S y n t h e s i z i n g P l a u s i b l e I n - - PowerPoint PPT Presentation

TU Wien, Vienna Austria Distributed Systems Group https://dsg.tuwien.ac.at

T h

• ma

s R a u s c h @ t h r a u a t Clemens Lachner Pantelis A. Frangoudis Philipp Raith Schahram Dustdar

S y n t h e s i z i n g P l a u s i b l e I n f r a s t r u c t u r e C

• n

fi g u r a t i

• n

s f

• r

E v a l u a t i n g E d g e C

• mp

u t i n g S y s t e ms

HotEdge’20, June 25–26, 2020

2

USENIX HotEdge’19

3

4

Edge systems evaluation tools

iFogSim

Gupta et al. iFogSim: A toolkit for modeling and simulation of resource management techniques in the Internet of Things, Edge and Fog computing environments

Mayer et al. EmuFog: Extensible and scalable emulation of large-scale fog computing infrastructures

5

What are representative infrastructure configurations for edge systems?

6

Industrial IoT Urban Sensing Multi-Access Edge Computing Vehicular Networking

Chen et al. Edge Computing in IoT-Based Manufacturing Eiza et al. Investigation of routing reliability of vehicular ad hoc networks Fadahuns et al. Locality sensitive request distribution for fog and cloud servers Array of Things, Huawei PoleStar

Details in the paper!

7

City-Scale Edge Computing Fabric

https://arrayofthings.github.io

8

Edge Topology Synthesizer

9

Conceptual model

Nodes Links

I

• T

C

• m

p u t e C e l l S h a r e d L i n k C e l l C l

• u

d l e t L A N C e l l

U p / D

• w

n l i n k B a c k h a u l I n t e r n e t N

• d

e L i n k C e l l C

• n

n e c t i

• n

L a t e n c y : . 5 m s B W: 3 M B / s

Industrial IoT Scenario Host

10

Parameterized cell synthesis

 Degree distributions? Probably not.  Cell (cluster) size  Heterogeneity:

How different are the devices in the cell?

 Density:

How many nodes are in this cell compared to others?

 ? What others would be useful

11

GeoCell Density

https://www.senderkataster.at/karte 25 50 75 100 125 150 175 Number of mobile base stations 0.00 0.01 Density ( = 0.47, s=73.02)

2 4 6 8 10 12 14 Number of AoT nodes 0.0 0.2 Density lognorm( = 0.82, s=2.02)

12

Programmatic creation of topologies

topology = Topology() aot_node = IoTComputeBox(nodes=[nodes.rpi3, nodes.rpi3]) neighborhood = lambda size: SharedLinkCell( nodes=[ [aot_node] * size, IoTComputeBox([nodes.nuc] + ([nodes.tx2] * size * 2)) ], shared_bandwidth=500, backhaul=MobileConnection('internet_chix')) city = GeoCell( 5, nodes=[neighborhood], density=lognorm((0.82, 2.02))) cloudlet = Cloudlet( 5, 2, backhaul=FiberToExchange('internet_chix')) topology.add(city) topology.add(cloudlet)

(= networkx graph) GraphML GML JSON YAML ...

13

Interconnecting topologies

Internet Backbone Graph (static) Application Graph (generated)

SYD FRA US East 352 ms 90 ms 206 ms

Different backhaul graphs:

• Internet exchanges
• AWS regions

Cloudlet(backhaul = FiberToExchange(‘aws_us_east’)) GeoCell(..., backhaul = FiberToExchange(‘aws_syd’))

14

Edg dge Networ work WLA LAN Ce Cell Cl Clou

• udl

dlet Up/ p/Down Downlink Up/Down Downlink

Visualization by Reinhold Preiner using TAM: https://github.com/rpreiner/tam Simulation data from faas-sim: https://github.com/edgerun/faas-sim

15

I

• T

C

• m

p u t e C e l l S h a r e d L i n k C e l l C l

• u

d l e t L A N C e l l

U p / D

• w

n l i n k B a c k h a u l I n t e r n e t N

• d

e L i n k C e l l C

• n

n e c t i

• n

L a t e n c y : . 5 m s B W: 3 M B / s

github.com/edgerun/ether

Dipl.-Ing. (MSc), BSc

Thomas Rausch

Research Assistant TU Wien Information Systems Engineering Argentinierstrasse 8-194-02, Vienna, Austria T: +43 1 58801-184838 E: trausch@dsg.tuwien.ac.at https://dsg.tuwien.ac.at/staff/trausch