QEST2013 paper

The final publication is available at www.springerlink.com from http://link.springer.com/chapter/10.1007/978-3-642-40196-1_19
(a)
TQ.csv
S
F1
A
G
F2
SI
result
.dat
(c)
TD.txt
U
F3
F2
F1
A
G
VWAP
(b)
CD.csv
S
F1
A
G
result
.dat
SI
CDR
SI
result
.dat
Twitter
(d)
ND.csv
S
F1
SI
result
.dat
Micro-benchmark
Key
S
Fn
AG
SI
Source
Functor n
Aggregate
Sink
Thousands
8
Throughput of four applications on S4
10
VWAP
100
Throughput (Tuples\s)
12
Micro-benchmark
CDR Optimized
80
Twitter
60
40
20
6
CDR Optimized
4
VWAP
Twitter
Micro-benchmark
2
Throughput (Events\s)
CDR
Thousands
Throughput of five applications on System S
120
Throughput (Events\s)
Thousands
140
60
Throughput of four Esper Applications
50
40
Micro-benchmark
30
Twitter
VWAP
20
CDR
10
CDR
0
0
0
2
4
6
8
Number of Nodes
(a)
10
12
14
0
0
2
4
6
8
Number of Nodes
(b)
10
12
14
0
2
4
6
8
Number of Nodes
(c)
10
12
14
90
100
Percentage of CPU_CLK_UNHALT events consumed by processes on each
node for running CDR on S4 using 12 Nodes
80
70
java
60
kernel
50
40
30
20
10
Percentage of counted
CPU_CLK_UNHALT events (%)
Percentage of counted
CPU_CLK_UNHALT events (%)
100
0
90
80
Percentage of CPU_CLK_UNHALT events consumed by processes on each
node for running micro-benchmark on S4 using 12 Nodes
70
java
60
kernel
50
40
30
20
10
0
sa01
sa02
sa03
sa04
sa05
sa06
sa07
Host Name
sb01
sb02
sb03
sb04
sb05
sa01
sa02
sa03
sa04
sa05
sa06
(b)
100
90
Percentage of counted CPU_CLK_UNHALT
events (%)
Percentage of counted CPU_CLK_UNHALT
events (%)
(a)
Percentage of CPU_CLK_UNHALT events consumed by processes on
each node for running CDR on System S using 12 Nodes
80
70
60
50
40
30
20
10
0
sa01
sa02
sa03
sa04
sa05
sa06
sa07
Host name
(c)
sb01
sb02
sb03
sb04
sb05
sa07
sb01
Host name
sb02
sb03
sb04
sb05
100
90
Percentage of CPU_CLK_UNHALT events consumed by processes on each node for
running micro-benchmark on System S using 12 Nodes
80
Legend for
(c) and (d)
70
60
libstreams-pe
50
kernel
40
libc
30
libperl
20
libstdc++
10
All System S
Components
0
sa01
sa02
sa03
sa04
sa05
sa06
sa07
sb01
(d)
sb02 sb03
Host Name
sb04
sb05
(libstreams-pe,
libc, libstdc++,
libperl)
6000
5000
Memory usage of S4 for different applications running on 12 nodes
4000
CDR
3500
VWAP
3000
Twitter
2500
Micro-benchmark
2000
1500
1000
Memory usage during experiment (MB)
Memory usage during experiment (MB)
4500
5000
Memory usage of System S for different applications running on
12 nodes
4000
CDR
3000
VWAP
Twitter
2000
Micro-benchmark
1000
500
0
sa01
sa02
sa03
sa04
sa05
sa06
sa07
Node Name
(a)
sb01
sb02
sb03
sb04
sb05
0
sa01
sa02
sa03
sa04
sa05
sa06
sa07
Node Name
(b)
sb01
sb02
sb03
sb04
sb05