We saw in the last video that the MAC layer enables several applications to share the same radio connection. But what if, these applications don’t have the same requirements in terms of latency or reliability? Or, in other words, different requirements of Quality of Service. In L. T. E, it is the R. L. C layer’s role to provide such features. R. L. C, or Radio Link Control is just above the MAC layer. Let’s see how it works. What is the definition of the quality of service, or QoS in short? It is the capability of the network to give applications a level of service corresponding to their needs, based on different criteria such as delay, throughput, or error rate. The thing to understand about QoS is that it is a compromise. For example, it is possible to increase the reliability, by re-transmitting lost messages. But, of course, it increases the delay. The right balance depends on the application. For a voice communication, we prefer to have a short delay even if we have some losses. Whereas, for bank transaction, we do not tolerate error and the delay is not that important. In the previous lesson, we saw some limitations of the MAC layer. For example, after a certain number of failures, the MAC layer gives up retransmitting messages. We also saw that H. A. R. Q can introduce de-sequencing in messages. Also, the MAC layer does not take care of the message segmentation. That is to say, chopping the messages into smaller blocks that can be inserted into transport blocks. The R. L. C layer offers services to increase the reliability of exchanges, to re-order messages, and to manage segmentation. But this is done at the expense of latency. This is why these services are optional. They can be used for certain data streams and de-activated for others. R. L. C proposes 3 modes. The simplest is transparent mode or TM. In this case, R. L. C does nothing. Because there is no segmentation, this mode is reserved for short messages. It is mainly used to transport signalling messages of L. T. E. The second mode is un-acknowledged Mode or UM. It takes care of re-sequencing received blocks as well as segmenting and concatenating messages depending on the needs of the MAC layer. The impact of this mode on latency is low but it does not improve reliability. It is therefore adapted to real time applications such as voice or video communication. The third mode is Acknowledged Mode or AM. It offers the same services as UM and adds a system to retransmit lost packets which increases reliability. But, of course, this is done at the cost of latency. It is therefore best adapted to applications such as file transfer and Web browsing. In terms of architecture, several R. L. C instances can run in parallel for a given link. For example, a Voice communication over IP can use an R. L. C instance in UM while files are being transferred with another instance in AM. As discussed previously, the MAC layer picks up packets on both instances depending on the parametrized priorities and throughput targets. How does the un-acknowledged mode work? When it receives a packet from the upper layer, R. L. C saves it in a buffer while waiting for the MAC layer to request a MAC-SDU. When the MAC layer requests a MAC-SDU, it specifies the size of the MAC-SDU it is expecting. R. L. C then assembles the stored packets to make an R. L. C-PDU of the requested size. To do that, it can concatenate packets, or split them to reach the specified size. R. L. C also adds headers to tell the receiver where to find the different fragments and which R. L. C instance they belong to. These headers also contain sequence numbers that will enable putting blocks in the correct order if they are de-sequenced by H. A. R. Q. Let’s stop for a second and look at two things. Firstly, the MAC layer can regroup in the same transport block SDUs from different R. L. C instances. Secondly, when trying to retrieve the correct sequence of messages, R. L. C in UM can remain blocked if an intermediate segment never arrives. To avoid this, after a certain time, R. L. C will eventually consider this packet as lost and will go on to the next. Acknowledged Mode works in the same way as UM. But it also manages the repetition of lost messages. To do that, R. L. C memorizes the MAC-SDUs it delivers to the MAC layer, so that it will be able to resend them if necessary. It regularly asks the recipient to send the status of the messages it has received. The receiver answers to this request by sending an R. L. C signaling message tha t is transferred on the data channel. Let’s note in passing that this is the first time we see this case in the radio interface. Until now, all control exchanges were made on dedicated control channels. But now, the answer from the recipient must be separated from the data flow and delivered to the R. L. C controller. The controller can then erase from its memory the messages that are being correctly received and resend ones that have not be delivered to the recipient. Thus, for Acknowledged Mode, it must be the same body that manages the transmission and the reception part while in UM, these processes are independent. To summaries, the R. L. C layer is located above the MAC layer. It takes care of : Re-sequencing of blocks held up by H. A. R. Q. It also manage the concatenation and segmentation of the data to fit the size requested by the MAC layer. And also handles the retransmission of lost blocks. Quality of service is a compromise, notably between reliability and delay. R. L. C defines three modes: transparent mode, un-acknowledged mode and acknowledged mode.
