기본
Quantitative
Capacity
Number of Transmission Links ( Routers )
ex) Network 를 보안에 따라 나누기
Buffer Sizes
버퍼크기 = RTT * Link_Capacity / root(number of TCP flow)
Size of the Switching Fabric
Metrics
Call Blocking => Capacity 보다 초과시 블럭
Packet loss Probability => Capacity, Buffer Size
Moments and Distribution Delay
Exp(D), Variation(D), Pr{D>x} ( Guarantee )
Throughput
Above is Statistical(확률적)
as Deterministic Quantity make Under Utilization
don't know parameters that make randomness
Arrival Packet time, Holding Time, Pack Size, Transmission Facility
=> Queueing Theory
이런 랜덤 가득한 네트워크의 성능을 분석하고 만들기 위해 필요
Queue is Variable
Lots of Job + Limited Resources => Make Queue
Delay Delay = Processing + Queue + Transmission + Propagation
Queue => 나머진 다 상수인데 얘만 가변적임 => Queue Theory
Theory of Queue
Server => S
Buffer => B
at least 1 Packet Length as It should store for Transmission
S < B imply Packet Loss Probability => 1/(S*B)
...
왜 1에 가까울수록 queue 확률이 커지냐를 수학적으로 증명하는건데 나중에