Transcript Group mobility

Supporting Group Mobility in MissionCritical Wireless Networks for SIPbased Applications
Project LaTe
Topics










Background
Session Initiation Protocol
SigComp
Group Mobility
Hierarchical State Routing
Group mobility models
Predictive Address Reservation
Simulation part
Conclusions
Final remarks & future work
Background: project LaTe 1/3


”Langattomien teknologioiden
käyttömahdollisuudet puolustusvoimien
tietoliikenneverkoissa” / ”Possibilities for
wireless technologies in defence networks”
funded by the Finnish Defence Forces
A joint research program of HUT Networking
Laboratory, Communications Laboratory and
the Finnish Defence Forces, commenced in
2003
Background: project LaTe 2/3

Contemporary disaster relief operations rely heavily on realtime wireless communications
–
–

The rapid development of civilian communications technology
has caused their prices to decline fast, making them an
attractive alternative for the military-grade equipment
–

these systems fall into category ”It Just Must Work”
the technology commonly used for these ends has had propensity
to be expensive
remember the price discrimination: a price charged from a
governmental authority is N-fold compared to the price charged
from a civilian party
Project LaTe is an attempt to find ubiquitous, affordable and
easily disposable wireless solutions to complement (and even
completely substitute) the aging authority communications
equipment currently in use
–
Commercial Off-The-Shelf (COTS)
Background: project LaTe 3/3
Netlab involvement (master’s theses)
2003 Wireless LAN Security (Ahvenainen, Marko)
2004 Mobility management with Mobile IP version 6 (Merger, Mikko)
2005 An Overview of Mobile IPv6 Home Agent Redundancy (Keränen, Heikki)
2006 Mobile IPv6 performance in 802.11 networks:
handover optimizations on the link and network layer (Hautala, Mikko)
2007 Analysis of Handoff Performance in Mobile WiMAX Networks (Mäkeläinen, Antti)
2007 Supporting Group Mobility in Mission-Critical Wireless Networks for SIP-based
Applications (Repo, Marko)
2008 …
Master’s thesis: the main themes

Session Initiation Protocol (SIP)
–
–
–

Seamless handoffs during mobility
–
–
–

flexible, scalable and reliable signaling protocol
inadequate in terms of bandwidth & security
good starting point for application-layer mobility
VoIP & data
inter-domain mobility assumed
scarce network bandwidth & resources
Group handoffs
–
–
–
”Group Mobility” is a term originally coined in the world of ad-hoc networks
assumes that network nodes exhibit group behavior (often realistic!)
attempt to forecast the future need of network resources and minimize the
required amount of signaling during handoff procedure
Session Initiation Protocol 1/3


Citing RFC3261, SIP is ”an application-layer control
(signaling) protocol for creating, modifying, and
terminating sessions with one or more participants”
Has undergone a lot of development during the last
half a decade
–
–

and still does (various interoperability forums and events
held by SIP Community)
and will do (3GPP NGN/IMS, IETF, Microsoft etc.)
Has gained a significant foothold as a signaling
protocol both in academia and private sector
companies, competing with ITU-T H.323 mainly
backed by the telecommunications industry
Session Initiation Protocol 2/3


Provides all needed primitives for establishing a connection
between 2-N end points
Transport independent
–


Supporting unicast and multicast
Extremely scalable
–

UDP, TCP, SCTP, …
Intended as a subscriber signaling protocol, but functions virtually
in every network core where the intelligence is located at the
edges
Intercompatible when required
–
–
–
ITU-T H.323
ISUP (SS7)
Q.931 (ISDN)
Session Initiation Protocol 3/3

Issues
–
UTF-8 ASCII format implies bandwidth inefficiency


–
”Light-weight”?


–
Way no. SIP is already as complex as H.323. By the date, the SIP
specifications contain thousands of pages
Irony underneath: the protocol design started from the need for a robust
signaling mechanism characterized by simplicity and lightness
Many open security questions


–
SIP was not designed for low-bandwidth wireless environment
Attempts to alleviate the bandwidth issue have spawned mechanisms
such as SigComp. Many problems and issues.
signaling
media
Virtually no support for seamless mobility

Cannot be handled with MIPv4/v6, due to the triangular routing
phenomenon (too high latencies involved!)
–

suitable for data connections with loose temporal requirements
The real-time streams problematic (VoIP can withstand <100ms latencies
without degradation)
SigComp



Attempt to address the bandwidth issue by binary
compressing text-based SIP messages
May improve efficiency especially on low-bandwidth
connections
However, SigComp has some severe shortcomings
–
–
–
–

consumes computing power for message processing
requires a lot of memory for storing state information
security issues (may subject to DoS attacks)
problems with mobility
After all, SigComp introduces another extra layer,
and thus more complexity. So, we’ll take a different
approach.
Group mobility 1/2
The fundamental problem with SIP:
It was never intended for narrowband
airlinks. The size of a single
message with a payload can range
anything between a few hundreds
of bytes to many kilobytes.
Ergo, even a modest number of
moving nodes may generate a
significant amount of SIP signaling
traffic during connection hand-off.
Group mobility 2/2
We may try to eliminate the
unnecessary signaling by
dealing with groups instead of
individual nodes.
Introducing group handoffs.
Another approach: WiMAX & MRS
Creating an isolated cell using a
mobile relay station (MRS),
which gains the control of the
moving mobile nodes.
Suitable for public transportation
vehicles (buses, trains, aeroplanes)
where groups guaranteed to stay
compact. Not suitable for loose or
scattered groups (e.g. infantry).
Hierarchical State Routing 1/3

HSR: A link state protocol
–

Applies hierarchical addressing to keep channel
utilization efficient
–


a low-latency routing solution for applications requiring
group mobility
conservative on routing table sizes
Unbundles the physical affinity from the logical
partition representing different logical or functional
levels where the nodes may reside
The amount of signaling remains low, since there is
no need for flooding
–
even when the location of the corresponding node is not
known
Hierarchical State Routing 2/3
Hierarchical State Routing 3/3

Better in terms of complexity (=fewer routing table entries) than
traditional flat routing schemes
– Let N : no. nodes, M : no. hierarchy levels; then
– Flat routing: O(NM);
– HSR: O(N X M).
Leads to better scalability

The flip side of the coin: constant need for updating
databases
– increased complexity + update latency
– dynamic cluster re-arrangement?
Handoff delay components

Link layer (L2) delay
–

Movement detection (L3)
–


RTT for re-INVITE and message processing, a major contributor
Packet transmission time
–

Duplicate Address Detection (DAD) is a major source of delay!
Re-configuration delay
SIP re-establishment delay
–

Router Solicitation / Router Advertisement
DHCP
–

scanning, authentication and reassociation
The time for first packet to be exchanged over the restored
connection
QoS + AAA (optionally)
–
Quality and security reservation introduce some latency when
used
PAR-SIP 1/4



Predictive Address Reservation (PAR) is a
mechanism attempting to alleviate incurred handoff
latency by eliminating the most significant sources of
delay: Duplicate Address Detection (DAD) during
DHCP and SIP connection re-establishment (reINVITE)
Allows approximate latencies of ~60 ms, allowing
possibly even better performance!
Allocate L3 addresses and the session
establishment proactively, so that the handoff
process is almost seamless
PAR-SIP 2/4
1.
MN starts searching for a new AP/BS when the Signal-to-Noise falls below
the Cell Search Threshold
2.
MN consults its internal database and chooses a suitable target BS (TBS),
then sends a reservation request to its serving BS (SBS)
3.
SBS consults its neighboring BS table to see whether the MAC of the TBS
belongs into the same (L3) domain or not
1.
If so, the SBS initiates a normal L2 handoff (L2HO) procedure
2.
If not, a network level (L3) handoff is needed. The SBS requests a new IP address
from the TBS, which obtains it using DHCP and allocates resources proactively.
Reservation reply containing procedure acknowledgments and a new IP address is
sent to the MN
PAR-SIP 3/4
4.
Subsequently, the MN sends a re-INVITE request to its
corresponding node (CN), using its newly reserved IP address
5.
The CN opens a new session in parallel with the old session
6.
The packet exchange happens through both sessions (bi-casting)
until the handoff procedure is completed

7.
for minimizing the amount of lost packets
When the handoff is completed, the old session will be torn down. All
traffic is now sent using the new session.
PAR-SIP 4/4
Group Mobility Models





Mobility models are needed for system analysis and protocol
during the design phase, but also for predicting the future
availability of wireless resources
Conventional models (Random Walk, Gauss-Markov) put the
emphasis on individual entities
In many cases, however, it makes sense to observe the
movement and interaction characteristics for groups instead
Group mobility is currently undergoing heavy research, mainly in
the world of ad-hoc networks
The future need of resources can be predicted with aid of group
mobility models.
–
logic: when a MN belonging into a group performs handoff, it can be
anticipated that that others will follow in a certain pattern

the rest is about queuing theory and e-λt:s…
Column Mobility Model
The most simple group mobility model. It is a conventional model for
representing e.g. field operations involving searching activity.
The group consists of MNs associated with a line of reference, which fully
characterizes the group behavior.
The participants also have a reference point on
the line, around which they may freely wander.
The movement of individual nodes does not
have effect on the location of group center.
Pursue Mobility Model
Another simple model representing e.g. a chasing scenario. A target node
(TN) takes now the place of the point of reference, which denotes the
group ”center”.
At any time t, the scenario can be modeled mathematically:
MN i (  1)  MNi ( )  Ai  RM i
Where MNi is place at any time t, A is an
acceleration vector of form F(TN – MNi ), i.e.
position of the target node TN and the Mobile
Node i. RMi is a random motion displacement
vector for any node i, RM << A
Nomadic Community Model
Describes activity of wandering tribes, camping for night. One may
imagine that the point of reference (RP) is the camp fire.
The group motion vector GM represents the movement of the campfire
(RP), and the mobile nodes are able to wander around it randomly.
The roaming distance can be set as a parameter.
Reference Point Group Mobility
RPGM is perhaps the most generally seen ad-hoc mobility
model. It can be considered of generalization of all the
presented. RPGM it is also maybe the most commonly
studied group mobility model as it comes to the ad-hoc
mobility. Has been an inspiration for several derivative
models.
The location vector for each
individual node i can be written now:
PNi (  1)  PNi ( )  GM  RM i
Simulation part 1/3

Carried out using network simulator ns-2
–
several contributed modules needed


–
C++ coding needed




insufficient 802.11b model
No way to model PAR
Attempt to demonstrate the benefits obtainable by deploying
GM-enhanced PAR-SIP with four plausible scenarios
–

Mobility enhancements (NIST HSNTG)
A SIP module by Rui Prior
simulating VoIP (RTP) and data (TCP) traffic
Indicators of interest: total traffic, hand-off latency and packet
loss during the hand-off process
As of May 2007, work still in progress!
Simulation part 2/3
Simulation part 3/3
Conclusions


The main goal of this thesis: minimizing signaling, minimizing
handoff latency!
SIP is the choice of the future, currently undergoing very rapid
& active development
–


However, yet a far cry from all-around protocol
There are many ways to mitigate the incurred handoff latency.
Predictive Address Reservation (PAR) is one of them.
Group mobility mechanisms aim at minimizing the unnecessary
signaling during handoff, allowing better channel utilization in
many scenarios, group handoffs (=group handovers) are their
realization.
Final remarks & future work

802.11x not necessarily the most realistic platform for such
wide-area scenarios
–
–


Vertical handovers? IEEE 802.21 (Media Independent
Handover) on the verge of introduction
How about voice and data taking different routes?
–

hybrid MIP-SIP
The research dealt solely with the most rudimentary transport
level protocols, UDP and TCP
–

as it comes to €uro$, very alluring (comparing to WiMAX!)
still undergoing evolution
how about more advanced protocols? DCCP? SCTP?
Hybrid networks? The strict division into infrastructured and adhoc networks is likely to disappear in the future
–
–
actually, this is happening already, slowly but steadily…
look at VIRVE/TETRA for instance, but also civilian applications
(WPANs, Bluetooth, UWB, …) although the scale is different
The End
Thank you!