The base station permanently but not continuously receives data (bits stored in a queue). Each bit (fifo) is stored in a buffer when this buffer has free space. This buffer may be either a simple array, in this case each bit has to be “tagged” as belonging to a given user, or several arrays each belonging to one user. Let's take one array per user. Those datas are then available to the “scheduler part” of the program. The scheduler may implement one of the three algorithms mentionned earlier, let's take round robin in this example.
Following the round robin algorithm: the scheduler chooses one queue (one user) and extracts data to send to this user (2048 bits corresponding to one OFDM symbol multiplied by the number of resources blocks allocated to the user). This data set is stored (and tagged with the user id) in another queue (fifo) that is destined to the “mapper part” of the program. A counter is then incremented and the scheduler does the same action for the next user. This happens again for each user, when the last user has been choosen, the counter is reseted to prepare the next timeslot.
In the same time the scheduler performs his algorithm, the mapper gets each set of data present on the queue. Those datas are pushed on a map which has time and frequency dimensions. A time slot corresponds to 0.5 ms and 7 OFDM symbols so the second OFDM symbol is then occupied on the first user ressources blocks (because the first OFDM symbol is used for Downlink Control information). The third OFDM symbol on the next user resources blocks is used to store the second user datas and so on. We then have a kind of “stairs topology” on this map (because we use round robin). At each iteration of the scheduler, the mapper allocates the wished ressources blocks (corresponding to the user whom the data are handled) and sends this. In our program it'll certainly send an array with only some elements used. This actions is performed 7 times (7 OFDM symbols) during 0.5 seconds, that's why we consider a “map” (the time dimension is added to the frequency dimension).
This description is about time scheduling because we consider the resources blocks and ressources blocks per user are already defined. So there is another of scheduling to implement. We have to schedule the attribution of each resources block group (RBG). Following the round robin algorithm, it's quite simple: each RBG is iteratively allocated to a user. For each ressource block we choose the next user and increment a counter when we have considered all users and if we still have resources blocks groups to attribute, we go back to the beginning of the users lists. Other algorithms as max CQI may include the transmission conditions parameter to allocate . Each RBG is composed of 1 to 4 resources block (in the case of LTE the channel bandwidth is about 20Mhz so we have 100 resources blocks which means 25 to 100 resources blocks groups).