Queue的一些有用的特性
Arrival process在像网络这样的系统中往往是非常复杂的,因此可以把它看作random process
The study of queues with random process is called queuing theory
当在某个interval中间的departure因为在这段时间内没有新的packet进入到queue,称作shadow departure(圈出的虚线箭头)
It's a departure opportunity that we could have sent a packet,but we didn't since the queue is empty
1. Burstiness increases delay(Bursty arrivals)
如果假设 packet follow a periodic arrival (1 packet per second)
Average occupancy of the queue will always be equal or less than 1
如果假设 packet follow a bursty periodic arrival (N packet per N second 但是聚集在一起)
因此即使rate不变,只是改成bursty arrival
Average occupancy of the queue will be much higher,and the variance of Qt will increase
2. Determinism minimize delay
(i.e. random arrival waits longer on average)
3. Little's result
L - Average number in the queue
lambda - arrival rate
d - average delay
Poisson Process
Why poisson process?
aggregation of many independent random events
makes the math easier
几个需要注意的问题
Example (M/M/1 Queue)
连续马尔可夫链