Op imizing on haul esou ce alloca ion
o 6G RAN
BSC Se e o Ochoa Doc o al Symposium · May 6-8, 2025 · Ba celona (Spain) · Pos e Session
Dynamic bandwid h alloca ion
Abs ac : As wi eless ne wo ks mo e owa d 6G, Radio Access Ne wo ks (RANs) ace inc easing demands o
highe speed, g ea e bandwid h e iciency, and lowe la ency, pa icula ly in he on haul segmen . Passi e Op ical
Ne wo ks (PONs) o e a cos -e ec i e way o suppo his, bu eme ging new applica ions like AI and ex ended
eali y make scheduling on haul a ic mo e complex. This wo k explo es sma e scheduling me hods o op imize
on haul ne wo k pe o mance, enabling e icien and eliable suppo o u u e 6G applica ions.
Xiang Lu1*, H. Shakespea -Miles1 , and Sima Ba zega 2
In Radio Access Ne wo ks (RANs), he on haul ne wo k
speci ically connec s he dis ibu ed uni (DU) and he adio
uni s (RUs). Passi e Op ical Ne wo ks (PONs) ha e eme ged
as a popula and cos -e ec i e solu ion o suppo on haul
ne wo ks and ha e g ea po en ial o deploying edge
compu ing nodes.As he need o high bandwid h and low
la ency becomes mo e c i ical, he on haul ne wo k becomes
a i al b idge be ween mobile de ices and da a cen e s, whe e
high-pe o mance compu ing (HPC) sys ems a e ypically
deployed.
Resul s
(1) Uni e si a Poli ècnica de Ca alunya (UPC), Ba celona, Spain
(2) Ba celona Supe Compu ing Cen e (BSC), Ba celona, Spain
*e-mail: [email p o ec ed]
Passi e Op ical Ne wo ks (PON)
A PON ypically consis o one Op ical Line Te minal
(OLT) connec ed o mul iple Op ical Ne wo k Uni s
(ONUs) ia a poin - o-mul ipoin opology.
Ca ying ups eam on haul a ic ia PONs
in oduces ex a la ency due o he s a us- epo ing
mechanisms. Fo his, Coope a i e DBA (CO-DBA)
was s anda dized o 4G/5G on haul o educe
la ency.
Howe e , eme ging 6G applica ions, such as AI and
ex ended eali y, p esen new and g ea e challenges
o on haul a ic managemen .
F on haul Ne wo k in RAN
We de eloped a Flexible Queue Managemen Dynamic
Bandwid h Alloca ion (FQM-DBA) algo i hm ha be e
suppo s he di e se la ency equi emen s o ime-sensi i e
se ices. This algo i hm can be e ec i ely applied o
on haul o e PON, capable o mee ing he pe o mance
demands o a ious ime-sensi i e 6G applica ions. whe e
de e minis ic la ency gua an ees and e icien bandwid h
u iliza ion a e essen ial.
Simula ions we e conduc ed using ou Py hon-based simula ion pla o m, conside ing a scena io o a50G-PON
sys em wi h one OLT se ing 64 ONUs. Bo h low (0.3) and high (0.9) ne wo k loads we e e alua ed.
T-CONT T-CONT T-CONT
High p io i y High la ency ole ance
Da a ame
Repo message
OLT
ONU
C ea e
Remo e
Da a ame g ouping
Queue moni o ing
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.2
0.4
0.6
0.8
1.0
PSR
(a) Load
IT 1 (Benchma k 1)
IT 2 (Benchma k 1)
IT 1 (Benchma k 2)
IT 2 (Benchma k 2)
IT 1 (FQM-DBA)
IT 2 (FQM-DBA)
CDF
(b) La ency (ms)
IT 1 (Benchma k 1)
IT 2 (Benchma k 1)
IT 1 (Benchma k 2)
IT 2 (Benchma k 2)
IT 1 (FQM-DBA)
IT 2 (FQM-DBA)
IT 1 la ency limi =0.8 ms
IT 2 la ency limi =1.5 ms
CDF
(c) La ency (ms)
IT 1 (Benchma k 1)
IT 2 (Benchma k 1)
IT 1 (Benchma k 2)
IT 2 (Benchma k 2)
IT 1 (FQM-DBA)
IT 2 (FQM-DBA)
IT 1 la ency limi =0.8 ms
IT 2 la ency limi =1.5 ms
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
0.0
0.2
0.4
0.6
0.8
1.0
PSR
(a) Load
OT 1 (Benchma k 1)
OT 2 (Benchma k 1)
OT 1 (Benchma k 2)
OT 2 (Benchma k 2)
OT 1 (FQM-DBA)
OT 2 (FQM-DBA)
CDF
(b) La ency (ms)
OT 1 (Benchma k 1)
OT 2 (Benchma k 1)
OT 1 (Benchma k 2)
OT 2 (Benchma k 2)
OT 1 (FQM-DBA)
OT 2 (FQM-DBA)
OT 1 ji e limi =0.4 ms
OT 1 la ency limi =1 ms
OT 2 la ency limi =2 ms
OT 2 ji e limi =0.6 ms
CDF
(c) La ency (ms)
OT 1 (Benchma k 1)
OT 2 (Benchma k 1)
OT 1 (Benchma k 2)
OT 2 (Benchma k 2)
OT 1 (FQM-DBA)
OT 2 (FQM-DBA)
OT 1 la ency limi =1 ms
OT 1 ji e limi =0.4 ms
OT 2 la ency limi =2 ms
OT 2 ji e limi =0.6 ms
Two se ice ypes we e conside ed: in- ime se ices equi e a maximum la ency gua an ee, and on- ime se ices
impose limi s on la ency and ji e .
Th ee schemes we e e alua ed: Benchma k 1 (IACG algo i hm + FIFO), Benchma k 2 (IACG algo i hm + S ic P io i y
scheduling), and ou p oposed FQM-DBA algo i hm.
FQM-DBA can e ec i ely suppo bo h ime-sensi i e se ices and signi ican ly ou pe o ms adi ional schemes,
pa icula ly unde high ne wo k load.
