27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

has shown that performance of the SAN model is very<br />

close to that of the rejection rate reported by SoapUI.<br />

III. RESULTS<br />

Sixteen tests have been conducted on the banking web<br />

service. Each test is run five times and the average of<br />

extracted data is considered as experimental data. Table II<br />

shows a sample data extracted from SoapUI. For each of<br />

the 16 different loads, avg and tps values returned by<br />

SoapUI are used in the SAN model (see Table I) and (5).<br />

Table III shows the results for the 16 tests obtained from<br />

SoapUI, the theoretical equation (5), and from running the<br />

SAN model. As it is shown in the table, the worst<br />

rejection rate happens when there are 380 users, which is<br />

0.13 – 0.08 = 0.05, and 0.14 – 0.08 = 0.06 for the<br />

theoretical and the SAN model, respectively.<br />

Users<br />

(threads<br />

)<br />

TABLE II.<br />

200 10<br />

350 10<br />

352 10<br />

354 10<br />

360 10<br />

370 10<br />

380 10<br />

TABLE III.<br />

Number of<br />

simultaneous<br />

users<br />

Runs<br />

per<br />

threa<br />

d<br />

SAMPLE DATA EXTRACTED FROM SOAPUI<br />

Avg<br />

Cnt (sec)<br />

200<br />

0 7.46<br />

350 13.7<br />

0 6<br />

352 13.8<br />

0 3<br />

354 13.6<br />

0 9<br />

360 14.5<br />

0 4<br />

370 15.1<br />

0 1<br />

380 13.7<br />

0 1<br />

Numbe<br />

r of<br />

request<br />

s<br />

rejected<br />

Total<br />

time<br />

Tps (sec)<br />

24.6<br />

8 82.55 0<br />

22.9 153.2<br />

2 5 0 0<br />

22.2 158.2<br />

5 1 0.5 0<br />

22.6 156.5<br />

1 8 5 0.001<br />

22.1 162.6<br />

5 7 24.2 0.006<br />

21.7 170.4<br />

8 6 185.4 0.05<br />

24.1<br />

3<br />

Request<br />

rejectio<br />

n rate<br />

157.6<br />

4 332 0.08<br />

REQUEST REJECTION RATE FOR ALL SIXTEEN TESTS<br />

Request<br />

rejection rate<br />

(SoapUI)<br />

Request<br />

rejection rate<br />

(rejection est /cnt)<br />

Request<br />

rejection rate<br />

(rejection SRN<br />

/cnt)<br />

200 0 0 0.01<br />

350 0 0 0.04<br />

352 0 0 0.03<br />

354 0.001 0 0.03<br />

360 0.006 0.02 0.04<br />

370 0.05 0.04 0.05<br />

380 0.08 0.13 0.14<br />

400 0.15 0.17 0.17<br />

500 0.4 0.4 0.4<br />

600 0.52 0.5 0.49<br />

650 0.54 0.52 0.52<br />

700 0.51 0.52 0.52<br />

800 0.5 0.54 0.53<br />

1000 0.54 0.51 0.51<br />

1200 0.55 0.52 0.52<br />

1500 0.51 0.5 0.49<br />

Fig. 5 displays graphically the request rejection rate<br />

for the three different models shown in Table III. The<br />

figure shows that the rejection rate of the analytical and<br />

the SAN models closely match the ones provided by<br />

SoapUI. Another interesting result from the experimental<br />

analysis is the throughput. Throughput is computed by:<br />

<br />

<br />

(6)<br />

Fig. 6 shows the throughput as a function of users. As<br />

expected, the throughput decreases as the users are<br />

increased. The tps range reported by SoapUI is from 21 to<br />

45 for an average of about 33, with a standard deviation of<br />

about 10. T he average throughput is about 22.5, which<br />

means on average, 30% of the requests are rejected.<br />

Rejection rate<br />

Rate of web services not serviced<br />

Throughput<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

Figure 6.<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

Figure 7.<br />

200 352 360 380 500 650 800 1200<br />

Number of virtual users<br />

Figure 5. HTTP rejection rate.<br />

Experimental model<br />

Analytical model<br />

200<br />

350<br />

352<br />

354<br />

360<br />

370<br />

380<br />

400<br />

500<br />

600<br />

650<br />

700<br />

800<br />

1000<br />

1200<br />

1500<br />

0<br />

Number of simultanous users<br />

Throughput based on virtual users (extracted<br />

from soapUI).<br />

Failure rate<br />

Total unsuccessful rate (HTTP rejections + failures)<br />

HTTP rejection rate<br />

0.01<br />

0.04<br />

0.08<br />

0.12<br />

0.16<br />

0.2<br />

0.24<br />

0.28<br />

0.32<br />

0.36<br />

0.4<br />

0.44<br />

0.48<br />

0.52<br />

Web service failure probability<br />

Web services rate not serviced using SAN model.<br />

The close approximate behavior of the SAN model,<br />

as shown in Fig. 5, allo ws for cases th at are more<br />

difficult to rep licate in real world, e.g. inducing failures<br />

in the application server (TJW activity in Fig. 4). As an<br />

example, consider Fig. 7 that shows the rate o f failures<br />

due to HT TP rej ections and web service failures. This<br />

309

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!