This means that the source node takes turns to transmit each of the source packets during encounters, 5 and a relay node selects uniformly at random a packet or a generation to transmit during an encounter.įor the buffer management, a drophead scheme has been considered by : when a node with a full buffer receives a packet, it drops the relay packet 6 that has resided in the buffer the longest. Global rarest policy an oracle scheme where a node chooses the packet that has the smallest number of copies in the network.įocusing on the potential benefits of network coding, have studied round robin scheduling for the source packets and randomized transmission scheduling for relay packets. Local rarest policy where each node chooses to transmit the packet that it has transmitted least up to that time instant Random policy that chooses each packet with the same probability In, the following different scheduling policies are considered for non-coding schemes: Larger weights lead to larger amounts of credit, which allow more use of the link. When using weights, the amount of credit allocated to each queue can simply be scaled by the queue's weight. Also, note that inactive queues (i.e., those without packets) cannot accumulate credit. Smaller quantum sizes provide better fairness, but also increase the computational requirements for the link scheduler as more rounds need to be computed before a packet can be transmitted. Larger quantum sizes increase the (temporary) unfairness between connections. The quantum size determines how fine-grained the scheduler operates. There are two scheduling examples using quantum sizes of 100 and 60. (Note that the credit is incremented by the quantum size only once per round, but multiple packet transmissions may be performed.) The process of DRR scheduling is shown in Figure 10-7. While the remaining credit is sufficient to send the next packet in the queue, the process continues. Once the packet is transmitted, the credit of the queue is reduced by the size of the packet. A packet can be transmitted if the accumulated credit is equal to or larger than the packet size. Then, the queue can use the credit to transmit packets. The round-robin operation is as follows: Every time the scheduler visits a queue, the deficit counter for this queue is increased by a fixed amount (i.e., “quantum size”). Credit is accumulated when a packet is delayed in the queue, and credit is used when a packet is transmitted.
This counter tracks the credit that each queue has with the link scheduler. In deficit round-robin (DRR) scheduling, a “deficit counter” is associated with each queue. The waiting time enforced between the transmissions of packets increases with larger packets. The main idea is to track the amount of bandwidth consumed by each queue. This scheduling algorithm was first described in Shreedhar and Varghese. The shortcoming of systematic unfairness due to large packets in round-robin scheduling is addressed in deficit round-robin scheduling. Time-critical program elements should not be interrupted, however, and this requirement will need to be written into the program.ĭimitrios Serpanos, Tilman Wolf, in Architecture of Network Systems, 2011 Deficit round-robin scheduling Therefore, complete context saving (all flags, registers and other memory locations) must be undertaken as the task switches. When the task is allowed to run again, it must be able to pick up operation seamlessly, with no side-effect from the pre-emption. The implications of this pre-emptive task switching, and its overheads, are not insignificant and must be taken into account. This is an example of a ‘pre-emptive’ scheduler. their execution is discontinued mid-flight. Tasks are not allowed to run to completion, but are ‘pre-empted’, i.e. All tasks are treated as being of equal importance and wait in turn for their slot of CPU time. On each clock tick, the current task is discontinued and the next is allowed to start execution. Tasks are selected in a fixed sequence for execution. In round-robin scheduling the operating system is driven by a regular interrupt (the ‘clock tick’). Tim Wilmshurst, in Designing Embedded Systems with PIC Microcontrollers (Second Edition), 2010 18.4.2 Round-robin scheduling and context switching
0 kommentar(er)
