ECEN 619: Internet Protocols & Modeling

𝑯𝒐𝒎𝒆𝒘𝒐𝒓𝒌 𝟑
Sridhar Mareguddi April 16, 2015
UIN: 823000772

Please complete the following problems in the context of [1]

1. Under first order approximation, derive the following performance parameters:

𝜆 max = 𝜇 + 𝛼𝜏

1
 ECEN 619: Internet Protocols & Modeling

The steady state behavior is defined by the following equations

1
𝜆(𝑡) = (𝜇 + 𝛼𝜏)𝑒𝑥𝑝(− 𝛽 (𝑡 − 𝑡1 − 𝑛𝑇)) for 𝑡1 + 𝑛𝑇 ≤ 𝑡 ≤ 𝑡4 + 𝑛𝑇

𝑡
𝑞(𝑡) = ∫ (𝜆(𝑡 − 𝜏𝐹 ) − 𝜇)𝑑𝑡
𝑡0

Since 𝜆(𝑡) increases linearly at rate 𝛼 for 𝑡 ≤ 𝑡1 , it follows that

𝜆𝑚𝑎𝑥 = 𝜇 + 𝛼𝜏
𝑡
The maximum value qmax of the bottleneck queue is given by qmax =∫𝑡0(𝜆(𝑡) − 𝜇 )𝑑𝑡 and since 𝜆(𝑡) =
1
( 𝜇 + 𝛼𝜏) exp(− (t-t1 )) for t1 ≤ t ≤ t2, we can substitute the value of 𝜆(𝑡) in qmax to get the following
𝛽

𝑡2
1
𝑞𝑚𝑎𝑥 = ∫ (𝜇 + 𝑎𝜏) exp(− (𝑡 − 𝑡1)) − 𝜇)𝑑𝑡
𝑡0 𝛽
1 𝜇
But the first order approximation of (𝜇 + 𝛼𝜏) exp(− (𝑡 − 𝑡1) is (𝜇 + 𝛼𝜏) − (𝑡 − 𝑡1), hence we get
𝛽 𝛽

𝑡2 𝜇
𝑞𝑚𝑎𝑥 = ∫𝑡0 (𝜇 + 𝑎𝜏) − 𝛽
(𝑡 − 𝑡1)) − 𝜇)𝑑𝑡 which gives

𝛼𝜏 2 𝑡2−𝑡1
𝜇
𝑞𝑚𝑎𝑥 = + ∫ (𝛼𝜏 − 𝑡) 𝑑𝑡
2 0 𝛽
𝛼𝜏2 𝜇 𝑡2
𝑞𝑚𝑎𝑥 = + 𝛼𝜏 [𝑡]𝑡2−𝑡1 - [ ]𝑡2−𝑡1
0
2 𝛽 2

𝛼𝜏2 𝜇
𝑞𝑚𝑎𝑥 = 2
+ 𝛼𝜏 (𝑡2 − 𝑡1) - 2𝛽 (𝑡2 − 𝑡1)2

𝜇
First order approximation of (μ + ατ) * e −(t 2−t 1)/β = μ gives (μ + ατ) − 𝛽(t2-t1 ) = μ which gives us (t 2 − t 1)
𝛼𝛽𝜏
= 𝜇
. Substituting this in the equation qmax gives the following

𝛼𝜏 2 𝛼𝛽𝜏 𝜇 𝛼𝛽𝜏 2
𝑞𝑚𝑎𝑥 = + 𝛼𝜏 ∗ − ( )
2 𝜇 2𝛽 𝜇

𝛼𝜏 2 𝛼 2 𝜏 2 𝛽 𝛼 2 𝜏 2 𝛽
𝑞𝑚𝑎𝑥 = + −
2 𝜇 2𝜇

2
 ECEN 619: Internet Protocols & Modeling

𝛼𝜏 2 𝛼 2 𝜏 2 𝛽
𝑞𝑚𝑎𝑥 = +
2 2𝜇

𝜶𝝉𝟐 𝜶𝜷
𝒒𝒎𝒂𝒙 = (𝟏 + )
𝟐 𝝁

T = (t1 - t0) + (t3 - t1) + (t4 - t3) + (t5 -t4)

Substituting these value in T, we get

𝜇− 𝜆𝑚𝑖𝑛
T = 2𝜏 + (𝑡3 − 𝑡1) + 𝛼

𝑡3
But ∫𝑡0 (𝜆(𝑡) − 𝜇)𝑑𝑡 = 0 which yields

𝛼𝜏 2 𝑡3 −𝑡1
𝜇
+∫ (𝛼𝜏 − 𝑡)𝑑𝑡 = 0
2 0 𝛽

𝛼𝜏 2 𝑡 −𝑡 𝜇 𝑡 2 𝑡 −𝑡
+ 𝛼𝜏[𝑡]03 1 − [ ]03 1 = 0
2 𝛽 2

𝛼𝜏 2 𝜇
+ 𝛼𝜏(𝑡3 − 𝑡1 ) − (𝑡 − 𝑡1 )2 = 0
2 2𝛽 3

Solving the quadratic equation for (𝑡3 − 𝑡1 ) gives

𝛼𝛽𝜏 𝜇 𝛼𝛽𝜏 𝜇
(𝑡3 − 𝑡1 ) = 𝑅𝑜𝑜𝑡( (1 + √1 + ) , (1 − √1 + ))
𝜇 𝛼𝛽 𝜇 𝛼𝛽

but since (𝑡3 − 𝑡1 ) can only be +ve , we consider the positive root for it.

Now we have
1
𝜆𝑚𝑖𝑛 = (𝜇 + 𝛼𝜏)𝑒𝑥𝑝(− (𝜏 + 𝑡3 − 𝑡1 )
𝛽
First order approximation of the above gives us

3
 ECEN 619: Internet Protocols & Modeling
𝜇
𝜆𝑚𝑖𝑛 = (𝜇 + 𝛼𝜏) − (𝜏 + 𝑡3 − 𝑡1 )
𝛽
therefore
𝜇
𝜇−𝝀𝒎𝒊𝒏 𝜇−((𝜇+𝛼𝜏)− (𝜏+𝑡3 −𝑡1 )) 𝜇
𝛽
𝜶
becomes 𝜶
which is −𝜏 + 𝛼𝛽 (𝜏 + 𝑡3 − 𝑡1 )

Substituting all the values in T we get

𝜇 𝜇
𝑇 = 2𝜏 + (𝑡3 − 𝑡1 ) − 𝜏 + 𝜏+ (𝑡 − 𝑡1 )
𝛼𝛽 𝛼𝛽 3

𝜇
𝑇 = (𝜏 + (𝑡3 − 𝑡1 ))(1 + )
𝛼𝛽

𝛼𝛽𝜏 𝜇 𝜇
𝑇 = (𝜏 + (1 + √1 + ))(1 + )
𝜇 𝛼𝛽 𝛼𝛽

Rearranging the terms gives us

𝛼𝛽 𝜇 𝜇
𝑇 = 𝜏(1 + )(1 + + √1 + )
𝜇 𝛼𝛽 𝛼𝛽

Hence proved.

𝜆̅ is the average steady state sending rate which can be found by the following

1 𝑡
𝜆̅ = ∫𝑡 5 𝜆(𝑡)𝑑𝑡
𝑇 0

which can be split as between the time intervals (𝑡0 , 𝑡1 ) , (𝑡1 , 𝑡4 ) and (𝑡4 , 𝑡5 ) as below

1 𝛼𝜏 2 𝑡4
1 1
𝜆̅ = [(𝜇𝜏 + ) + ∫ (𝜇 + 𝛼𝜏) 𝑒𝑥𝑝(− (𝑡 − 𝑡1 ) 𝑑𝑡 + (𝜆𝑚𝑖𝑛 (𝑡5 − 𝑡4 ) + (𝜇 − 𝜆𝑚𝑖𝑛 )(𝑡5 − 𝑡4 ))]
𝑇 2 𝑡1 𝛽 2

the above expression is obtained from the fact that during the interval (𝑡0 , 𝑡1 ) the area under the graph
is a trapezium and during the interval (𝑡1 , 𝑡4 ) we have an exponentially decreasing function and during
the remaining interval (𝑡4 , 𝑡5 ) the area under the graph is again a trapezoid.

Now we have the below simplification

4
 ECEN 619: Internet Protocols & Modeling

1 𝛼𝜏 2 𝑡4 −𝑡1
1 1
𝜆̅ = [(𝜇𝜏 + )+∫ (𝜇 + 𝛼𝜏) 𝑒𝑥𝑝(− (𝑡)) 𝑑𝑡 + (𝜇 + 𝜆𝑚𝑖𝑛 )(𝑡5 − 𝑡4 )]
𝑇 2 0 𝛽 2

1 𝜇
from first order approximation we know that (𝜇 + 𝛼𝜏)𝑒𝑥𝑝(− 𝛽 (𝑡)) is (𝜇 + 𝛼𝜏) − 𝛽 𝑡

Substituting the same in the above expression we get

1 𝛼𝜏 2 𝑡4 −𝑡1
𝜇 1
𝜆̅ = [(𝜇𝜏 + )+∫ (𝜇 + 𝛼𝜏 − 𝑡)𝑑𝑡 + (𝜇 + 𝜆𝑚𝑖𝑛 )(𝑡5 − 𝑡4 )]
𝑇 2 0 𝛽 2

𝜇
also we know that first order approximation of 𝜆𝑚𝑖𝑛 = (𝜇 + 𝛼𝜏) − 𝛽 (𝜏 + 𝑡3 − 𝑡1 )

𝜇−𝝀𝒎𝒊𝒏 𝜇
and (𝑡5 − 𝑡4 ) = 𝜶
gives us (𝑡5 − 𝑡4 ) = 𝜏 + 𝛼𝛽 (𝜏 + 𝑡3 − 𝑡1 )

Substituting the above approximations in the 𝜆 we get the following

1 𝛼𝜏 2 𝜇𝑡 2 𝑡4 −𝑡1 1 𝜇 𝜇
𝜆̅ = [(𝜇𝜏 + ) + [(𝜇 + 𝛼𝜏)𝑡 − ]0 + (𝜇 + 𝜇 + 𝛼𝜏 − (𝜏 + 𝑡3 − 𝑡1 ))(𝜏 + (𝜏 + 𝑡3
𝑇 2 2𝛽 2 𝛽 𝛼𝛽
− 𝑡1 ))]

1 𝛼𝜏 2 𝜇 1 𝜇 𝜇
𝜆̅ = [(𝜇𝜏 + ) + (𝜇 + 𝛼𝜏)(𝑡4 − 𝑡1 ) − (𝑡4 − 𝑡1 )2 + (2𝜇 + 𝛼𝜏 − (𝜏 + 𝑡3 − 𝑡1 ))(𝜏 + (𝜏
𝑇 2 2𝛽 2 𝛽 𝛼𝛽
+ 𝑡3 − 𝑡1 ))]

1 𝛼𝜏 2 𝜇 1 𝜇
𝜆̅ = [(𝜇𝜏 + ) + (𝜇 + 𝛼𝜏)(𝜏 + 𝑡3 − 𝑡1 ) − (𝜏 + 𝑡3 − 𝑡1 )2 + (2𝜇 + 𝛼𝜏 − (𝜏 + 𝑡3 − 𝑡1 ))(𝜏
𝑇 2 2𝛽 2 𝛽
𝜇
+ (𝜏 + 𝑡3 − 𝑡1 ))]
𝛼𝛽

𝛼𝛽𝜏 𝜇 𝛼𝛽 𝜇 𝜇
also we have derived that (𝑡3 − 𝑡1 ) = 𝜇
(1 + √1 + 𝛼𝛽) and 𝑇 = 𝜏(1 + 𝜇
)(1 + 𝛼𝛽 + √1 + 𝛼𝛽)

5
 ECEN 619: Internet Protocols & Modeling

substituting the same in above equation gives the following

𝜆̅ =
𝛼𝜏 2 𝛼𝛽𝜏 𝜇 𝜇 𝛼𝛽𝜏 𝜇 1 𝜇 𝛼𝛽𝜏 𝜇 𝜇 𝛼𝛽
[(𝜇𝜏 + ) + (𝜇 + 𝛼𝜏)(𝜏 + (1 + √1 + )) − (𝜏 + (1 + √1 + ))2 + (2𝜇 + 𝛼𝜏 − (𝜏 + (1 + √1 + )))(𝜏 + (𝜏 +
2 𝜇 𝛼𝛽 2𝛽 𝜇 𝛼𝛽 2 𝛽 𝜇 𝛼𝛽 𝛼𝛽 𝜇
𝛼𝛽 𝜇 𝜇
𝜏(1 + )(1 + + 1+ )
𝜇 𝛼𝛽 √ 𝛼𝛽

After some rigorous simplifications we get the following

𝛼𝜏 𝜇 𝜇
𝜆̅ = 𝜇 − ( + √1 + − 1)
2 𝛼𝛽 𝛼𝛽

Hence proved.

Now that we have the following

𝜆𝑚𝑎𝑥 = 𝜇 + 𝛼𝜏
𝜏2 𝛼𝛽
𝑞𝑚𝑎𝑥 = 𝛼 (1 + )
2 𝜇

𝛼𝛽 𝜇 𝜇
𝑇 = 𝜏(1 + )(1 + + √1 + )
𝜇 𝛼𝛽 𝛼𝛽

𝛼𝜏 𝜇 𝜇
𝜆 = 𝜇− ( + √1 + − 1)
2 𝛼𝛽 𝛼𝛽

if we have the condition 𝛼𝛽 = 𝜇 and we substitute the same in the above equations we get

𝜆𝑚𝑎𝑥 = 𝜇 + 𝛼𝜏

6
 ECEN 619: Internet Protocols & Modeling

𝜏2 𝜇
𝑞𝑚𝑎𝑥 = 𝛼 (1 + )
2 𝜇
𝜏2
𝑞𝑚𝑎𝑥 = 𝛼 (1 + 1)
2
𝑞𝑚𝑎𝑥 = 𝛼𝜏 2

𝜇 𝜇 𝜇
𝑇 = 𝜏(1 + )(1 + + √1 + )
𝜇 𝜇 𝜇

𝑇 = 𝜏(1 + 1)(1 + 1 + √1 + 1)

𝑇 = 2𝜏(2 + √2)

𝛼𝜏 𝜇 𝜇
𝜆̅ = 𝜇 − ( + √1 + − 1)
2 𝜇 𝜇

𝛼𝜏
𝜆̅ = 𝜇 − (1 + √2 − 1)
2

𝛼𝜏
𝜆̅ = 𝜇 −
√2

Hence proved.