Data in GPRS Networks
Antti Kantee <pooka@cubical.fi>
Overview of our Slight Plight
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GSM Data
what is needed to provide data on top of GSM what’s wrong with the picture
GPRS:
overall architecture
components & interconnections protocols
roaming
problems with GPRS
PoC: GPRS use in "real life"
GSM Data: packets in the circuit
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GSM data is circuit switched
establish connection, use connection, disband connection setup time
charging is time-based
Internet-connection via (separate) ISP
GSM -> modem -> PSTN -> modem -> Internet GSM -> ISDN -> Internet
High-Speed Circuit Switched Data (HSCSD)
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
one GSM TCH/F provides 9.6kbps user data total number of TCH’s available: 8
very simple idea: use several TCH/F’s
up to 57.6kbps total speed
8x9.6 = 76.8kbps, but A I/F is limited to 64kbps
de/multiplexing handled at MS and BSC+MSC
transparent to RF
transparent to outside network
transparent to whoever uses the stream
symmetric and asymmetric connections
SMS: User and Control Plane Mixup
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
(ab)use signalling plane for transmitting user data
people have been doing this forever, e.g. non-answered phonecalls which "transmit" data using CID
main advantage: don’t need user-plane components for installations only wishing to do SMS
two types: broadcast and point-to-point
teenagers found mobile equipment, and the rest is history
SMS business view
Quick Review of GPRS Itself
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
packet switched link
dynamic allocation of radio resources speeds up to 171.2kbps
that’s theory for you
"integration" with packet switched networks, such as the internet
billing on transfer volume, not connection duration
image source: http://www.cs.hut.fi/~hhk/GPRS/gprs_own.html
GPRS Architecture: Overview
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GPRS: diffs to GSM
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GPRS was designed as a plugin to GSM
business POV: easy to sell add-on to existing architecture
GGSN - GPRS Gateway Support Node
tunnel packets between SGSN and PSPDN using GTP
SGSN - Serving GPRS Support Node
mobility management and packet routing
PCU - Packet Control Unit
handle radio link layer for data frames
CGF - Charging Gateway Function
make sure nasty users are charged for all the trouble they cause
image source: http://www.cs.hut.fi/~hhk/GPRS/gprs_own.html
GPRS Protocols: Signalling Plane (MS<->SGSN)
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
image source: http://www.cs.hut.fi/~hhk/GPRS/gprs_own.html
GPRS Protocols: User Plane
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GPRS Protocols: SNDCP
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
Subnetwork Dependency Convergence Protocol used between MS and SGSN on user plane
maps external PDN to GPRS network
features:
multiplex PDPs onto LLC connections compression
de/fragmentation
GPRS Protocols: LLC
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
Logical Link Control
provides a logical link between the MS and SGSN multiplexing between one SGSN and several MS’s abstracts radio characteristics
features:
un/acknowledged data transfer flow control
data confidentiality
variable-length frame support
I heard it through the GPRSvine
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
as known, GPRS uses TDMA radio 52-multiframe structure
4 consecutive blocks in same slot = radio block RLC layer:
provide un/acknowledged transfer between MS and BSC
MAC layer:
provide medium access control (surprise!) collision avoidance, detection
arbitration
Filling the slots
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
MS and BSS need to agree on which timeslots to fill - non-trivial problem
slots reserved for voice, but not used if silent slots reserved for data
slots available to GPRS users
predictive best-fit timeslot assignment, "tetris"
fixed allocation
dynamic allocation
extended dynamic allocation exclusive allocation
GPRS Mobility Management
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
GPRS defines three different states for MM
tries to optimize battery consumption & uplink usage
idle:
network does not track MS location standby:
network knows MS location per RA busy:
network knows MS location per cell
GPRS Session Management
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
before being able to actually transmit/receive PDU’s, the MS must establish a PDP context
e.g. dial "*99#" on your phone
MS sends activation request to SGSN, which selects GGSN for session
PDP context includes
PDP address for MS QoS parameters
PDP context will time out if not used
keepalives
IP on GPRS (basically the previous slide again)
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
MS tells the SGSN that it wants to use IP
along with other parameters, such as desired IP address and QoS parameters
SGSN checks if MS has credentials for that and does authentication
also establish an encryption
SGSN selects GGSN for MS
establish a GTP tunnel
IP context created in MS, SGSN and GGSN
MS can now transfer packets into IP network and back
user plane protocol stack
Roaming in GPRS
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
possibilities
1: exit PLMN as early as possible, use visitor location GGSN 2: exit PLMN from home location GGSN
the advantage of option number 1 is that it is smart from the POV of network topology
packets can directly be routed to the correct destination
however, option 1 makes billing difficult
therefore, option 2 is a winner
option 2 also makes mobility simple
you can always have the same PDP address
Latency in GPRS
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
several factors contribute to latency in GPRS mobile station
process datagram and request radio resource
radio resource reservation
MS must request TBF
if none is active, negotiation can take a while usually source of significant portion of delay
transmit over air interface
MS -> BSS -> SGSN
transmit through core network
SGSN -> GGSN roaming?
(source: SourceO2.com)
Some figures for latency
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
measured: 500 byte IP packet transmission MS delay: 50-200ms
TBF establishment
1 uplink TBF: 320-750ms
2 download TBF: 290-1700ms
once established, 0ms delay (surprise!)
air delay: 260-460ms
depends on number of TBF in use
core network delay: 20ms
non-roaming situation
PoC: Push-to-talk over Cellular
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
use data channels to transmit voice
VoIP
structured so, that you can have calls "always open"
use tangent to signal when you’re actually transmitting GPRS: pay only for traffic
especially group calls benefit from this feature
GPRS & latency
"push-to-wait"?
Wrapup
Antti Kantee, T-110.456 Next Generation Cellular Networks. 2004 .
data on the Internet is packet-based
GSM data causes annoyances in Internet use
GPRS is packet-switched and easy to integrate with PDN’s
GPRS has been easy to plug into the existing GSM architecture
well, it was designed for that
GPRS falls short on a couple of aspects
data rates & latency issues
GPRS works "well ’nuf"
too well for 3G to be tempting?