You are on page 1of 14

Net w or k Rout i ng

A maj or c omponent of t he net w or k l ayer


r out i ng pr ot oc ol .
Rout i ng pr ot oc ol s use r out i ng al gor i t hms.
J ob of a r out i ng al gor i t hm: Gi ven a set of
r out er s w i t h l i nk s c onnec t i ng t he r out er s,
f i nd a good pat h f r om t he sour c e t o t he
dest i nat i on.
Model i ng a Net w or k
A net w or k c an be model ed by a gr aph.
- Rout er s/sw i t c hes ar e r epr esent ed by
nodes.
- Physi c al l i nk s bet w een r out er s/sw i t c hes
ar e r epr esent ed by edges.
- At t ac hed c omput er s ar e i gnor ed.
- Eac h edge i s assi gned a w ei ght
r epr esent i ng t he c ost of sendi ng a
pac k et ac r oss t hat l i nk .
The t ot al c ost of a pat h i s t he sum of t he
c ost s of t he edges.
The pr obl em i s t o f i nd t he l east -c ost pat h.
Rout i ng Al gor i t hms
Rout i ng al gor i t hms t hat sol ve a r out i ng
pr obl em ar e based on shor t est -pat h
al gor i t hms.
Tw o c ommon shor t est -pat h al gor i t hms ar e
Di j k st r as Al gor i t hm and t he Bel l man-For d
Al gor i t hm.
Rout i ng al gor i t hms f al l i nt o t w o gener al
c at egor i es.
Li nk -St at e Al gor i t hms
The net w or k t opol ogy and al l l i nk c ost s ar e
k now n.
Ex ampl e: Di j k st r as Al gor i t hm.
Mor e c ompl ex of t he t w o t ypes.
Nodes per f or m i ndependent c omput at i ons.
Used i n Open Shor t est Pat h Fi r st (OSPF)
pr ot oc ol , a pr ot oc ol i nt ended t o r epl ac e
RI P.
Di st anc e-Vec t or Al gor i t hms
Nodes r ec ei ve i nf or mat i on f r om t hei r
di r ec t l y at t ac hed nei ghbor s.
Ex ampl e: Bel l man-For d Al gor i t hm.
Si mpl er of t he t w o t ypes.
May have c onver genc e pr obl ems.
Used i n Rout i ng I nf or mat i on Pr ot oc ol (RI P).
Di j k st r as Al gor i t hm
Named af t er E. W. Di j k st r a.
Fai r l y ef f i c i ent .
I t er at i ve al gor i t hm.
At t he f i r st i t er at i on, t he al gor i t hm f i nds t he
c l osest node f r om t he sour c e node w hi c h
must be a nei ghbor of t he sour c e node.
At t he sec ond i t er at i on, t he al gor i t hm f i nds
t he sec ond-c l osest node f r om t he sour c e
node. Thi s node must be a nei ghbor of
ei t her t he sour c e node or t he c l osest node
f ound i n t he f i r st i t er at i on.
Di j k st r as Al gor i t hm
At t he t hi r d i t er at i on, t he al gor i t hm f i nds
t he t hi r d-c l osest node f r om t he sour c e node.
Thi s node must be a nei ghbor of ei t her t he
sour c e node or one of t he f i r st t w o c l osest
nodes.
The pr oc ess c ont i nues. At t he k
t h
i t er at i on,
t he al gor i t hm f i nds t he f i r st k c l osest nodes
f r om t he sour c e node.
Ex ampl e
The sour c e node i s s = 1.
Ex ampl e
The bot t om ent r y i n eac h D-c ol umn i s t he
mi ni mum c ost t o go f r om t he st ar t node 1
t o t hat node.
Quest i on: How c an you det er mi ne t he pat h
w hi c h gi ves t he mi ni mum c ost t o a
dest i nat i on node?
Answ er : The t abl e not onl y gi ves t he
mi ni mum c ost s. I t al so gi ves t he
pr edec essor node of eac h node al ong a
l east -c ost pat h f r om t he sour c e node. By
k eepi ng t r ac k of t he pr edec essor nodes, w e
c an c onst r uc t a l east -c ost pat h.
Least -Cost Pat h Tr ee
Rout i ng Tabl e f or Sour c e Node 1
Compl ex i t y of Di j k st r as Al gor i t hm
Suppose t her e ar e n nodes not c ount i ng t he
sour c e node.
I n t he f i r st i t er at i on, w e need t o sear c h
t hr ough n nodes t o det er mi ne t he node not
i n N w i t h mi ni mum c ost .
I n t he sec ond i t er at i on, w e need t o c hec k
n1 nodes.
I n t he t hi r d i t er at i on, n2 nodes. And so on.
Compl ex i t y of Di j k st r as Al gor i t hm
The t ot al number of nodes w e need t o
ex ami ne i s
1 + 2 + 3 + + n = n(n+1)/2
Thus, Di j k st r as Al gor i t hm as pr esent ed i s
O(n
2
)
A mor e sophi st i c at ed i mpl ement at i on of t he
sec ond st ep usi ng a heap w oul d f i nd t he
mi ni mum i n l ogar i t hmi c i nst ead of l i near
t i me. Thi s i mpr oves t he per f or manc e t o
O(n l og n)

You might also like