Technique for adaptively distributing web server requests
View Patent ↗A technique for adaptively distributing a web server request in a system having a plurality of web servers is disclosed. In one embodiment, the technique is realized by first generating a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, wherein each of the plurality of web servers is assigned a respective probability range based upon each respective web server request distribution function. A random probability number is then generated for a web server request. The particular probability range encompassing the random probability number is then determined so as to identify the corresponding web server to which the web server request is to be distributed.
1. A method for adaptively distributing a web server request in a system having a plurality of web servers, the method comprising the steps of:
generating a web server request rejection probability based upon performance measures of each of the plurality of web servers;
admitting a web server request if the web server request rejection probability is less than or equal to a predetermined web server request rejection probability threshold, wherein the web server request is admitted before the web server request is sent or transmitted to a web server;
generating a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, each of the plurality of web servers being assigned a respective probability range based upon each respective web server request distribution function, wherein the web server request distribution function for each of the plurality of web servers is based on recursive filtering functions of the performance measures of each of the plurality of web servers, and wherein the recursive filtering functions of the performance measures of each of the plurality of web servers comprises:
f
^
i
,
p
(
n
)
=
(
1
-
η
)
f
^
i
,
p
(
n
-
1
)
+
η
max
{
0
,
ρ
av
(
n
)
-
ρ
i
(
n
)
ρ
max
(
n
)
}
,
0
<
η
<
1
f
^
i
,
q
(
n
)
=
(
1
-
μ
)
f
^
i
,
q
(
n
-
1
)
+
μ
max
{
0
,
q
av
(
n
)
-
q
i
(
n
)
q
max
(
n
)
}
,
0
<
μ
<
1
wherein n represents the sampling time, i represents a web server, ρ i (n) and q i (n) represent performance measures of the ith web server at sampling time n, ρ av (n) and q av (n) represent average of performance measures across all web servers, and ρ max (n) and q max (n) represent the largest of each performance measure;
generating a random probability number for the admitted web server request; and
determining the particular probability range encompassing the random probability number so as to identify the corresponding web server to which the admitted web server request is to be distributed.
2. The method as defined in claim 1 , wherein the random probability number has a value ranging from 0 to 1.
3. The method as defined in claim 1 , wherein the performance measures of each of the plurality of web servers are weighted.
4. The method as defined in claim 3 , wherein the performance measures of each of the plurality of web servers comprise at least one of a web server processor utilization, a web server queue length, a web server memory utilization, a web server response time, a number of web server sessions, and a number of web server connections.
5. The method of claim 1 further comprising the step of rejecting the web server request if the web server request rejection probability is more than a predetermined web server request rejection probability threshold.
6. The method of claim 5 further comprising the step of queuing the rejected web server request if there is room in an admission control queue.
7. The method of claim 1 wherein the performance measures are provided by each of the plurality of web servers.
8. The method of claim 1 wherein the web server request distribution function for each of the plurality of web servers based upon performance measures of each the plurality of web servers, comprises:
f
i
(
n
)
=
f
i
^
(
n
)
∑
j
=
1
N
f
j
^
(
n
)
,
i
=
1
,
2
,
…
,
N
wherein n represents the sampling time, and i represents a web server.
9. The method of claim 1 wherein the web server request rejection probability is defined as p r (n)=1−p a (n) at sampling interval nΔt, n=1, 2, . . . , wherein p a (n) is probability of accepting the web server request and t is time.
10. The method of claim 1 wherein the web server request is admitted for processing before a web server to which the request is to be delivered to is identified or determined.
11. An apparatus for adaptively distributing a web server request in a system having a plurality of web servers, the apparatus comprising:
a web server request rejection probability generator for generating a web server request rejection probability based upon performance measures of each of the plurality of web servers;
admission control logic for admitting a web server request if the web server request rejection probability is less than or equal to a predetermined web server request rejection probability threshold, wherein the web server request is admitted before the web server request is sent or transmitted to a web server;
a web server request distribution function generator for generating a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, each of the plurality of web servers being assigned a respective probability range based upon each respective web server request distribution function, wherein the web server request distribution function for each of the plurality of web servers is based on recursive filtering functions of the performance measures of each of the plurality of web servers, and wherein the recursive filtering functions of the performance measures of each of the plurality of web servers comprises:
f
^
i
,
p
(
n
)
=
(
1
-
η
)
f
^
i
,
p
(
n
-
1
)
+
η
max
{
0
,
ρ
av
(
n
)
-
ρ
i
(
n
)
ρ
max
(
n
)
}
,
0
<
η
<
1
f
^
i
,
q
(
n
)
=
(
1
-
μ
)
f
^
i
,
q
(
n
-
1
)
+
μ
max
{
0
,
q
av
(
n
)
-
q
i
(
n
)
q
max
(
n
)
}
,
0
<
μ
<
1
wherein n represents the sampling time, i represents a web server, ρ i (n) and q i (n) represent performance measures of the ith web server at sampling time n, ρ av (n) and q av (n) represent average of performance measures across all web servers, and ρ max (n) and q max (n) represent the largest of each performance measure;
a random number generator for generating a random probability number for the admitted web server request; and
web server request distribution logic for determining the particular probability range encompassing the random probability number so as to identify the corresponding web server to which the admitted web server request is to be distributed.
12. The apparatus as defined in claim 11 , wherein the random probability number has a value ranging from 0 to 1.
13. The apparatus as defined in claim 11 , wherein the performance measures of each of the plurality of web servers are weighted.
14. The apparatus as defined in claim 13 , wherein the performance measures of each of the plurality of web servers comprise at least one of a web server processor utilization, a web server queue length, a web server memory utilization, a web server response time, a number of web server sessions, and a number of web server connections.
15. The apparatus as defined in claim 11 , wherein the web server request distribution function for each of the plurality of web servers is based on recursive filtering functions of the performance measures of each of the plurality of web servers.
16. The apparatus as defined in claim 11 , wherein the web server request distribution function for each of the plurality of web servers is based on tracking functions of the performance measures of each of the plurality of web servers.
17. The apparatus of claim 11 wherein the web server request is admitted for processing before being delivered to a web server.
18. An article of manufacture for adaptively distributing a web server request in a system having a plurality of web servers, the article of manufacture comprising:
at least one processor readable medium; and
instructions stored on the at least one medium;
wherein the instructions are configured to be readable from the at least one medium by at least one processor and thereby cause the at least one processor to operate so as to:
generate a web server request rejection probability based upon performance measures of each of the plurality of web servers;
admit a web server request if the web server request rejection probability is less than or equal to a predetermined web server request rejection probability threshold, wherein the web server request is admitted before the web server request is sent or transmitted to a web server;
generate a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, each of the plurality of web servers being assigned a respective probability range based upon each respective web server request distribution function, wherein the web server request distribution function for each of the plurality of web servers is based on recursive filtering functions of the performance measures of each of the plurality of web servers, and wherein the recursive filtering functions of the performance measures of each of the plurality of web servers comprises:
f
^
i
,
ρ
(
n
)
=
(
1
-
η
)
f
^
i
,
ρ
(
n
-
1
)
+
η
max
{
0
,
ρ
av
(
n
)
-
ρ
i
(
n
)
ρ
max
(
n
)
}
,
0
<
η
<
1
f
^
i
,
q
(
n
)
=
(
1
-
μ
)
f
^
i
,
q
(
n
-
1
)
+
μ
max
{
0
,
q
av
(
n
)
-
q
i
(
n
)
q
max
(
n
)
}
,
0
<
μ
<
1
wherein n represents the sampling time, i represents a web server, ρ i (n) and q i (n) represent performance measures of the ith web server at sampling time n, ρ av (n) and q av (n) represent average of performance measures across all web servers, and ρ max (n) and q max (n) represent the largest of each performance measure;
generate a random probability number for admitted web server request; and
determine the particular probability range encompassing the random probability number so as to identify the corresponding web server to which the admitted web server request is to be distributed.
19. The article of manufacture as defined in claim 18 , wherein the random probability number has a value ranging from 0 to 1.
20. The article of manufacture as defined in claim 18 , wherein the performance measures of each of the plurality of web servers are weighted.
21. The article of manufacture as defined in claim 20 , wherein the performance measures of each of the plurality of web servers comprise at least one of a web server processor utilization, a web server queue length, a web server memory utilization, a web server response time, a number of web server sessions, and a number of web server connections.
22. The article of manufacture as defined in claim 18 , wherein the web server request distribution function for each of the plurality of web servers is based on recursive filtering functions of the performance measures of each of the plurality of web servers.
23. The article of manufacture as defined in claim 18 , wherein the web server request distribution function for each of the plurality of web servers is based on tracking functions of the performance measures of each of the plurality of web servers.
24. A method for adaptively distributing a web server request in a system having a plurality of web servers, the method comprising the steps of:
generating a web server request rejection probability based upon performance measures of each of the plurality of web servers;
admitting a web server request if the web server request rejection probability is less than or equal to a predetermined web server request rejection probability threshold, wherein the web server request is admitted before the web server request is sent or transmitted to a web server;
generating a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, each of the plurality of web servers being assigned a respective probability range based upon each respective web server request distribution function, wherein the web server request distribution function for each of the plurality of web servers is based on tracking functions of the performance measures of each of the plurality of web servers, and wherein the tracking functions of the performance measures of each of the plurality of web servers comprises:
f
^
i
,
p
(
n
)
=
max
{
0
,
f
^
i
,
p
(
n
-
1
)
[
1
+
ρ
tar
,
i
-
ρ
i
(
n
)
max
{
ρ
tar
,
i
,
ρ
i
(
n
)
}
]
}
f
^
i
,
q
(
n
)
=
max
{
0
,
f
^
i
,
q
(
n
-
1
)
[
1
+
q
tar
,
i
-
q
i
(
n
)
max
{
q
tar
,
i
,
q
i
(
n
)
}
]
}
wherein n represents the sampling time, i represents a web server, ρ tar,i and q tar,i represent performance targets, and max{ρ tar,i ,ρ i (n) and max{q tar,i ,q i (n) represent normalization terms;
generating a random probability number for the admitted web server request; and
determining the particular probability range encompassing the random probability number so as to identify the corresponding web server to which the admitted web server request is to be distributed.
25. An apparatus for adaptively distributing a web server request in a system having a plurality of web servers, the apparatus comprising:
a web server request rejection probability generator for generating a web server request rejection probability based upon performance measures of each of the plurality of web servers;
admission control logic for admitting a web server request if the web server request rejection probability is less than or equal to a predetermined web server request rejection probability threshold, wherein the web server request is admitted before the web server request is sent or transmitted to a web server;
a web server request distribution function generator for generating a web server request distribution function for each of the plurality of web servers based upon performance measures of each of the plurality of web servers, each of the plurality of web servers being assigned a respective probability range based upon each respective web server request distribution function, wherein the web server request distribution function for each of the plurality of web servers is based on tracking functions of the performance measures of each of the plurality of web servers, and wherein the tracking functions of the performance measures of each of the plurality of web servers comprises:
f
^
i
,
p
(
n
)
=
max
{
0
,
f
^
i
,
p
(
n
-
1
)
[
1
+
ρ
tar
,
i
-
ρ
i
(
n
)
max
{
ρ
tar
,
i
,
ρ
i
(
n
)
}
]
}
f
^
i
,
q
(
n
)
=
max
{
0
,
f
^
i
,
q
(
n
-
1
)
[
1
+
q
tar
,
i
-
q
i
(
n
)
max
{
q
tar
,
i
,
q
i
(
n
)
}
]
}
wherein n represents the sampling time, i represents a web server, ρ tar,i and q tar,i represent performance targets, and max{ρ tar,i ,ρ i (n) and max{q tar,i ,q i (n) represent normalization terms;
a random number generator for generating a random probability number for the admitted web server request; and
web server request distribution logic for determining the particular probability range encompassing the random probability number so as to identify the corresponding web server to which the admitted web server request is to be distributed.