Considerations For Wired and Wireless QoS

Lancez-Vous. C'est gratuit
ou s'inscrire avec votre adresse e-mail
Considerations For Wired and Wireless QoS par Mind Map: Considerations For Wired and Wireless QoS

1. Typical traffic Classes

1.1. Voice

1.2. Mission Critical

1.3. Transactional

1.4. Best Effort

1.5. Scaveger

2. Classification & Marking

2.1. Layer 2

2.1.1. Class Of Service (CoS) Uses only 3 Bits Only Present in Trunk Links

2.2. Layer 3

2.2.1. IP Precedence Uses only 3 Bits Present in Access Links

2.2.2. DSCP Uses 6 Bits The First 3 Create Classes The Second Used for Drop Probability Th last one would be 0 Present in Access Links

3. Congestion Management

3.1. FiFo

3.2. Priority Queuing

3.2.1. Guarantees strict service of a queue (servicing this queue in priority, regardless of the number of packets in the other queues), which is good for the prioritized queue but may starve the other queues.

3.3. Custom Queuing

3.3.1. Custom queuing (CQ) allocates a percentage of the bandwidth to each queue, ensuring good load balancing to each queue, but no real prioritization

3.4. Class-Based Weighted Fair Queuing

3.4.1. Sends in priority smaller packets (low-bandwidth consumption) and packets with a higher ToS value, using an automatic classification algorithm. Allows you to define classes and the amount of bandwidth each should get

3.5. Low Latency Queuing

3.5.1. Commonly used for VoIP, is a variation from CBWFQ where one queue is serviced in priority (in a PQ logic), but only up to a certain amount of bandwidth (so as not to starve the other queues)

4. Congestion Avoidance

4.1. CAC

4.1.1. Allows only a certain number of concurrent calls, avoiding oversubscription of the queue allocated for this traffic. CAC is usually implemented on the platform managing the calls.

4.2. WRED

4.2.1. Drops additional TCP packets as the link becomes more congested (before link saturation), recognizing that TCP windowing will resend the lost packets and reduce the TCP window (thus slowing down the flow). This mechanism avoids link oversubscription, but is adapted only for TCP

4.3. Policing

4.3.1. Drops or Marks Packets when they reach predefined limits. Used on input & output interface

4.4. Sahpping

4.4.1. Buffer Packets, Later if the bandwidth gets lower, these buffered packets can be sent.

4.5. Header Compression

4.5.1. Enables you to drop packet header size from 40 bytes (Layer 4 to Layer 2) to 2 bytes or 4 bytes, by simply indexing each packet. This technique works only on point-to-point links.

5. Distributed Coordinated Function

5.1. Sending a Frame

5.1.1. The DCF coordination is distributed which allows each device to take care of itself. When a device needs to send a data frame, it starts picking a random number. It will then count down from that number. While counting down the station listen the media. Every time it detects a signal, it stops the countdown. The total amount of time waited is: back off time + Time waited during the transmission = Contention Windows

5.1.2. Two ways of determining if the media is free Logical based on Nav Physical through Clear Channel Assessment

5.2. Interframe Spaces

5.2.1. 802.11b SIF = 10 Slot Time = 20

5.2.2. 802.11g SIF = 10 Slot Time = 9

5.2.3. 802.11a SIF = 16 Slot Time = 9

5.3. After a Frame is Sent

5.3.1. Wireless network rely on a system of acknowledgment confirming that the frame was received. The receiving station sends its acknowledge after a short amount of time called SIF When a station sends, it actually reserves the medium for: The duration of its frame + SIF + The duration of the expected ACK

6. Point Coordination Function

6.1. The AP Usually take the Control sending a beacon with the duration field set to 32768.

6.1.1. During the CFP the AP Polls the stations, asking each of them to send. If the polled stations don't have any packet to send the AP can release the cell sending another special beacon with duration field set to 0.

7. 802.11E

7.1. Access Category

7.1.1. AC_VO Network Control (NC) 7 Voice (VO) 6

7.1.2. A_VI Video (VI) 5 Controller Load (CL) 4

7.1.3. AC_BE Excellent Effort (EE) 3 Best Effort (BE) 0

7.1.4. AC_BK Background (BK)

7.2. HCCA

7.2.1. A random access protocol allows fast collision resolution The AP requieres information that hats be update but he stations, information like: Identity of stations that need to be polled. At which times. For Which Duration.

7.2.2. Close to PCF

8. 802.11e Designation

9. 802.11P Priority

10. WMM

10.1. Is a Partial implementation Of 802.11e & Ensure the compatibility between vendors.

10.2. WMM EDCA Queues

10.2.1. When a station needs to send it starts classifying the packet into one of four Categories. (Queues)

10.2.2. Then a backoff timer is picked up, it will depend on Traffic Categories, instead of being the Default aCWIn

10.2.3. All stations run their back off timer in Parallel. If two or more stations reach zero at the same time an internal scheduler avoids the virtual collision by giving priority to the queue with higher QoS.

10.3. WMM Interframe Space

10.3.1. In DCF, after a back off timer reaches 0, the station listens the media,and, if the media is free Waits a DIF before sending. In EDCA, the station waits what is called Arbitrated Interframe Space (AIFS) AIFS can be Fixed or Variable, its variation can depend upon TC.

10.4. TXOP

10.4.1. Enables a station to transmit multiple frames consecutively within a burst after it gains the channel.

10.5. QBSS Information Element

10.5.1. Helps clients to decide which AP to Associate or Roam to. How Chenking Station Count. Checking Channel Utilization. Chenking Available Admission Capacity.

10.5.2. No real Interaction between client an AP.

10.6. Traffic Specification

10.6.1. Allows a WMM station to signal its traffic requirements toe the AP. TSpec in include in ADDTS Action Frame TSpec include in Association & Re-associated messages.

10.7. Power Save Features

10.7.1. PS-POLL With standard 802.11, a station saves battery power by sending a null (empty) frame to the AP with the Power Management bit in the header set to 1. The AP then buffers subsequent traffic for that station. The station wakes up at regular intervals to listen to the AP beacon, which contains a field called the Traffic Indication Map (TIM), indicating the list of stations for which some traffic is buffered. Some beacons also contain a Delivery Traffic Indication Message (DTIM), indicating that the AP has broadcast or multicast traffic that is going to be sent just after the beacon If the station sees its number in the TIM, it sends a special frame (PS-Poll) to the AP to ask for the first buffered packet. The AP contends for medium access and sends the frame, indicating with the More Data bit in the frame header if more packets are buffered. The process is repeated until the More Data bit is set to 0, showing that no more traffic is buffered. This process consumes a lot of frames

10.7.2. Scheduled APSD (S-APSD) The station and the AP negotiate a wakeup time The AP recognizes when the station is going to wake up, and simply sends the buffered traffic in due time

10.7.3. Unscheduled APSD (U-APSD) The station can inform the AP about its awaken state at any time, with any frame where the Power Management bit is set to 0 The AP then empties its buffer in a burst

10.8. Blocks Acknowledgment

10.8.1. The BA Mechanism aggregates several acknowledgments into one Frame. To send several frames,a station first check the TXOP to determine how many frames it can send within the TXOP, It then send an Add Block Acknowledgments After the Block is sent the emitter sends a last frame that contains a BA Request. The receiving station then just need to send an acknowledgment frame to validate the entire Block.