802.11 protocol Intensive reading 10: Energy Saving mode (PSM)

Preface

In the 802.11 major editions, a total of four energy-saving models were defined, this article focuses on the most initial PSM mode, for the ASPD added in 802.11e and the psmp,smps mechanism added in 802.11n, we will discuss in the next article.

PSM (Power Save mode): The initial energy-saving mode in the 802.11 protocol, which defines the power-saving mechanisms in both the infrastructure mode and the IBSS mode, and the specific MAC layer working mechanism in the DCF and PCF modes is also different.
As we described earlier, the basic idea of the 802.11 energy-saving mode is that the AP caches the downstream data, and the AP makes feedback on the downstream data only when the node is active to the AP after the end of hibernation. In fact, there is a problem, that is, the node does not know whether the AP has its own cache data. So the actual idea should be that the AP periodically to the corresponding node of its broadcast buffer situation, so that the node can know whether they are cached. At the end of hibernation, the node that is cached data requests data, and then continues to hibernate. This can effectively prevent the node from making some meaningless data requests (that is, the data is not cached on the AP, but the node makes the data request).

So this article we first to answer two questions: 1) How the AP broadcasts its own cache information (i.e., aid,tim and bitmap mechanism), 2) When the AP broadcast the corresponding node buffer information (that is, tsf,tbtt,listen Interval field and CFP Repetition interval). After that, we re-work on the specific MAC layer in the protocol.

PS: The concept of this article is more and more complex, and with the previous description of the DCF and PCF Basic Working mode association is larger. Therefore, this article has tried to follow the first demand after the design of ideas to describe, other deficiencies, please forgive me. Some of the specific energy-saving modes, such as the 802.11v Wnm-sleep mode, are also defined in some of the 802.11 branch versions, which we do not expand.

Aid,tim and Bitmap

In the 802.11 protocol, a mechanism is designed to broadcast its own cache information with fewer fields, for example, we first describe its general idea:

In fact, the AP is a bitmap structure, used to inform the node's own buffer information, if it as a matrix, then the matrix of each row has 8 columns, the bottom contains 1 rows, the maximum contains 251 rows, in other words, the maximum storage space for the matrix is 251 by. Each position in the matrix represents a node, such as the red box position represents the STA0, the Blue box position represents the STA4, the Purple box position represents the STA24. and a specific element of the matrix represents the case of buffer, if the element is 1, then the corresponding node has a data cache, if the element is 0, then there is no data cache. AP cycle broadcast such a bitmap, the node will see if its corresponding location is 1 or 0, and then decide whether to send Ps-poll request data.

Then the specific location of the matrix, corresponding to the node's association ID (AID).

• AID (Association IDentifier): Association ID, which is equivalent to an alias for Sta. There is a association ID table on the AP, where each aid is bound to the macaddr that corresponds to the STA. The scope of the aid is from 0~2007, so it also shows that an AP can associate up to 2007 nodes in the protocol. The location of the aid=0 is a reserved field and is not assigned to a node to represent all multicast and broadcasts.
  For example, the red box describes the Aid=0 has a data cache, which is the existence of multicast or broadcast cache (that is, the STA0 does not exist), the blue box is aid=4 node (that is, the actual node STA4) its data is cached, the purple box is aid=24 node, its data is cached.

aid allocation: When a node (STA) initiates an association request (association requests) to the AP, the AP will respond with the Associated frame (association Response). The aid is also assigned in the process and informs the node (PS: During the re-association process, the aid is also allocated, but we do not discuss it here). For example, is a association response frame format which contains the association ID of this parameter.

Usually the AP in the allocation of aid parameters, should be according to starting from 1, the next to the node, while here, although the display is 2 Byte, that is 16 bits, but in order to be compatible with the Duration/id field we mentioned later, so its highest two bits are 1, as a reserved field, So the range is 1~2007. Because the aid assignment is not the same as the DHCP protocol, an expiration time can be set, that is, the AP does not actively reclaim the aid that has been allocated to certain nodes. So when a new node is added to the AP, the allocated aid is sometimes compared, and the bitmap design in the Tim field behind our analysis has some consideration. Is the display of the specific aid in the grab bag:

When the aid is allocated, the node can use bitmap to broadcast its cached buffer information, in the 802.11 protocol, because the buffer is also periodic feedback, so it is placed in the beacon frame, as a field is carried, the field is the Tim field.
TIM (Traffic Indication map): Traffic indicator, in effect, is a traffic indicator based on the bitmap structure that identifies the cache information of the AP. Its specific structure is as follows:

• Element ID: A code identifying the different fields contained in a beacon frame.
• Length, which describes the length of the element, in fact the element ID and length are the elements necessary for the information element in a general management frame, which is not expanded in detail here.
• DTIM Count , DTIM period:dtim count, and the time interval. In the 802.11 protocol, we can see three concepts, Tim,dtim,atim. Tim is a basic structure of the flow indicator, the standard Tim only indicates the unicast information of the AP cache, DTIM (Delivery traffic indication MAP) is a special Tim, which in addition to the cached unicast information, also indicates the AP cache multicast information. In general, each beacon frame contains a Tim message, but the TIM is not dtim, you need to consider Dtim count and Dtim period two parameters. DTIM period is a cycle that is a fixed value that represents a DTIM after several Tim. While Dtim count is a count value, it is a change, and when Dtim count=0 is represented, the Tim is a dtim. In fact, if DTIM period is set to 1, the DTIM count equals 0 in each Tim field, so each Tim is DTIM. PS: As for Atim,atim is a frame, in the IBSS mode is used, because this article is mainly about the infrastructure mode of the wireless network, so this is not expanded.
• Bitmap control,partial Virtual Bitmap : This field is the specific field of Bitmap, and in fact there are some differences from the BITMAP structure we described at the outset. Below we highlight the specific bitmap structure used in the protocol.
  Bitmap: In front, we give a BITMAP structure that is the most initial, which has some drawbacks
  Occupy a longer transmission time: because in the 802.11 protocol, the actual transmission of beacons is generally the lowest rate, which contains two reasons, 1) Beacon Frame is a broadcast frame, it does not have ACK feedback, so cannot set retransmission mechanism, 2) beacon Frame is broadcast AP basic information, so I hope all the Nodes are able to receive this data effectively, thus adopting a lower rate to ensure that a node with poor signal can receive this information. Therefore, if each beacon frame contains a complete 251 bytes bitmap, then it will occupy a relatively long transmission time, reduce the utilization of the channel, and affect the overall throughput. Therefore, in the actual application, for the aid position which is not used in the back, we can omit the transmission, thus reducing the bitmap space.
  Lack of flexibility: because each location in bitmap corresponds to a node's aid, and as we described earlier, the AP's allocation of aid does not have a recycling mechanism. So it is possible that the aid from the 0~300 location is no longer used node information, and the really active nodes are in the aid from the position around 300~330. That way, if you just get rid of the useless aid position, you will be more than a waste of space around bits (that is, the range of aid that was previously 0~300 without re-use). Therefore, it is not only necessary to omit the bitmap space behind, but also to optimize the bitmap space in the front.
  
  For example, we can see its shortcomings more clearly and think of improved ideas. For example, in the case of this diagram, there is only one node in the aid=120~127 location, the data cache exists. The rest of the aid<120 node does not have the data cache, the AID>127 node also does not have, so the specific transmission, we just notify the AID to 120~127 node, the remaining two parts are redundant, that is meaningless, so we want to compress. In 802.11, the structure of bitmap control and partial virual bitmap is designed to solve this problem, as shown in
  

In the 802.11 protocol, BITMAP control and partial Virtual Bitmap are used together, where the Bitmap control field is divided into two parts:

• The [0] bit is used to indicate whether there is a multicast/broadcast packet being cached, and this is a special bit. If it is 1, then there is multicast/broadcast data being cached and vice versa.
• The [1:7] bit is used to identify bitmap offset, which indicates the offset of the aid. This parameter is used to describe the initial aid in the partial Virtual bitmap, in which the start aid is 0-based since the bitmap offset section is 0 (PS: Because the protocol specifies that aid=0 is a reserved field that identifies all broadcast and multicast information, So there is no allocation to the node for use. But the first bit in the bitmap control field is also the function, so actually aid=0 is not used. ), then the aid=1 represents node 1, while aid=8 represents node 8. The partial virtual bitmap is a variable-length field that is 1~251, that is, if the subsequent aid does not have a corresponding cache, it is not stored in the partial virtual bitmap. (PS: It is important to note that because this is stored in byte form, it is stepping in accordance with 8 bits, if it is not enough 8 bits, it needs to be complete)

We now focus on the use of Bitmap offset, as mentioned earlier, Bitmap offset is used to describe the first aid offset, if the field is 0, the offset value is 0, so in the partial Virtual Bitmap is the beginning of the aid=0 count. Now we describe bitmap offset is not 0, first we give bitmap control every bit of x1~x7, where the order of labels and the general binary representation of the order exactly the opposite. Each bit represents a specific offset, these offsets are calculated as 8 odd, if only X1 is 1, the rest is 0, then the aid start is 2*8=16, if only X4 1, the rest is 0, then the start of the aid is 16*8=128. Here are some more examples:
If there is no multicast/broadcast data, and the AID=24 node has data, then the bitmap control field is "0 1 0 0 0 0 0 0", and the Partial Virtual bitmap is "0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0". In fact, aid=24, is equal to 3*8, if it is binary words, then can be decomposed into (2^1+2^0) *8, that is, there are two positions 1, but because the smallest unit of bitmap offset is 2, rather than 1 (you can see, 2^0 is x0 this bit is actually being multicast /Broadcast data hint this one occupies, so does not use), so the closest starting position with Aid=24 is aid=16, which determines the bitmap control field, and then the partial Virtual bitmap to find the aid=24 location, labeling can , and the last one is a byte.
If there is multicast/broadcast data, and the AID=100 node has data, then the bitmap control field is "1 0 1 1 0 0 0 0", and the Partial Virtual bitmap is "0 0 0 0 1 0 0 0", the exact same calculation method as above.

PS: In the 802.11 protocol, actually gave the bitmap C language implementation of the demo, in the 07 version of the Protocol Annex l have given, with interest can be read by themselves.

What we described earlier is the application of the aid in the Tim field sent by the AP, which is actually the aid parameter used in the node's request to the AP. In the Ps-poll frame where the node requests data from the AP, the duration/id is identified as an aid, which is a ps-poll frame structure:

Where the Duration/id field is directly identified as the aid field, the specific contents are as follows:

Thus the node sends the PS-POLL frame to the AP, directly does not contain the source address, but uses the aid as the substitution, when the AP receives the ps-poll, according to its aid, searches its corresponding cache content in the cache space, then feeds back to the node.

tsf,tbtt,listen Interval field and CFP repetition Interval

As previously described, the AP is periodically sending beacon frames, and the Tim field in the Beacon frame contains information about the corresponding buffers. However, if the TIM information for all nodes is included in each frame, there are two drawbacks:

Limit the sleep time of the node, because each beacon frame contains the TIM information, so the node needs to wake up to receive the corresponding beacon, which is also more energy-intensive, if the node more sleep, it will be more energy-saving
Added Tim Information, if the beacon frame does not contain all the nodes of the Tim information, then you can follow our previous bitmap control and partial Virtual bitmap further compressed space, reduce beacon transmission time waste
So in the 802.11 protocol design, the node is periodically woken up and the AP knows the node's corresponding wake cycle. When the node wakes up, the AP will indicate the buffer information in the corresponding Beacon's Tim field. (In practice, the AP wakes up after the corresponding sleep cycle arrives, and also calculates when the Dtim arrives, since the latter is broadcast information, the former is unicast information).

So in order to achieve the requirements described above, the AP and the node first to achieve time synchronization, and then to specify a cycle, and then complete the periodic acquisition of Tim information such a requirement.

• TSF (Timing synchronization function): The protocol uses the TSF mechanism to describe the timing synchronization function. The TSF timer has a total of 64 bits, in us. In infrastructure mode, timed synchronization is done by a beacon frame sent by a synchronous AP, with a timestamp field in Beacon, which is also 64 bits, and is not stored as element in Beacon, So there must be a timestamp in each beacon frame. When the STA receives the beacon frame of the AP, it extracts the timestamp of the timestamp field and adds a locally estimated delay (received from the antenna port to the last processing local delay) to complete the node-to-AP time synchronization function.

is the beacon frame, timestamp the corresponding field, after which we can see a beacon interval field, which is actually seen in the router configuration, is generally described as beacon time slot, and the size of 100ms, That is 0.1s (PS: Usually, the speed is in accordance with the 10-step, that is 1kbps=1000bps, K is the meaning of kilo. In the 802.11 protocol, a time unit is specified here Tu,time unit,tu is a rare unit of stepping in accordance with the binary, 1tu=1024us, here is actually the kilo-binary counting method. So generally we set the 0.1s, but in the actual beacon frame is 0.1024s, here is a difference), this parameter and TBTT time is consistent.

• TBTT (Target Beacon transmission time): The Beacon is scheduled to be transferred, in fact this is a period of time after the sending/receiving Beacon action, the period of which is determined by beacon interval. When the TBTT time arrives, the AP will actively send beacon frames, and the node will actively accept the beacon frame (including the sleep mode node, will also wake up to accept the beacon), and then use beacon for time synchronization, and view the Tim Field, Without its own data cache, the node continues into hibernation mode until the next TBTT time arrives.

Beacon frames are periodically sent in accordance with TBTT time, but the nodes will not be strictly every beacon need to listen, in order to more efficient design of energy-saving mode, the node should be a few TBTT intervals per interval, and then listen to beacon frames, so that you can extend their sleep time.

• Listen Interval: The listening interval refers to the number of times the workstation wakes up between two times, after how many TBTT, that is, how many beacon frames have been skipped. Longer listening intervals, the longer the node sleeps, the more energy-efficient, but consumes the AP's buffer space and increases the access delay.

Listen interval is a node sent to the AP through the association request frame, so that the AP knows the wake-up time of the node, when the corresponding node wakes up, it will be in the beacon frame, indicating the state of buffer. This field altogether 16 bits, its unit is Beacon's period interval, namely TBTT time, if listen interval set to 2, then represents the node each after two beacon cycle, only then wakes up once. In the capture package, as shown in:

In the 802.11 time period this piece is actually more complicated, because we are talking about the complete energy-saving mode of the protocol, the energy-saving mode is actually divided into DCF and PCF under different working modes, so we also need to understand the CFP repetition interval, A relationship between the CFP cycle and the TBTT time above.

• CFP Repetition interval: The CFP recurrence is a scheduling cycle that coordinates two different modes of operation for PCF and DCF, which includes the CFP and CP two sections, which are used to work with PCF time, CP time is for the DCP to work. This part of the content we discussed in the discussion of PCF.

Finally, with a picture, we identify the relationship between these scheduling cycles:

The TBTT cycle is the period between beacons (the upper blue part), and each interval Tbtt,ap sends a beacon, which may contain the TIM information (the red part of the figure) or dtim information (the purple part of the figure), separated by Dtim The period parameter is set. At the same time, we can also focus on the transmission of beacons may be in the CFP time, it is possible in the CP time, it is actually only in accordance with the TBTT interval can be sent, if within the CFP time, Beacon sent interval is very accurate equals TBTT, and in the CP time, Because of the competitive access mechanism of DCF, there are some minor deviations in the interval between beacons, roughly equal to TBTT time.
The CFP repetition interval is a more noticeable interval (the green part of the figure), and the CFP interval is equal to the multiple of the TBTT time. In fact, we know that the CFP's NAV time setting is done through the beacon Frame's NAV or CF Parameter set mechanism (specifically to see our previous PCF narrative), so the start of the CFP must have a beacon frame, and the end is not necessarily. At the same time, multiple beacon frames may appear in the CFP or CP. As in, the first CFP must start with a beacon frame, and the end is not at the same time as Beacon. Then after the CFP time is over, until the end of the CFP cycle, the rest is the CP time. And in this picture, we can see that there are 2 beacon frames in the first CFP cycle.

PSM mode (Power Save mode)

In the initial 802.11 PSM model, there are different working mechanisms in DCF and PCF modes. Here we first describe how the node enters PSM mode, then describes the DCF work form, and finally we describe the PCF. (PS: As for the two time-period scheduling relationship, we have already mentioned before, here will not be expanded.) ）

1. How to enter PSM mode

node if you want to work in PSM mode, first tell the AP that you will be working in energy saving mode. In 802.11 frames, this information is contained in the MAC layer header in the Frame Control field, that is, any one frame can be used to switch the mode of operation, the node can be associated with the AP when the switch to PSM mode, can also be in the working state, Switch to PSM mode. In the Frame Control field, we have two things to note, such as:

This two content is PWR MGT (Power Management) and more Data.

• Pwr Mgt (Power Management): This field is used to identify whether the node will enter the power-down mode after the frame (PSM mode). If it is 1, then it enters PSM mode, which in turn maintains the current working state. Because the AP itself has a power supply in the infrastructure mode, and the AP is responsible for the core management of the entire network, the field is set to 0 by default in the downlink frame emitted by the AP.

• More Data: The field is the AP indicates whether the node has cached data that is not sent. If 1, the AP also has a corresponding node of the data cache, and vice versa. As we mentioned earlier, in PSM mode, the request data is actually a "ping pong" mechanism, and a request will only have one feedback. So the AP will continuously indicate whether the node is still in the cache, and if there is a cache, the node will continue to request the data when it feeds down the downlink frame. The node will be re-entered into hibernation only after all the cache for the node in the AP has been emptied.

2, DCF under the PSM mode

DCF is a competitive mode of operation in which all nodes require access to the channel to compete, including APS and nodes that work in power-saving mode.
Because the energy-saving mode is related to more parameters, with a legend to illustrate the difficulties, so we try to speak clearly, the lack of the place also forgive me.

Then describes the 1 ap,2 energy-saving node work sequence diagram, the black axis represents the packet sending and receiving time series, purple on the line to represent the wake of the node. The listen Interval of STA1 in the figure is equal to the listen interval=1 of 2,STA2 (there is no identification on the diagram for simplicity of illustration).

2. At first, we assumed that a beacon frame (i.e. 1st Beacon) was sent by the AP, because the beacon frame was the beacon at the initial moment, so the node needed to wake up and receive the beacon frame in advance. As shown in the Beacon frame, the Tim Field identifies that the STA1 node has packets to be passed, and the STA2 node has no packets. So STA1 after receiving beacon, the wake up state is maintained, and STA2 is switched to sleep state.
4. STA1 since the known APS have their own packets, it is first to backoff (here the procedure for Difs,backoff,sifs is the same as the standard DCF, so this is omitted), when the Backoff is complete, it sends a ps-poll frame to the AP, Used to request data.
6. When the AP receives the Ps-poll frame, it first has to feedback an ACK (in the preliminary discussion of the energy-saving mode, we do not emphasize the ACK feedback, In fact, the unicast frames in 802.11 all require ACK feedback, and when the ACK feedback is received, the AP sends the data to STA1, and in the Frame control field of the data, identifies more data=1.
8. When STA1 receives the data from the AP, it needs a feedback ack first, and the more data field in the frame is viewed, because the field is equal to 1, so STA1 cannot be moved to the sleep state. It also needs to continue to request data to the AP, then it passes Backoff again, then sends the PS-POLL frame to the AP.
10. When the AP receives Ps-poll again, the ACK is first, and then the data is fed back. Assuming that this feedback packet is STA1 cached on the AP's last packet, the AP will set the more data field to 0 in the feedback.
12. When the node receives the packet, because more data=0, the STA1 will enter hibernation after the ACK is completed.
14. When the second beacon is transmitted (2nd Beacon), due to STA1 's listen interval=2, it is still dormant and STA2 listen interval=1, so every Beacon cycle needs to wake up, Receive the Beacon. Assuming that the second beacon has no data (not indicated on the figure), the STA2 is then switched to sleep mode after it receives the beacon.
16. When three beacons are transmitted (3rd beacons), STA1 and STA2 will wake up and receive the beacon. At the same time, the Beacon identifies that two nodes have packets that are cached in the AP, so two nodes need to compete for access. In, we assume that STA2 first completes the Backoff and accesses the channel. When the STA2 occupies the channel transmission, the STA1 detects that the channel is busy, so it does not send at the same time, which is actually the process of DCF competition.
18. When STA2 sends Ps-poll to the AP and receives the data from the AP, STA2 checks that the more data field in the packet is 0, so it enters sleep after the packet is received. STA1 also competes with the access channel until it receives the cache corresponding to the AP because it has not yet received the data, and then it enters sleep mode again.

Above, we describe a PSM-DCF basic mode of operation, here we also need to pay extra attention to the point is Dtim time, because the drawing situation is more complex, so we can only describe the situation. In PSM mode, the node wakes up to two conditions, reaching one of them and waking up.
The node wakes up according to the listen interval, that is, every listen interval time, the node will wake up one time until it receives its own cache in the AP before it goes to sleep mode.

Dtim cycle, because we know that Dtim is actually used to distribute the multicast/broadcast frames cached on the AP, all nodes need to wake up at this point and receive this frame. So as long as the beacon is carrying the dtim, all the nodes will also wake up (even if they are not in the listen interval prescribed wake-up time). After the node wakes up, the AP transmits its cached multicast and broadcast frames first, the node receives the multicast or broadcast frames, and is not in the Tim cache at the same time (I understand that because the AP knows the node's listen interval, so not every beacon will carry its Tim information), Then the node will enter sleep until the time it has been set for itself to wake up again.

At the same time, we do not specify the upstream data, when the node local data needs to be sent to the AP, after its backoff, it will actively send data. As to whether the node will send ps-poll or local data first, the author is not fastidious.

3. PSM mode under PCF

PCF is a scheduling-based access method, and the access sequence of the nodes is done through the polling of the APS.

In fact, because of the PCF scheduling mechanism, so generally more energy saving scheme, in, we assume STA1 listen interval=1,sta2 listen interval=1, and we assume dtim period for 4, and the first Beacon, That is, the 1st Beacon carries the Dtim.

2. First the AP sends a beacon, the 1st Beacon, and the node is guaranteed to wake up before the beacon cycle to ensure that it receives information from the Beacon. In 1st beacons, the Dtim information is carried, and a cache of broadcast packets is identified in the AP.
4. When the beacon time is over, the AP will first send the cached broadcast frame, because it is a broadcast frame, so there is no need to wait for ACK feedback. When a broadcast frame is sent, the AP sends a data frame to the node in turn and polls the node. Because the information that is carried in Dtim indicates that both STA1 and STA2 have data that is cached by the AP, STA1 and STA2 remain wake up until the downstream packet is received.
6. The AP will first send a data+cf-poll frame to STA1, which is a frame, but it has two functions (i.e. sending data to STA1 and requesting the STA1 upstream data), which has already been mentioned in the PCF working mode introduction.
8. When the STA1 receives the data, it sends an ACK and its own cache to the AP, and similarly, the information is encapsulated in a single frame, which is a one-time feedback in the data+cf-ack frame, saving some inter-frame spacing. At the same time, because the data field of the AP is equal to 1, so STA1 receives the data, it also needs to keep the wake up state, waiting for the remaining data below the AP.
10. In PCF, the AP polls the node sequentially in ascending order of the aid, so even if the data field is sent to STA1, more data=1, but it still has to poll the next node instead of staying on STA1. The AP sends a data+cf-ack+cf-poll frame to the STA2, which can be noted that there are three functions in this frame, both ACK acknowledgement to STA1, sending data to STA2, and data request.
12. When the STA2 receives the data, it also wants the AP to send the Data+cf-ack frame, and since the more data field equals 0, the STA2 transmits the data and enters the sleep state. And since STA1 and STA2 's listen interval are equal to 1, STA2 will wake up before the next beacon arrives and receive the beacon data.
14. In the second Beacon, the 2rd Beacon,tim indicates that the STA1 has data, the STA1 and AP continue to Exchange data (omitted in the figure), and STA2 does not have a data cache, so after receiving Beacon, STA2 will sleep.
16. In the third Beacon, the 3rd beacon,tim field indicates that STA1 has data, and STA2 has no data. So STA2 still received beacon into sleep, and STA1 first receive the downlink data sent by the AP, here because we assume dtim period is 4, so the beacon is carrying Tim information, that is, the AP does not have multicast/broadcast packets need to be sent first. So the AP sends Data+cf-poll to STA1 at the beginning, and since the more data field equals 0 in that frame, the AP enters sleep after it successfully receives the data frame and feedback the ACK.

802.11 protocol Intensive reading 10: Energy Saving mode (PSM)