update code for Figure 6(c)

Author: tjulym

Figure/matlab/motivation/GsightLatency.m

@@ -1,70 +1,33 @@
-% clc
-% clear
-load GsightLatencyData;
+clc
+clear
+load GsightLatencyDataNew;
 figure1 = figure('PaperSize',[20.98404194812 29.67743169791]);
 
 axes1 = axes('Parent',figure1,'YGrid','on');
 box(axes1,'on');
 hold(axes1,'all');
-GsightLatencyData=GsightLatencyData;
-GsightLatencyData(:,[5 3 2])=GsightLatencyData(:,[5 3 2])*0.8;
-a1 = GsightLatencyData(:,5);%EFRA390  252
-b1 = GsightLatencyData(:,1);%ElaX469  343
-c1 = GsightLatencyData(:,2);%EFRA390  425
-d1 = GsightLatencyData(:,4)-200;%peak401 550
-e1 = GsightLatencyData(:,6)-200;%PRESS681 642
-
-hold on
-lw = 3 
-xi = linspace(min(a1),max(a1),100);
-F = ksdensity(a1,xi,'function','cdf');
-plot(xi,F,'--', 'LineWidth',lw ,'color',[35 31 32]/255); %红色  
- 
-xi = linspace(min(b1),max(b1),100);
-F = ksdensity(b1,xi,'function','cdf');
-plot(xi,F,':', 'LineWidth',lw ,'color',[35 31 32]/255); %紫色
+lw=3;
+plot(GsightLatencyDataNew(:,2),GsightLatencyDataNew(:,1),'--', 'LineWidth',lw ,'color',[35 31 32]/255); %红色  
+plot(GsightLatencyDataNew(:,4),GsightLatencyDataNew(:,3),':', 'LineWidth',lw ,'color',[35 31 32]/255); %紫色
+plot(GsightLatencyDataNew(:,6),GsightLatencyDataNew(:,5),'-', 'LineWidth',lw ,'color', [238 41 47]/255);  %绿色   
+plot(GsightLatencyDataNew(:,8),GsightLatencyDataNew(:,7),'-', 'LineWidth',lw ,'color', [35 31 32]/255) %黄色  
+plot(GsightLatencyDataNew(:,10),GsightLatencyDataNew(:,9),'-.', 'LineWidth',lw ,'color', [35 31 32]/255) %黄色  
 
-xi = linspace(min(c1),max(c1),100);
-F = ksdensity(c1,xi,'function','cdf');
-plot(xi,F,'-', 'LineWidth',lw ,'color', [238 41 47]/255);  %绿色   
-
-xi = linspace(min(d1),max(d1),100);
-F = ksdensity(d1,xi,'function','cdf');
-plot(xi,F,'-', 'LineWidth',lw ,'color', [35 31 32]/255) %黄色  
-
-xi = linspace(min(e1),max(e1),100);
-F = ksdensity(e1,xi,'function','cdf');
-plot(xi,F,'-.', 'LineWidth',lw ,'color', [35 31 32]/255) %黄色  
-% aa1 = plot(a1,x,'-', 'LineWidth',lw ,'color',[255 215 0]/255); %红色  
-% bb1 = plot(b1,x,'-', 'LineWidth',lw ,'color', [153 102 153]/255); %紫色
-% cc1 = plot(c1,x,'-', 'LineWidth',lw ,'color', [51 153 51]/255);  %绿色   
-% dd1 = plot(e1,x,'-.', 'LineWidth',1.5,'color', [149 14 8]/255) %黄色  
-% ee1 = plot([0 1400],[0.99 0.99],'-.', 'LineWidth',1.5,'color', [0 0 255]/255) %  
-% hold on;
-% plot(967,0.9,'or', 'MarkerSize',12,'LineWidth',2.5) %黄色  
-% hold on;
-% plot(542,0.9,'or', 'MarkerSize',12,'LineWidth',2.5) %黄色  
 set(gca,'YLim',[0  1]);%X轴的数据显示范围
-set(gca,'YTick',[0 : .2: 1]);%设置要显示坐标刻度
-% %set(gca,'yticklabels',{'0' ,'24'  ,'48',  '72', '96',  '120'});
+set(gca,'YTick',[0 : .2: 1]);%设置要显示坐标刻度 
 set(gca,'XLim',[100  1300]);%X轴的数据显示范围
 set(gca,'XTick',[100 : 300: 1300]);%设置要显示坐标刻度
 set(gca,'xticklabels',[50 : 150: 1300]);%设置要显示坐标刻度
-set(gca, 'Fontsize' ,18)
- %set(gca,'XTicklabel',{0:2:10});%设置要显示坐标刻度
-
-% title('EMU of redis with BE Tasks', 'FontSize' , 13)
+set(gca, 'Fontsize' ,18) 
+ 
 xlabel('Latency (ms)', 'Fontsize' ,18)
 ylabel('CDF', 'Fontsize' ,18)
 set(gca, 'GridLineStyle', ':','ticklength',[0.002 0]) 
 set(gca,'xcolor',[0 0 0]);
- set(gca,'ycolor',[0 0 0]);
-%set the position of figure and axis 
- set(gcf,'position',[100 100 300 240])
-%  set(gca,'units','normalized','position',[0.2 0.3 0.6 0.5],'box','off')
- set(gca,'units','normalized','position',[0.24 0.3 0.7 0.65],'box','off')
- %legend content  
-legend({'KNN','LR','RFR','SVR','MLP','90th'},'FontSize',12)
+set(gca,'ycolor',[0 0 0]); 
+set(gcf,'position',[100 100 300 240]) 
+set(gca,'units','normalized','position',[0.24 0.3 0.7 0.65],'box','off') 
+legend('KNN','LR','RFR','SVR','MLP','90th')
 box on
 grid on
 

Figure/matlab/motivation/GsightLatencyDataNew.mat

@@ -3,18 +3,40 @@
 ## Benchmark Installation
 1. Deployment guide of OpenFaaS for Kubernetes is available here: https://docs.openfaas.com/deployment/kubernetes/.
 2. Run benchmarks /install.sh to build & deploy all functions.
+
 Notice: this operation takes some time.
 We can also build & deploy functions according to the description below: https://docs.openfaas.com/cli/build/.
 
 ## Workload Generator Installation
 Guidance is available here: https://github.com/tjulym/gray/blob/SC21ORO/loadGen/README.md.
+
 Notice: Be sure that OpenFaaS functions and workload generator work well before running thefollowing script. The project LoadGenSimClient cannot work if the LoadGen has not yet been deployed.
 
 ## Metrics Collector
-Compile gray/gsightCollector/ to generate a runnable jar package. Then, run java -jar collector.jar func_name interval result_dir_name to collect metrics of corresponding functions under their solo-run. Run models/collector/start.sh , and it will create a csv file that stores the metrics under co-locating. You can edit start.sh to set the QPS of LS workloads, edit models/collector/get_ml_data.py to set the amount of data to be collected, and edit models/collector/runBEPara.py to configure tasks that co-locate with the LS of social network.
+Compile gray/gsightCollector/ to generate a runnable jar package. Then, run
+```bash
+java -jar collector.jar func_name interval result_dir_name 
+```
+to collect metrics of corresponding functions under their solo-run. Run 
+```bash
+./models/collector/start.sh
+```
+and it will create a csv file that stores the metrics under co-locating. You can edit start.sh to set the QPS of LS workloads, edit models/collector/get_ml_data.py to set the amount of data to be collected, and edit models/collector/runBEPara.py to configure tasks that co-locate with the LS of social network.
 
 ## Model Training
-The initial training dataset is in models/algorithm/data/. Run models/algorithm/RFR_model_training.py.py , and it will create a file named "RFR" to store the RM model, and csv file to store importance of the metrics.
-Run models/algorithm/plots_imports.py to review the impurity based importance of the 16 metrics, and run models/algorithm/plots_imports_all.py to show the importance of 16 metrics under all combinations of workload and server.
+The initial training dataset is in models/algorithm/data/. Run 
+```python
+python models/algorithm/RFR_model_training.py
+```
+and it will create a file named "RFR" to store the RM model, and csv file to store importance of the metrics.
+Run 
+```python
+python models/algorithm/plots_imports.py
+```
+to review the impurity based importance of the 16 metrics, and run 
+```python
+python models/algorithm/plots_imports_all.py
+```
+to show the importance of 16 metrics under all combinations of workload and server.
 
 ## Scheduler