1. Solving IMS problems using P2P technology
by Adrian Georgescu
International SIP Conference
Paris, February 28-th, 2007
Adrian Georgescu

2. Who am I? My name is Adrian Georgescu Founder and CEO of AG Projects
Our focus is the delivery of SIP infrastructure
We specialize in serving companies like cable operators and Internet service providers
We provide simple solutions for complex problems
I have studied Electrical Engineering at Institute of Polytechnics in Bucharest
Adrian Georgescu

3. Short history of PSTN
PSTN is a centralized network where various elements are chained to provide a voice service
The service follow a “create and manage bottleneck” architecture
The need for cost effective growth and the possibility of introducing services brought NGN in the picture
Initially, the network was expensive to build and maintain, pricing was calculated based on distance and time of day.
PSTN became eventually ubiquitous, but arbitrary points of control remained enabled to justify the business models based on resource scarcity.
Adrian Georgescu

4. Introduction to Next Generation Network
NGN is a model proposed by ITU-T that aims to implement a a model similar to PSTN by using Internet protocols
network is application aware
control resides in the network
QoS is a matter of central control
ITU-T networks, such as ISDN, BISDN/ATM/NGN are blueprints based on grand design.
Whether we like it or not, ITU network designs are not based on a natural evolution.
Technically speaking, changes from TDM to ATM to IP are significant discontinuities in the ITU-T architectures.
Adrian Georgescu

5. ITU-T NGN System Architecture
The model is fairly complex.
Strangely enough, the Internet is showed on the side in a small rectangle!
Adrian Georgescu

6. Motivation for IP communications
Lower the operational expenditure by converging the data and voice networks into one single network (cost reduction)
Increase revenues by introducing new services beyond traditional voice services (generate revenues)
Back in 98 it seemed like a very good idea to combine the minute terminating business with cost effective IP technology.
As years passed by, the failure to deploy left room for the pure Internet players to steal parts of the telecom market.
Unfortunately, innovation did not happen from the Telecom side.
Adrian Georgescu

7. IMS, the NGN solution for mobile networks
IMS initially developed by 3GPP to replace the mobile networks
Based on an operator controlled, walled-garden Internet
Based on the SIP protocol developed within IETF
Internet protocols have been chosen for their efficiency while trying to maintain a closed network
IMS was designed to fill the gap between PSTN and IP, by enabling all TDM based services on IP while maintaining a single data network.
It gave a boost to SIP protocol adoption, BUT
It introduced several change request to the initial SIP protocol to accommodate functionality for the existing PSTN model
Adrian Georgescu

8. 3GPP IMS Architecture – all subsystems
The basic design comprises about 12 boxes and 22 interfaces. Many, not showed here, have been later added to solve intrinsic design issues
Components like DNS and NAT traversal and ENUM have been later added to the mix.
Adrian Georgescu

9. IMS, the NGN solution for fixed networks
IMS was designed to implement an all-IP telecommunications environment for the fixed line operators too
IMS supported by ETSI (TISPAN) with its extensions for the fixed line networks (DSL/cable)
Based again on an operator controlled, walled-garden Internet
Because IMS is transport agnostic it is suitable for any kind of network as long as it talks IP.
Another delays were introduced by new changes of the SIP protocol pushed from ETSI to IETF.
Adrian Georgescu

10. ETSI TISPAN IMS ArchitectureRf
/Ro Ut Rf
/Ro Ut
Application Servers
Charging
Functions Sh Rf
/Ro
UPSF
ISC Dh
SLF Iw Cx Dx
IWF P3 Ib
IMS /
PSTN
Simulation
PSTN
Emulation (R2) Mw
Mw/Mk/Mm
I/S -
CSCF
IBCF Ic
AGCF Mi Mk P2
BGCF Mk Mw Mr Mj Gq ' e2 P -
CSCF Mg
SGF
MRFC
MGCF Ie
Other IP Networks P1
This added more complexity and changes to the initial design.
Again, Internet is not part of the picture. Other networks? Gq '
Network Gq ' Gm
SPDF
Attachment
SPDF
PSTN/ISDN
Resource &
Admission
Control
Resource &
Admission
Control Mp Mn
Subsystem
A-RACF e4 MG Re Ia
MRFP T -
MGF UE I -
BGF
Adrian Georgescu
RCEF
BGF
IP Transport (Access and Core)
CNG -

11. IMS is clogged with technical problems
Follows a classic telephony design with chained components, enforcing resource scarcity
End-to-end communication not possible between end-points
Innovation possible and allowed only in the network core
Complexity. 12 components with at least 22 interfaces.
Why, where are the problems coming from?
Adrian Georgescu

12. Why is IMS so complex?
decomposing devices into most granular functions and links
tracking and controlling user behavior
The proliferation of boxes and protocols for the state management required for data tracking leads to cognitive overload but adds little value.
Complexity is ugly
IMS standardizes interfaces instead of behavior.
Generating a billing event is more important than enhancing the end-user experience.
Adrian Georgescu

13. IMS costs are prohibitive
Cost of deployment
Cost of maintenance, many components, boxes, links
Development costs, services deployed only from the core
It will deliver for a huge cost less than your consumers got from Skype yesterday
Do not forget how supermarkets won against corner stores.
Skype customers provision themselves, and Skype has no helpdesk.
You do have controlled provisioning and helpdesk but your do not have 100 million customers.
Adrian Georgescu

14. Do we still need walled gardens?
Telecom industry is definitely not known for its innovations
Raising barriers cost money and returns nothing back
All innovation is done outside on the Internet
By the time you finish raising up your walled garden, the customers are safely outside
Tens of millions of users today are using Skype, Yahoo, MSN.
They are using the Internet.
They have build an walled garden but not bound to a limited network.
Adrian Georgescu

15. Internet is simple and simple is beautiful
Internet is a simple network
Internet services (called applications) are performed at the edge
More CPU power on the edge, more applications available
Increase of bandwidth eliminates the source of QoS problems
It is based on the end-to-end principle
I tried to find a picture to symbolize both simplicity and beauty. I was told it is politically not correct what I have chosen. I apologize.
Network provides a simple routing function ,it moves packets from one IP address to another. It can survive a route failure.
DNS provides friendly names in place of numbering.
Adrian Georgescu

16. the end-to-end principle of the Internet
Network is unaware of the applications, this makes it scalable
Nothing should be done in the network that can be done in an end-system
Quality of Service (QoS) is a matter of bandwidth availability and not of central control
E2E is the most fundamental Internet’s architectural principles.
Internet evolved in speed and number of applications but it is basically the same concept powering it now as it was 30 years ago.
Internet reached current status through natural evolution.
Adrian Georgescu

17. Why do Internet business models flourish?
Services are available on the edge, spread virally
Internet is an “eat all you can” model with easy to bill flat-fee models
New applications are rolled out faster then ever
Free services attract naturally large amount of users to premium services
+10 years work to rollout a service caller Id on PSTN.
Took few months to roll out a new global service like You-Tube.
Adrian Georgescu

18. Real-time communications over the Internet
SIP - Signalling protocol to setup and close sessions
ENUM - Translate telephone numbers into SIP addresses
Presence - enable applications beyond voice
P2P is the most significant Internet development after the world wide web.
And both were not invented by the Internet community.
W3c and IEEE are organizations that are behind many developments we see today on the Internet.
Adrian Georgescu

19. P2P - the most successful Internet application
Widely known for file-sharing and IM applications
P2P today accounts for more than 70% of the Internet traffic
What P2P does, it creates an overlay network for a set of specific applications
P2P provides actually a suite of technologies that solves today the problems of IMS
P2P is wrongly perceived as an evil-doer by many.
Adrian Georgescu

20. What is so special about P2P technology?
Join/Leave: Nodes may come and go at any time, the network topology reconfigures itself
Lookup: The network can locate deterministically the node that serves a specific user or service
Routing: The network can route a message regardless of network topology changes and without manual re-configuration
P2P technology provides the primitives that allow self-organizing networks.
SIP combined with P2P technology provides the functionality and solves at a fraction of the cost, the problems of IMS.
Adrian Georgescu

21. How can P2P solve the problems of IMS?
Self-organizing network (automatic disaster recovery)
Scalability limited by hardware
Cheap hardware resources in place of expensive servers
No idle components, all hardware assets are used
No need for monitoring, operations, engineering, planned maintenance
Let me rephrase this:
One single FTE can run a network of millions of users.
Think about P2P being a server with a built-in technician.
Adrian Georgescu

22. Take only necessary functions from IMS
Drop the overhead!
Adrian Georgescu

23. IMS design should be drastically optimized
Drop the overhead!
And add what the initial designers forgot to add.
We have seen some components like DNS/ENUM added to the picture.
NAT traversal and session border controllers…
Adrian Georgescu

24. Isolate the necessary functions
SIP Proxy/Registrar
Media Proxy
DNS/ENUM
Voic
Presence
User profile database
Accounting
You need some ingredients, but also a good cook.
You may add any other application servers without reconfiguration of the whole network.
Just route based on application to the right node.
Adrian Georgescu

25. Fold all functions into a single box
A single server today can easily handle 20K SIP subscribers and more than 200 media sessions.
Adrian Georgescu

26. Play the hat-trick
Build a self-organizing overlay network by using Peer-to-Peer technology and use the IMS-in-a-box as nodes
Adrian Georgescu

27. A P2P circle of IMS nodes.
Adrian Georgescu

28. Resources working together to address many SIP users.
Adrian Georgescu

29. From complexity to simplicity
We should not be afraid to see through our own mistakes.
Adrian Georgescu

30. to easy accomplishments
From hard work
to easy accomplishments
Adrian Georgescu

31. Questions? Thank you, You may contact me at: ag@ag-projects.com
Adrian Georgescu