![Page 1: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/1.jpg)
Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models
Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia
Chair of Software EngineeringUniversity of Würzburg
http://se.informatik.uni-wuerzburg.de/
SIMUtools2015, Athens, Greece, 25.08.2015
![Page 2: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/2.jpg)
Motivation
2 Piotr Rygielski
(dst_IP>*.*.*.128) ? port1 : port0;
(src_TCP==80 && src_TCP==443) ? port1 : port0;
What if…
What if…
What if…
Current performance known – monitoring.
Goal: predict performance after a change.
![Page 3: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/3.jpg)
Research Gap
3 Piotr Rygielski
End-to-end performance analysis not detailed enough
Existing network models too coarse or too fine grained
Other approaches focus only on selected technologies/protocols
Flexibility in modeling is missing
Black-box models Detailed simulations
Time overhead
Accuracy
Model
Model
Model
Model
![Page 4: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/4.jpg)
Approach
4 Piotr Rygielski
Real network
Model extraction
Model transformation(s
)
Descriptive model
Performance model(s)
![Page 5: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/5.jpg)
Approach
5 Piotr Rygielski
Real network
DNI meta model (modeling language)
Structure model
Traffic model
Configuration model
Model-to-model transformations
to QN to OMNeT++
to QPN to ns3
to formulas other...
Performance models
singlemodel
script
![Page 6: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/6.jpg)
Models and Transformations
6 Piotr Rygielski
![Page 7: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/7.jpg)
miniDNI Meta-Model
7 Piotr Rygielski
When not enough data to build full DNI instance
Very coarse-granular modeling
![Page 8: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/8.jpg)
DNI Meta-Model
Structure model Traffic model Configuration model
SoftwareComponent
NetworkInterface
Link
PerformanceDescriptions
Node
TrafficSource
Workload
Flow
Start Stop
Wait Transmit
Loop Sequence
Route
ProtocolStack
NetworkProtocol
DNI Meta-Model (short)
8 Piotr Rygielski
![Page 9: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/9.jpg)
Transformation mDNI-to-QPN
9 Piotr Rygielski
QPN model of a network node, e.g., Switch, Server (mDNI)
Aspects: None, Generator, Receiver, Traversal
![Page 10: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/10.jpg)
Transformation mDNI-to-QPN
10 Piotr Rygielski
QPN model of a network link (mDNI)
Delays from Interfaces and links integrated in queueing place
![Page 11: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/11.jpg)
Transformation mDNI-to-QPN
11 Piotr Rygielski
![Page 12: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/12.jpg)
Transformations - comparison
12 Piotr Rygielski
![Page 13: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/13.jpg)
Traffic Management System
GPS Sensors
Traffic Light
Sensors
http://www.cl.cam.ac.uk/research/time/
Induction Loops
Traffic Cameras
Case study – SBUS/PIRATES
13 Piotr Rygielski
![Page 14: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/14.jpg)
Case study – SBUS/PIRATES
Event Bus
Bus Sensors
TrafficControl
LicensePlate
RecognitionCamCam
Speeding
Toll
LocationBus
Proximity
14 Piotr Rygielski
![Page 15: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/15.jpg)
Case study – SBUS/PIRATES
15 Piotr Rygielski
![Page 16: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/16.jpg)
Model Calibration
16 Piotr Rygielski
![Page 17: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/17.jpg)
Experiment - Hardware
17 Piotr Rygielski
![Page 18: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/18.jpg)
Results – Prediction Accuracy
18 Piotr Rygielski Motivation & Approach DNI & Transformations Current Focus Planning
![Page 19: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/19.jpg)
Results – Simulation Time
19 Piotr Rygielski Motivation & Approach DNI & Transformations Current Focus Planning
Dumbbell topology
![Page 20: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/20.jpg)
Results – Simulation Time
20 Piotr Rygielski Motivation & Approach DNI & Transformations Current Focus Planning
Dumbbell topology
![Page 21: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/21.jpg)
Conclusions
21 Piotr Rygielski Motivation & Approach DNI & Transformations Current Focus Planning
Automatically generated three predictive models
Prediction errors up to 18% for DNI (fully automatic process)
miniDNI-QPN: accuracy loss (~4%) with speedup up to 300x
Support for network virtualization in DNI (SDN planned)
Model calibration is important. Modeling support tools needed
![Page 22: Flexible Performance Prediction of Data Center Networks using Automatically Generated Simulation Models Piotr Rygielski, Samuel Kounev, Phuoc Tran-Gia](https://reader035.vdocuments.net/reader035/viewer/2022081519/56649e425503460f94b34c33/html5/thumbnails/22.jpg)
Thank You!
http://se.informatik.uni-wuerzburg.de
Code & more info:
http://go.uni-wuerzburg.de/aux