Você está na página 1de 31

ALEVIN Software User Guide

Authors:
Mchae Due and Dane Schosser (Unversty of Wuerzburg)
|uan Fepe Botero and Xaver Hessebach (Unverstat de Cataunya)
Andreas Fscher (Unversty of Passau)
Table of Contents
1Introducton...................................................................................................2
2Instaaton Requrements..............................................................................3
2.1GLPK LP sover framework nstaaton....................................................3
2.1.1Unpackng the dstrbuton fe.........................................................4
2.1.2Confgurng the package..................................................................4
2.1.3Compng the package.....................................................................4
2.1.4Check the package...........................................................................4
2.1.5Insta the package...........................................................................4
3Edtng............................................................................................................6
3.1Create, Edt, Import, Export and Manage Scenaros...............................6
3.1.1Scenaro Creaton From Scratch.......................................................6
3.1.2Scenaro Edtng...............................................................................6
3.1.3Edtng Nodes and Lnks...................................................................8
3.1.4Scenaro Import and Export..............................................................9
3.1.5Scenaro Generaton.......................................................................10
4Currenty Supported Agorthms..................................................................15
4.1Agorthms bref descrpton..................................................................15
4.1.1Vrtua Node Mappng Agorthms...................................................15
4.1.2Vrtua Lnk Mappng Agorthms.....................................................16
4.2Agorthms and Mappngs......................................................................17
4.2.1Mappngs........................................................................................17
4.2.2Impemented Agorthms................................................................18
5GUI Features................................................................................................25
5.1The Graph Pane....................................................................................25
5.1.1Move Nodes....................................................................................25
5.1.2Zoom n and out.............................................................................25
5.2The Seecton Pane...............................................................................25
5.3The Mappng Pane................................................................................26
5.4The Consoe..........................................................................................27
5.5Menu Optons........................................................................................28
5.5.1Fe Menu........................................................................................28
5.5.2Vew Menu......................................................................................28
5.5.3Generators Menu............................................................................29
5.5.4Agorthms Menu............................................................................29
5.5.5Metrcs Menu..................................................................................30
1 Introduction
The focus n the deveopment of ALEVIN s on moduarty and effcent handng
of arbtrary parameters for resources and demands, and the support of
ntegraton of new and exstng vrtua network embeddng agorthms and
evauaton metrcs.
For patform ndependency, ALEVIN s wrtten n |ava. ALEVIN's graphical user
interface (GUI) and mut-ayer vsuazaton component s based on the
MuLaVTo pro|ect MuLaVTo (http://muavto.sf.net) whch enabes us to
vsuaze and hande the substrate and arbtrary vrtua networks as drected
graphs.
The next fgure depcts the archtecture of ALEVIN and hghghts the moduar
nteracton of parameters for substrate as we as vrtua networks, agorthms,
and evauaton.
ALEVIN provdes the abty to ustrate the depoyment of resources n the
substrate network and demands n an arbtrary number of vrtua networks as
we as the mappng of demands on resources cacuated by a VNE agorthm.
Moreover, ALEVIN can be used to create VNE scenaros as we as mport and
export them usng an XML-based exchange format.
ALEVIN s competey moduar regardng the addton of new parameters to the
VNE mode. Usng the |ava visitor design pattern n a sophstcated way, we are
abe to avod any casts to concrete demand/resource casses. Thus, the
number of parameters s not performance-reevant and a convenent
mpementaton of arbtrary parameters s possbe.
To ncrease ALEVIN's moduarty and to make t a fexbe and extensbe
patform to compare exstng and upcomng agorthms, the mpementaton of
agorthms s kept ndependent of the resource/demand mpementaton. To
that end, a smpe nterface s provded defnng the rough structure of an
agorthm and connectng ts output to the GUI as ustrated n the prevous
pcture.
2 Installation e!uire"ents
Beng mpemented n |ava, Aevn s patform ndependent and shoud work on
MS Wndows, Mac OS X and Lnux. For proper operaton of Aevn, however, the
foowng software packages have to be nstaed:
SUN |ava |DK verson 1.5 or ater (recommended: verson 6, update 21 or
ater). Do not use other Runtme Envronments, as they are known to create
probems!
The GNU Lnear Programmng Kt (GLPK): http://www.gnu.org/software/gpk/ -
ths s used by some agorthms.
2#1 GL$% L$ sol&er fra"ewor' installation
The GLPK (GNU Lnear Programmng Kt) package s ntended for sovng arge-
scae near programmng (LP), mxed nteger programmng (MIP), and other
reated probems. It s a set of routnes wrtten n ANSI C and organzed n the
form of a caabe brary (http://www.gnu.org/software/gpk/).
Pro|ect GLPK for |ava devers a |ava anguage bndng
(http://sourceforge.net/pro|ects/gpk-|ava/).
Some agorthms of ALEVIN use near programmng to reaze the VNE. ALEVIN
mpementaton rees on GLPK to sove the VNE node and nk mappng stages,
when they requre Lnear Programmng. As ALEVIN s deveoped n |ava, we
have use the |ava bdng nterface provded n
http://sourceforge.net/pro|ects/gpk-|ava/.
The nstaaton nstructons are paced n the /doc foder of the gpk-|ava
software. The fe "gpk-|ava.pdf" expans how the nstaaton must be done.
However, here a step by step expanaton w be provded for each patform:
1. Wndows: The GLPK for |ava |NI brary can be comped from source code.
The bud and make fes are n drectory w32 for 32 bt Wndows and n w64
for 64 bt Wndows. The name of the created brary s gpk 4 45 |ava.d for
revson 4.45.
eco""ended: A precomped verson of GLPK for |ava s provded at
http://wngpk.sourceforge.net
The brary (the name of the brary s gpk_4_45_|ava.d) has to be n the
search path for bnares, t can be found n the w64 or w32 foder (dependng of
your computer archtecture) of the percomped verson
(http://wngpk.sourceforge.net). Ether copy the brary to a drectory that s
aready n the path (e.g. C:\wndows\system32) or update the path n the
system settngs of Wndows. The |ar fe, needed aso for the executon of
ALEVIN s stored aso n the same w64 or w32 foder.
The brary has to be n the CLASSPATH. Update the casspath n the system
settngs of Wndows or specfy the casspath upon nvocaton of the appcaton,
e.g. |ava -casspath ./gpk-|ava.|ar;. MyAppcaton
1. Lnux: To nsta the LINUX verson, the orgna GLPK brary
(http://www.gnu.org/software/gpk/) must be comped n frst pace. The
GLPK for |ava |NI brary can be comped from source code. The foowng
nstructons are aso contaned n the fe INSTALL provded n the source
dstrbuton (http://sourceforge.net/pro|ects/gpk-|ava/):
2#1#1 Un(ac'in) the distribution file
Copy the dstrbuton fe to a workng drectory. Check the MD5 checksum wth
the foowng command:
md5sum glpk-java-X.Y.tar.gz
Unpack the archve wth the foowng command:
tar -xzf glpk-java-X.Y.tar.gz
Now change to the new drecotry gpk-|ava-X.Y
2#1#2 Confi)urin) the (ac'a)e
Open swg/Makefe n a text edtor, e.g. wth the foowng command
vi swig/Makefile
Ad|ust the nstaaton path "prefx".
Ad|ust the ncude path for gpk.h.
Ad|ust the verson nformaton concernng GLPK.
Save the text fe.
On Mac OS X |n.h s n the foowng path:
/System/Library/ramew!rks/"ava#M.framew!rk/$eaders
2#1#* Co"(ilin) the (ac'a)e
To remove a fes from pror compng use the command
make %lea&
The package s comped wth the command
make
2#1#+ Chec' the (ac'a)e
To check f everythng s but correcty use the command
make %'e%k
2#1#, Install the (ac'a)e
To nsta the package you must be root or a suodoer. As sudoer use the
command
sud! make i&stall
1. Mac Os: The process to nsta GLPK-|ava n MacOS s the same as t s n
nux. Remember that on Mac OS X |n.h s n the foowng path:
/System/Library/ramew!rks/"ava#M.framew!rk/$eaders
* Editin)
*#1 Create- Edit- I"(ort- E.(ort and /ana)e Scenarios
*#1#1 Scenario Creation 0ro" Scratch
To create a new, empty scenaro seect "New empty scenaro / ayers" from the
Fe Menu or use the "Ctr + N" Hotkey. In the next step, you w be promted
for the number of vrtua networks of the scenaro to be created:
A scenaro consstng of a substrate network and the desred number of vrtua
networks s created and shown n the graph pane. A created ayers, substrate
or vrtua w be empty. To add nodes and nks, foow the steps descrbed n
the Scenaro Edtng secton beow.
*#1#2 Scenario Editin)
Addin) Networ's
To add vrtua networks seect "New empty scenaro / networks" from the Fe
Menu. Next you w be promted for the number of vrtua networks to add:
The desred number of vrtua networks s added to the graph. If we add two
more networks to our exempary scenaro, the graph w be as foows:
Note that the new networks are e"(t1.
Creatin) Nodes
A node can be created n an exstng network (substrate or vrtua network), at
a desred poston. To create a node, rght-cck at the desred poston and
seect "Create Node" from the pop-up menu:
Next you need to add a constrant to the node to be created:
and set the parameter(s) for the seected constrant:
A constrant s re!uired for the node to be created.
After foowng ths steps, the node w be created at the desred poston and
shown n the graph pane.
Creatin) Lin's
To create a nk seect two nodes and rght-cck on the ayer they beong to.
You then have to choose source and destnaton of the nk to be created. To
make thngs smpe, the two aternatves are dspayed n the pop-up menu:
After the desred nk s seected from the pop-up menu you need to add a
constrant to the nk, for t to be created. For ths purpose, a daog as
descrbed n the Creatng Nodes secton w be dspayed. As n the case of
nodes, a constrant s re!uired for a nk to be created. After addng a
constrant, the nk w be created and shown n the graph pane.
*#1#* Editin) Nodes and Lin's
To edt a node or a nk, rght-cck on t. You have the foowng optons:
Add Constraint
If you choose to add a constrant you w need to seect the constrant type to
add and set t's parameter(s), as descrbed n the Creatng Nodes secton. Note
that each node / nk can ony have a snge resource of a gven type.
Edit Constraint
If you seect the edt constrant opton, you can edt the parameter(s) of the
node's / nk's constrants. Frst you need to seect the constrant to be edted:
then you can set the new vaues of the parameter(s):
e"o&e Constraint
To remove a constrant, seect t as descrbed n the Edt Constrant secton
above and cck on the "Remove"-button.
2elete Node 3 Lin'
If you seect ths opton, the network entty aong wth ts constrants and
mappngs w be deeted.
If you deete a node, a ts ncdent nks w be deeted as we.
*#1#+ Scenario I"(ort and E.(ort
Scenario I"(ort
To mport a scenaro seect "Import" from the Fe Menu or use the "Ctr + O"
Hotkey. Next, seect the fe to mport from:
Import file chooser
The mported scenaro w then be dspayed n the graph pane of the GUI. For
our exempary scenaro ths ooks as foows:
Scenario E.(ort
To export the current scenaro, seect "Export" from the Fe Menu or use the
"Ctr + S" hotkey. You w then need to choose the fe to export to and seect
"Save". The Scenaro, ncudng constrants and mappngs w be saved as
XML-data to the fe specfed.
*#1#, Scenario Generation
Networ' To(olo)1 Generation
ALEVIN provdes a random scenaro generator, the scenaro wzard. For the
scenaro generaton the number of nodes n each network s set and the nks
are added randomy usng a Waxman generator. It creates edges wthn a 11
square wth probabty , where , , s the Eucdan dstance between vertex and
vertex , and s the maxmum dstance between any two nodes. An ncrease n
the parameter apha ncreases the probabty of edges between any nodes n
the graph, whe an ncrease n beta yeds a arger rato of ong edges to short
edges.
For more nformaton on the waxman generator pease refer to
(arti%le)*axma&++,
aut'!r - )*axma&, .er&ard M./,
title - )0!uti&g !f Multip!i&t 1!&&e%ti!&s/,
j!ur&al - )2333 "!ur&al !f Sele%ted 4reas i& 1!mmu&i%ati!&/,
pages - )5657--5688/,
&umber - )9/,
v!lume - )6/,
m!&t' - de%,
year - )59++/,
/
Pror to settng any parameters, the scenaro wzard ooks as foows.
Exemplary scenario import
To generate a new scenaro the foowng parameters must be set
For the substrate network
The number of substrate nodes.
The vaue of the apha and beta parameters.
The number of vrtua networks.
The defaut number of vrtua nodes per network.
For each vrtua network
The number of nodes, f dfferent from the defaut.
The vaue of the apha and beta parameters.
To set parameters for a specfc vrtua network, |ust cck on the respectve
tabe row. If the mouse stands st over one of the tabe coumns, a short nfo
about the parameter s dspayed, as shown n the graphc beow.
After settng the parameters as needed, the aspect of the scenaro wzard
changes sghty.

Constraints Generation
Generatng constrants s possbe usng the constrants generator.
It conssts of two sectons, one for the resources and one for the demands. The
most mportant component of the constrants generator s the tabe used for
seectng the constrants to add and settng the requred parameters. The
resources pane contans one such tabe, whe the demands pane cossts of
one tabe for each vrtua network, the tabes beng packed n dfferent tabs.
For constrants generaton, a constrant needs to be seected for the desred
network (substrate or vruta) and the correspondng parameters set, f needed.
Ths s done usng the tabes descrbed above. Currenty ony CPU and
bandwdth constrants have parameters to set. A other constrants need no
further user acton to be generated correcty. Demand genraton can be done
ndvduay on a per vrtua network bass.
After seectng some constrants and settng the needed parameters, the
constrants generator w ook as seen beow. Notce that, as descrbed above,
Constraints Generator
settng addtona parameters s not requred for a resource/demand types.
e"o&in) All Constraints
It s aso possbe to remove a constrants of the current scenaro, for exampe
n order to regenerate them.
Constraints Generator with selected constraints and
parameter set
+ Currentl1 Su((orted Al)orith"s
+#1 Al)orith"s brief descri(tion
In ths secton, a bref descrpton of the vrtua node and nk mappng exstng
agorthms s gven. The expanaton w separate the vrtua node mappng and
nk mappng.
+#1#1 Virtual Node /a((in) Al)orith"s
Greed1 Stress A((roach 4GS5
To understand the ob|ectve of ths approach. Fgure 2 defnes the term stress.

0i)ure 2
The numbers above the nks an nodes are the respectve stresses. The
ob|ectve of ths approach s to mnmze the baanced stress (wegthed sum of
node and nk stresses) n the network. To reach ths ob|ectve, the vrtua node
mappng stage s performed by assgnng the nodes wth greater degree n
vrtua network to the nodes wth greater potential (for each node n the SN,
the potenta s the mutpcaton of the node stress by the sum of ad|acent nk
stresses) n the substrate network.
Greed1 A&ailable esources 4GAS5
Ths vrtua node mappng approach s very smar to the GS. The avaabe
resources concept s defned n Fgure 3.

0i)ure *
The vrtua node mappng GAS s a greedy agorthm. Nodes wth greater
demand are assgned wth greater avaabe resources.
/i.ed Inte)er $ro)ra""in) Solution
Substrate Network Graph s augmented creatng Meta-nodes (representng the
vrtua nodes) and meta-edges (|onng meta-nodes wth the canddate nodes to
be mapped n the SN), they are added to the SN graph as ndcated n Fgure 4.

0i)ure +
In fgure 4, a w be mapped to ether A or C. (A and C are the canddates
nodes to map a). After the foowng steps, the map s decded.
1. The probem s soved usng near programmng (LP) -> The reaxed BIP
formuaton s soved.
2. The LP souton contans ratona vaue for each of the meta-edges (|onng
the meta-nodes wth nodes n the substrate network). A randomzed or
determnstc roundng s performed, among the meta-edges of each meta-
node, to choose one of them; the SN connected to the chosen meta-edge s
then chosen as the mapped node n the SN.
+#1#2 Virtual Lin' /a((in) Al)orith"s
It s very mportant to state that the vrtua nk mappng stage starts after
vrtua node mappng.
%6Shortest $aths Al)orith" 4%S$5
Ths agorthm cacuates, for each vrtua nk, the k-shortest paths from one
mapped vrtua node to the other end (n the substrate network). The mappng
s performed by assgnng, to each vrtua nk, the frst shortest path that
accompshes the bandwdth and CPU demands.
/ulti6$ath Al)orith" 4/$5
Ths agorthm soves a LP formuaton of the vrtua nk mappng stage (the
same formuaton as the reaxed BIP). Ths formuaton s equvaent to the
mut-commodty fow probem and s soved usng the tradtona LP methods
(SIMPLEX, IPM, etc.) A exampe of ths approach s shown n Fgure 6.

0i)ure 7
oundin) /ulti6$ath 4/$5
The probem s formuated as the mut-commodty fow probem and s soved
usng optma near programmng agorthms (SIMPLEX, IPM, etc.). Then, the
mut-path souton, each path w have a ratona percentage of the demand, s
rounded (n a determnstc or randomzed way) and |ust one drected path s
used to map each vrtua nk.
Sub)ra(h iso"or(his" detection heuristic 4SI25
The mappng n nodes and nks s done smutaneousy by tryng to fnd a
subgraph nsde the substrate network beng somorphc to the vrtua network
request. Graph somorphsm s expaned n fgure 5.
+#2 Al)orith"s and /a((in)s
+#2#1 /a((in)s
Co"(ute /a((in)s
To compute mappngs for the current scenaro you need to run an agorthm
from the Agorthms Menu. To do so, |ust seect t from the menu. Agorthms
can not be run on scenaros that aready have mappngs.
To better understand the nput parameters requested by each agorthm, see a
bref and detaed expanaton of the agorthms.
Whe an agorthm s runnng, a progress daog s shown:
+#2#2 I"(le"ented Al)orith"s
Unt now, there are fve agorthms avaabe:
Smpe D|kstra: Test vrtua nk mappng agorthm (wth node mappng
assumed as performed) mappng each vrtua nk to the shortest path.
Greedy Avaabe wth K-Shortest Path: Vrtua node and nk mappngs for
each VNR are performed separatey. Node mappng s performed usng a
greedy agorthm that maps vrtua nodes wth hgher revenues to SN nodes
wth hgher "avaabe resources" (GAS). Lnk mappng s accompshed by
mappng each vrtua nk to the shortest-path, accompshng capacty
constrants, between the correspondng mapped nodes n the SN.
Greedy Avaabe wth Path Spttng: Vrtua node and nk mappngs for each
VNR are performed separatey. Node mappng s aso performed wtha greedy
avaabe resources (GAS). Lnk mappng s accompshed wth a mut- path
(MP) optma souton that avods the NP-competeness of the probem.
Coordnated Node and nk mappng wth Path Spttng: The node mappng
stage s performed by defnng an augmented graph over the substrate
network; ntroducng a set of meta-nodes, one per vrtua node, each
connected to a custer of canddate SN nodes obeyng ocaton and capacty
constrants. The agorthm soves the VNE probem by usng a Mxed Integer
Programmng (MIP). Its ob|ectve s to mnmze the cost, .e. the weghted
sum of the bandwdth and CPU aocated n the SN nks to fuf VNR
demands, of embeddng a VNR. To avod the NP-competeness of the MIP, ts
near programmng reaxaton s soved, and the obtaned souton s rounded
n two ways: determnstcay or randomy. The nk mappng stage s
performed usng a Mut-Path approach (MP).
Coordnated Node and nk mappng wth k-Shortest Path: The same vrtua
node mappng stage s performed n the same way (MIP) as the prevousy
descrbed agorthms. The nk mappng stage s performed usng a K-Shortest
paths approach.
In(ut $ara"eters
Each agorthm has dfferent nput parameters. For the mpemented
agorthms, the parameters are as foows:
Greed1 A&ailable esources 4Node /a((in)5 8 '6Shortest (aths
4Lin' /a((in)5: The ony parameter needed n ths agorthm s the number
of shortest paths to be cacuated (the parameter k). An nput wzard s
Algorithms progress dialog
shown:
The parameters presented n the agorthm's wzard are the foowng:
': Is the number of shortest-paths (n terms of hops) used n the agorthms for
the vrtua nk mappng stage.
2istance Considered: The defaut vaue of ths booean vaue s "No". Ths s
not a vaue presented n the orgna agorthm. Ths vaue s ncuded to
provde the possbty of comparng ths agorthm wth subsequent proposas.
If the vaue s chosen to "No" the norma behavor of the agorthm s
presented. The vaue refers to the maxmum dstance that a substrate node
can be from a vrtua node to be consdered a canddate node to be mapped.
Ths vaue must be between 0 and 100 (the area of the space where each
network s potted s 100100 ).
Node 9&erload: As the prevous parameter the node overoad s not a vaue
presented n the orgna agorthm. Current agorthms do not aow that more
than one vrtua node, beongng to a vrtua network, are mapped n one
substrate node. Ths vaue modfes the agorthm to avod ths constrant.
Agan, If the "No" opton s chosen, the agorthm w have ts norma behavor.
Greed1 A&ailable esources with $ath S(littin) 4Lin' /a((in)5: To
reaze the nk mappng agorthm t s needed to provde the weght that
user gves to the CPU and BW.
The parameters presented n the agorthm's wzard are the foowng:
C(u and :andwidth ;ei)ths: Ths vaues aow to gve more revenue vaue
to Cpu or Bw when the vrtua network embeddng s beng performed. .e., f
Cpu weght s hgher than bandwdth weght, the agorthm w perform a
mappng tryng to mnmze the CPU consume n the substrate network.
2istance Considered: It has the same meanng than n the prevous
agorthm
Node 9&erload: It has the same meanng than n the prevous agorthm
Prevous two agorthms are proposed n:
(arti%le)Yu:+,
aut'!r - )Yu, Mi&la& a&d Yi, Yu&g a&d 0exf!rd, "e&&ifer a&d 1'ia&g, Mu&g/,
title - )0et'i&ki&g virtual &etw!rk embeddi&g; substrate supp!rt f!r pat'
splitti&g a&d migrati!&/,
j!ur&al - )S2<1=MM 1!mput. 1!mmu&. 0ev./,
v!lume - )>+/,
issue - )8/,
m!&t' - )Mar%'/,
year - )8::+/,
iss& - ):5?6-?+>>/,
pages - )57--89/,
&umpages - )5>/,
url - )'ttp;//d!i.a%m.!rg/5:.55?5/5>557>?.5>557>7/,
d!i - )'ttp;//d!i.a%m.!rg/5:.55?5/5>557>?.5>557>7/,
a%mid - )5>557>7/,
publis'er - )41M/,
address - )@ew Y!rk, @Y, AS4/,
keyw!rds - )&etw!rk virtualizati!&, !ptimizati!&, pat' migrati!&, pat'
splitti&g, virtual &etw!rk embeddi&g/,
/
Coordinated Node 4Node /a((in)5 and Lin' /a((in) with '6Shortest
$aths 4Lin' /a((in)5: Ths souton uses the heurstc souton of the Mxed
Integer Programmng approach n the node mappng phase and the k-
shortest path approach n the nk mappng phase.
The parameters presented n the agorthm's wzard are the foowng:
': Is the number of shortest-paths (n terms of hops) used n the agorthms for
the vrtua nk mappng stage.
C(u and :andwidth ;ei)ths: Ths vaues aow to gve more revenue vaue
to Cpu or Bw when the vrtua network embeddng s beng performed. .e., f
Cpu weght s hgher than bandwdth weght, the agorthm w perform a
mappng tryng to mnmze the CPU consume n the substrate network.
2istance Considered: It refers to the maxmum dstance that a substrate
node can be from a vrtua node to be consdered a canddate node to be
mapped. Ths vaue must be between 0 and 100 (the area of the space where
each network s potted s 100100 ).
Node 9&erload: The node overoad s not a vaue presented n the orgna
agorthm. Current agorthms do not aow that more than one vrtua node,
beongng to a vrtua network, are mapped n one substrate node. Ths vaue
modfes the agorthm to avod ths constrant. Agan, If the "No" opton s
chosen, the agorthm w have ts norma behavor.
Al)orith" T1(e: Ths agorthm work, n the node mappng phase, by sovng
a reaxed verson of the NP-Compete Unspttabe Fow Probem (UFP). After
that, the reaxed souton s rounded by two method (determnstc or
randomzed). Ths parameter contans the roundng type.
Coordinated Node 4Node /a((in)5 and Lin' /a((in) with $ath
S(littin) 4Lin' /a((in)5: Ths souton uses the heurstc souton of the
Mxed Integer Programmng approach n the node mappng phase and a path
spttng approach n the nk mappng phase.
The parameters presented n the agorthm's wzard are the foowng:
C(u and :andwidth ;ei)ths: Ths vaues aow to gve more revenue vaue
to Cpu or Bw when the vrtua network embeddng s beng performed. .e., f
Cpu weght s hgher than bandwdth weght, the agorthm w perform a
mappng tryng to mnmze the CPU consume n the substrate network.
2istance Considered: It refers to the maxmum dstance that a substrate
node can be from a vrtua node to be consdered a canddate node to be
mapped. Ths vaue must be between 0 and 100 (the area of the space where
each network s potted s 100100 ).
Node 9&erload: The node overoad s not a vaue presented n the orgna
agorthm. Current agorthms do not aow that more than one vrtua node,
beongng to a vrtua network, are mapped n one substrate node. Ths vaue
modfes the agorthm to avod ths constrant. Agan, If the "No" opton s
chosen, the agorthm w have ts norma behavor.
Al)orith" T1(e: Ths agorthm work, n the node mappng phase, by sovng
a reaxed verson of the NP-Compete Unspttabe Fow Probem (UFP). After
that, the reaxed souton s rounded by two method (determnstc or
randomzed). Ths parameter contans the roundng type.
To see more detas of prevous two agorthms, pease refer to:
(2@B0=133C2@<S)1$:9,
aut'!r-)1'!wd'ury, @.M.M.D. a&d 0a'ma&, M.0. a&d .!utaba, 0./,
b!!ktitle-)2@=1=M 8::9, 2333/, title-)#irtual @etw!rk 3mbeddi&g wit'
1!!rdi&ated @!de a&d Li&k Mappi&g/,
year-)8::9/,
m!&t'-)april/,
v!lume-)/,
&umber-)/,
pages-)7+> -795/,
keyw!rds-)2&ter&etE%!!rdi&ated &!deE'euristi%-based alg!rit'msEli&k
mappi&gEmixed i&teger pr!gramEmultiple 'eter!ge&e!us virtual &etw!rksE&etw!rk
virtualizati!&Esubstrate &etw!rk augme&tati!&Evirtual &etw!rk
embeddi&gE2&ter&etEembedded systemsEvirtual ma%'i&esE/,
d!i-)5:.55:9/2@1=M.8::9.5:659+7/,
2SS@-):7?>-566X/,/
Coordinated Node 4Node /a((in)5 and oundin) /ulti(ath 4Lin'
/a((in)5: Ths souton uses the heurstc souton of the Mxed Integer
Programmng approach n the node mappng phase and a new approach n
nk mappng to reaze the vrtua nk mappng usng snge path. Snge path
s reached by roundng the mut-path souton n the same way than n the
prevous proposa (Determnstcay or randomy).
': Is the number of shortest-paths (n terms of hops) used n the agorthms for
the vrtua nk mappng stage n case that rounded paths do not accompsh
the vrtua nk demands.
C(u and :andwidth ;ei)ths: Ths vaues aow to gve more revenue vaue
to Cpu or Bw when the vrtua network embeddng s beng performed. .e., f
Cpu weght s hgher than bandwdth weght, the agorthm w perform a
mappng tryng to mnmze the CPU consume n the substrate network.
2istance Considered: It refers to the maxmum dstance that a substrate
node can be from a vrtua node to be consdered a canddate node to be
mapped. Ths vaue must be between 0 and 100 (the area of the space where
each network s potted s 100100 ).
Node 9&erload: The node overoad s not a vaue presented n the orgna
agorthm. Current agorthms do not aow that more than one vrtua node,
beongng to a vrtua network, are mapped n one substrate node. Ths vaue
modfes the agorthm to avod ths constrant. Agan, If the "No" opton s
chosen, the agorthm w have ts norma behavor.
Al)orith" T1(e: Ths agorthm work, n the node mappng phase, by sovng
a reaxed verson of the NP-Compete Unspttabe Fow Probem (UFP). After
that, the reaxed souton s rounded by two method (determnstc or
randomzed). Ths parameter contans the roundng type.
<idden <o(s Selection
In the next step, you can seect the hdden hop demands that are to be
consdered when computng the mappng. Ths s done usng the Hdden Hops
Daog.
It conssts of a tabe that enabes you to seect the HH Demands to use and set
the factor used for computng them.
9ut(ut $ara"eters
After termnaton some nformaton messages are dspayed n the Consoe. If
the agorthm fnshes normay, the computed mappngs are added to the
scenaro and dspayed n the Mappng Pane.
e"o&e All /a((in)s
To remove a mappngs of the current scenaro, seect the "Remove a
mappngs" opton from the Agorthms Menu. Ths acton w aso free up a
resources n the substrate network.
Hidden Hops Dialog
, GUI 0eatures
,#1 The Gra(h $anel
The graph panel showing our exemplary scenario.
The graph pane s taken from MuLaVTo and s the centra part of the GUI,
where the scenaros are vsuazed. For each ayer of the scenaro, substrate or
vrtua, the graph of the network s dspayed.
The graph pane enabes you to:
,#1#1 /o&e Nodes
To move nodes, smpy drag them wth the mouse. Note that ths acton w
have an effect on the node's coordnates.
,#1#2 =oo" in and out
To zoom n or out hod the "Ctr" key and use the mouse whee.
,#2 The Selection $anel
The seecton pane shows the network enttes currenty seected as we as
ther constrants n a tree structure. Here s an exampe usng our exempary
scenaro. Note that some eements of the seecton tree are not shown because
ther parent eement are not popped up.
,#* The /a((in) $anel
The mappng pane shows the mappngs of the current scenro n a tree
strucure. If no mappngs exst, the vrtua nks w be the eaf eements of the
tree:
If mappngs do exst, they are shown as chd eements of the vrtua nks.
When seectng a vrtua nk from the mappngs tree, t s hghghted aong
wth the substrate nks t s mapped to. Hdden hops are dspayed usng a
dfferent coor (magenta).
election panel example
!apping panel showing only
the virtual lin"s as no
mappings exist
Note that the "Vrtua Network (2)" node s not popped up n ths exampe.
The Mappng pane aso enabes reverse hghghtng. If you seect a substrate
nk, the vrtua network enttes mapped on the nk, or ts ncdent nodes are
hghghted.
,#+ The Console
The Consoe s taken from MuLaVTo and dspays nformaton, debug and error
messages.
The Consoe can be ceared or redrected to the norma termna consoe.
It can aso be cosed and then restored by usng the "Hdden Panes" opton n
the Vews Menu.
!apping panel with mappings and highlighting
!apping panel with mappings and highlighting
,#, /enu 9(tions
,#,#1 0ile /enu
#ile !enu
From the Fe Menu you can seect one of the foowng optons
Creatng a new empty scenaro or addng new ayers to an exstng scenaro
Importng a scenaro
Exportng a scenaro
Cosng the Scenaro
Extng the Appcaton
,#,#2 View /enu
view menu
without hidden
panels
You can check the "Show node abes" or "Show nk abes" optons. Ths w
dspay the abes of nodes / nks n the graph pane.
The "Hdden Panes" menu entry provdes a means to restore cosed or hdden
GUI components. Currenty ony the consoe can be cosed and restored. Ths
functonaty s provded by MuLaVTo and s ony avaabe f there are cosed
GUI components avaabe for restoraton.
view menu with hidden
console
,#,#* Generators /enu
Generators !enu
The Generators Menu provdes the foowng three optons
Generate a scenaro usng the Scenaro Wzard.
Generate constrants for an exstng scenaro.
Remove a constrants of the current scenaro.
,#,#+ Al)orith"s /enu
Algorithms !enu
From the agorthm menu you can ether seect an agorthm to be rum or
remove a mappngs of the current scenaro.
un an Al)orith"
You can run an agorthm n order to create mappngs between demands and
resources.
See the graphcs above for a st of the currenty avaabe agorthms.
An agorthm can ony be run f the current scenaro has no mappngs.
e"o&e all /a((in)s
Ths opton removes a exstng mappngs of the current scenaro and frees up
a resources n the substrate network.
,#,#, /etrics /enu
!etrics !enu
The Metrcs Menu enabes the evauaton of the current scenaro usng one of
the avaabe metrcs. It s generated automatcy usng refecton, to ensure
that a avaabe evuatons are dspayed.
To perform an evauaton on the current scenaro seect t from the menu. The
resut w be dspayed n a daog that aso enabes copyng t. An exape for
the average node stress s shown beow.

!etrics Dialog