Method for the early indication of serialization bottlenecks
View Patent ↗The present invention provides a method for identifying a serialization bottleneck in a computer system (FIG. 1 ). The method comprises the steps of providing statistical data with regard to the actual response time (FIG. 2 ) of a computing resource, providing statistical data with regard to an expected response time of the computing resource and applying a mathematical formula to the actual response time to derive a relationship between the actual and expected response time. The relationship between the expected response time and the actual response time provides an indication of the presence of a serialization problem in a computing system.
1. A method for identifying a serialization bottleneck in a computer system, comprising the steps of:
calculating the probability of a transaction request being queued before the transaction request is processed utilizing the mathematical formula:
p
q
=
∑
n
=
1
c
-
1
cL
n
n
!
*
cL
c
c
!
(
1
-
L
)
where p q is the probability of a transaction being queued; c is the total number of computing resources (expressed as an integer) and L is the utilization of each computing resource;
providing an actual response time of a computing resource;
providing an expected response time of the computing resource, and comparing the actual response time and the expected response time to derive a relationship between the actual and expected response time, whereby the relationship between the expected response time and the actual response time provides an indication of the presence of a serialization problem in a computing system.
2. A method in accordance with claim 1 , wherein the expected response time is calculated utilizing the mathematical formula:
E
r
=
d
*
(
1
+
p
q
c
*
(
1
-
L
)
)
where E r is the expected response time, p q is the probability of a transaction being queued; c is the total number of computing resources (expressed as an integer); L is the utilization of each computing resource; and d is the average time taken for a transaction to be processed.
3. A system in accordance with claim 1 , wherein the expected response time is calculated utilizing the mathematical formula:
p
q
=
∑
n
=
1
c
-
1
cL
n
n
!
*
cL
c
c
!
(
1
-
L
)
where E r is the expected response time, p q is the probability of a transaction being queued; c is the total number of computing resources (expressed as an integer); L is the utilization of each computing resource; and d is the average time taken for a transaction to be processed.
4. A method in accordance with claim 1 , the step of deriving the relationship further comprising dividing the actual response time by the expected response time to produce a ratio of the actual response time to the expected response time.
5. A method in accordance with claim 4 , further comprising the step of providing a plurality of actual response times and expected response times for a computing resource, each of the plurality of actual and expected response times being derived when the computing response is subjected to a different load, and comparing the actual response time and expected response time at each respective load.
6. A method in accordance with claim 5 , wherein the ratio of the actual response time and the expected response time is plotted on a graph against the load on the computing resource, to provide a visual representation of the relationship between the ratio and the load.
7. A data processing system for identifying a serialization bottleneck in a computer system implemented using software executing within a computing system, comprising:
means to calculate the probability of a transaction request being queued before the request is processed, wherein the probability of queuing is calculated utilizing the mathematical formula:
E
r
=
d
*
(
1
+
p
q
c
*
(
1
-
L
)
)
where p q is the probability of a transaction being queued; c is the total number of computing resources (expressed as an integer) and L is the utilization of each computing resource;
means for providing an actual response time of a computing resource, means for providing an expected response time of the computing resource, and means comparing the actual response time and
the expected response time to derive a relationship between the actual and expected response time,
wherein the relationship between the expected response time and the actual response time provides an indication of the presence of a serialization problem in a computing system.
8. A system in accordance with claim 7 , wherein the means for deriving the relationship further comprises means for dividing the actual response time by the expected response time to produce a ratio of the actual response time to the expected response time.
9. A system in accordance with claim 8 , further comprising means for providing a plurality of actual response times and expected response times for a computing resource, each of the plurality of actual and expected response times being derived when the computing response is subjected to a different load, wherein the means for comparing the actual response time and expected response time compares the times at each respective load.
10. A system in accordance with claim 9 , further comprising means to plot the ratio of the actual response time and the expected response time against the load on the computing resource, to provide a visual representation of the relationship between the ratio and the load.