but before contention value reaches 0, it sense another node transmitting (3)
- Src shouldn’t transmit
- Retain contention value it had when detected other transmission
- Use again in the next contention opportunity
Fragmentation Threshold - 1500 bytes
RTS / CTS threshold - 700 bytes
Src node commences contention. Channel remains free until the src’s contention window reaches zero.
Next packet = 500 bytes. No errors occur during the transmision of the data frames (3)
- Below CTS / RTS threshold therefore transmits the data frame
- Due to no errors, dst should transmit ACK
- Src should reset contention range
Fragmentation Threshold - 1500 bytes
RTS / CTS threshold - 700 bytes
Src node commences contention. Channel remains free until the src’s contention window reaches zero.
Next packet = 500 bytes. Errors occur during the transmision of the data frames (3)
- Below CTS / RTS threshold therefore transmit the data frame
- Due to error, dst does NOT respond w/ ACK
- Src should double its CW range + generate new random value
Block Diagram of ARQ Codec + FEC codec
- ARQ Coder/Decoder
- FEC Coder/Decoder
- FEC Codec, reduces: BER, probability of packet error
- ARQ Codec, eliminates residual packet errors
ARC encoder 1st
HARQ attempts to correct errors at reciever
SDU definition
a unit of data above a particular layer
PDU definition
a unit of data below a layer
From perspective of the PHY layer @ dst, explain what would be the i/p a+ o/p + typical overheads (3)
- PHY @ dst i/p is a PHY PDU, processed to a PPDU to produce a PSDU (PHY SDU), i/p to MAC layer
- PHY removes overheads from the PPDU such as preamble + info on the modulation + coding used in the PPDU
- Use these in equalisation, demodulation + decoding to help recover the correct PSDU content.
Sub-carrier Choice: Arbitrary + Static
requires minimum signalling overhead as the BS will only inform SSs of their resource allocation once
Sub-carrier Choice: Random (3)
- mitigate intercell interference effects especially @ cell edge
- SSs near edge average out the interference effects over time if random allocation changes
- follows a pre-defined pttern to minimise signalling but this pattern MUST be different between cells
Sub-carrier Choice: Channel Aware (2)
- enables BS to exploit channel conditions to ‘best’ effort
- greedy strategy to maximise system throughput, fair strategy or some comprimise between these such as proportional fair strategy
Differences between UL + DL in BS (4)
- DL - multiplex, makes little difference how resources are allocated in time + frequency
- UL - multiple access. Frequency domain multiple access yields a processing gain
- better to share resources in freq rather than time
- simultanteously transmits using a fraction of the total b/w
Why does scheduler not allocate resources in a way that maximises PHY performance (2)
- Scheduler should allocate resources according to traffic needs + maximise PHY performance
- If one SS requires more throughput, it should get more resources even if this lowers processing gain
