Transcript Powerpoint 2007

A simulation study of GELS (GMPLS-controlled
Ethernet Label Switching) for Ethernet over WAN
Muhammad Saqib Ilyas ([email protected])
School of Science and Engineering
LUMS, Lahore, Pakistan
Co-authors:
Atif Nazir, Fawaz Saleem Bokhari, Zartash Afzal Uzmi (LUMS)
Fahad Dogar (CMU, Pittsburgh)
Adrian Farrel (Old Dog Consulting)
IEEE Globecom 2007
Washington, D.C.
Wednesday, Nov 28, 2007
Work sponsored by:
Siemens Corporate Technology Division
Munich, Germany
Agenda
GMPLS – Background
 Ethernet – Background
 GELS Architecture

◦ GMPLS as the control plane for Ethernet
Simulation Modeling and Setup
 Simulation Results
 Summary and Conclusions

IP Routing
Dest: 150.10.10.1
Dest IP
Next
hop
I’face
…………..
…..
…
150.0.0.0
…..
…
150.10.0.0
.….
…
150.10.10.0
…..
…
…………..
…..
…
…………..
…..
…
Longest prefix
match
Forwarding in MPLS
Label: 10
Label: 13
Label lookup
Labelin
Labelout
I’face
…
…..
…
8
8
…
9
15
…
10
13
…
…
…..
…
…
…..
…
MPLS challenges

Newer devices are capable of switching on the basis of:
◦ Interface (FSC)
◦ Wavelength (LSC)
◦ TDM timeslot

MPLS works with packet switch devices only
◦ Looks at the label and forwards an incoming packet
Incompatibility of MPLS with newer devices

Solution:
◦ Generalize MPLS to GMPLS (RFC 3945)
GMPLS offers a control plane for
devices with ANY data plane
Ethernet
Dominant LAN transport technology
 Speed and reach grew substantially in the
last 25 years
 Very flexible and cost-effective transport


Ethernet is seeing increasing deployment
in service provider networks
Ethernet in the core - challenges

Existing control plane (STP)
◦ Network link utilization – Low
◦ Resilience mechanism – Slow
◦ Rudimentary support for QoS and TE
Spanning
Spanningtree
tree
computed
recomputed
Link failure
GELS

Proposes to use GMPLS control plane for
Ethernet
Bridge
the Ethernet data
plane!
GELS is in draft stages in IETF
 No quantitative performance
comparison available so far

Our work
Simulation based evaluation of GELS
 Rapid STP (RSTP) versus GMPLS

◦ How does old control plane compare with
new control plane?

Considered:
1. Normal network operation
2. Single element failures
Evaluation Criteria
How efficiently can
we use the
network?
Average link
utilization
Normal network
condition
Number of LSPs
placed
Total bandwidth
placed
Evaluation criteria
Failed network
condition
Single link failure
RSTP convergence
time
Restoration
Single node failure
GELS recovery
Protection
How quickly can we
recover from
failure?
GELS
Recovery
Schemes
Evaluation challenges

How to compare contention-based
Ethernet with reservation based GMPLS?
◦ Allow partial placement of LSPs in GMPLS
instead of YES/NO placement
Available:
15
0
Request: 25
Placed: 15
0
GMPLSGMPLS
with Compromised
with CSPF CSPF
LSP not
placed
placed
Bandwidth placed: 0%
60%
Capacity:
100
GELS: Convergence time
Restoration: trest = tsig + tproc + tres + tsw
Reserve new LSP
Switch traffic
onto new LSP
tres: Reservation
delay
tsw: Switching
delay
Protection:
tprot
= tsig + tsw
Compute new LSP
tproc: Processing delay
Potential new
path
Link failure
Ingress
Failure notification
sent to ingress
tsig: Signaling delay
LSP
Egress
Nearest
upstream node
to the failure
Timing parameter values
 tsig(Signaling
delay):
◦ Based on 1ms/200 km link propagation delay
 tproc(Processing
delay):
◦ 5ms
 tres(Reservation
delay):
◦ Based on 1ms/200 km link propagation delay
 tsw(Switching
◦ 1ms
delay):
GELS restoration recovery time
LSP 1
LSP 2
Ingress has lost
multiple LSPs
NearestSequentially
upstream
node for LSP 1
1. Compute
2. Reserve
3. Switch
Sequentially
Or
In parallel
Link failure
Convergence
time is tmax
Convergence
time is tmin
Nearest
Sequentially
upstream node
for LSP 2
Failure signaled
to ingress
Simulation setup - networks
(1)
CopenhagenHelsinki
(1)
Oslo (2)
COST
COST 266:
239: 11
50 nodes
nodes
Stockholm (3)
Glasgow (4)
Belfast (5)
Dublin (7)
Copenhagen (6)
Liverpool (8)
Birmingham (9)
Amsterdam (3)
Amsterdam (11)Hamburg (12) Berlin (13)
London (10)
Brussels (15) Dusseldorf (16)
Leipzig (18)
London (2)
Berlin (4)
Warsaw (14)
Krakow (23)
Brussels (5)
Frankfurt (17)
Prague (22)
Strasbourg (20) Munich (21)
Luxembourg (6)
Paris (19)
Bordeaux (30)
Basel (25) Zurich (26)
Vienna (24)
Salzburg (27) Graz (29)
Lyon (31)
Milan (32)
Zagreb (33)
Toulouse (34)
Paris (8)
Porto (39)
Prague (7)
Budapest (28)
Marseille (42)
Zaragoza (40)
Turin (35)
Zurich (9)
Belgrade (37)
Bukarest
Vienna
(10) (38)
Bologna (36)
Sofia (46)
Lisbon (43)
Madrid (44)
Barcelona (41)
Rome (45)
Neapel (48)
Milan (11)
Seville (47)
Palermo (49)
Athens (50)
Traffic matrices
LSP requests arrive one-by-one
 Randomly chosen ingress and egress
nodes
 Bandwidth request 1, 2 or 3 Gb/s chosen
with equal probability

Simulation environment

Based on:
◦ Bridgesim1 for native Ethernet
◦ TOTEM2 for GMPLS-controlled Ethernet

Enhancements to simulators:
◦ Implementation of C-CSPF
◦ Computation of recovery time
1: http://www.cs.cmu.edu/~acm/bridgesim/index.html
2: http://totem.info.ucl.ac.be/
Results: LSP placement percentage
GELS
with
protection
GELS with restoration places more
LSPs
than
RSTP places fewer LSPs than RSTP
Results: Bandwidth
placement
GELS with restoration places more bandwidth than RSTP
GELS with protection places less (primary) bandwidth than RSTP
Results: Average link utilization
GELS with protection quickly approaches almost full link utilization
GELS approaches 92% average link utilization
RSTP has lowest average link utilization
Results: RSTP convergence time vs cost to root
RSTP convergence time is highest if the root bridge fails
Convergence time decreases as cost to root increases
Results: Single link failure
convergence time
Single link failure average convergence time
Topology
RSTP
(ms)
Restoration (ms)
tmin
tmax
Protection
(ms)
11 nodes
0.7
32.67
41.61
3.89
50 nodes
102.4
38.13
39.61
6.18
More links closer to root bridge in COST 266
More LSPs were restored in COST 239
Results: Node failure convergence time
Small value
10
50+ ti
t1 - t10 are in milliseconds
i 1
11
Single link failure average convergence time
Topology
RSTP
(ms)
Restoration (ms)
tmin
tmax
Protection
(ms)
11 nodes
4850
30.07
39.34
2.56
50 nodes
3365
42.25
44.24
6.1
10
50+ ti
i 1
50
Small value
t1 – t49 are in milliseconds
Summary

About 45% improvement with GELS over
native Ethernet in:
◦ LSP acceptance
◦ Bandwidth placement

Failure recovery time orders of
magnitude less for GELS than for native
Ethernet
Conclusion
Ethernet is a flexible, cost effective and
efficient transport mechanism for
metro/core networks
 GMPLS promises to be a useful control
plane for Ethernet in metro/core
 Tremendous administrative benefits of
using a single control plane
 Vendors actively working on
standardization of GELS

Questions?

Contact:
[email protected]

Simulator:
http://suraj.lums.edu.pk/gels/