This presentation takes you through the challenges network operators are facing as they bring in more and more bandwidth-intensive applications to their network. There are ways to optimize the network from the RAN to the Core -- and improve QoS.
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Core Network Optimization: The Control Plane, Data Plane & Beyond
1. Welcome!
October 4
Mobile Data Offloading Optimization
November 1
Core Network Optimization: The
Control Plane, Data Plane and Beyond
December 6
Optimizing Value Added Services (VAS)
for Greater Revenue Generation
1
2. Today’s Topic & Presenters
Core Network Optimization:
the Control Plane, Data Plane
and Beyond
Presenters:
Karl Wale, Director, Product Marketing
Prashant Sharma, Systems Architect (CTO Office)
Dikshit Sawhney, Product Manager
James Radley, Systems Architect (CTO Office)
November 1, 2012
2
4. End-to-End LTE Infrastructure
Radio Access Network Evolved Packet Core Policy Control IP Multimedia Subsystem
User Policy &
Mobility Charging
Equipment IMS
Management Routing
Equipment
Entity Function
Home eNodeB
Application Media
Server Resource
Function
User Policy &
Equipment Charging
Equipment Enforcement Internet
Function
eNodeB LTE Security Serving Packet
Gateway Gateway Gateway
75+ Customer Wins 10G 40G ATCA Traffic Management Audio Video Conf
Macro Small Cells ~40% ATCA Share Dumb Smart Pipes ~65% Market Share
4
5. Mobile Traffic Profile
More users using more of their
data allowance
Web/Internet
More sessions…more
applications…more signaling
3x
Overall
Video flooding radio & 50% CAGR
transport network Until 2016
Video Stream
Web/Internet 7x
Video Stream
2x Audio Stream
Audio Stream
P2P, Voice Etc. P2P, Voice Etc.
2012 2016
5
6. Optimization vs. Customer Need
Voice Era Data Era
Slow / no internet
Dropped calls Poor video streaming
Poor Call Poor quality Poor Data Cant get email
Quality QoS
Capacity – RAN or core?
Coverage ? Bearer or signaling ?
Why ? Handovers ? Why ? Internet or access network
Policy setting ?
Churn Leave voice… Churn Churn due to data
… promote …promote based
coverage and voice on data capabilities
quality
Solution needed for 3G today, not only LTE problem
6
7. Optimization Goals
Avoid Churn…
QoS
Improve
Cost Service Revenue
Reduction Services
Efficiency Plans
Opex Reduction Tiered Services
Efficiency to defer Capex Core Content Based Pricing
Augment to offload etc. Network Tailored Plans
Optimization
Plan Deploy Optimize
What you invoke and when depends on problem and lifecycle of network
7
8. Core Network Optimization Tools
Network
DPI & Policy Probes
Intelligent Switch LTE & 3G
& Load Balancer Network Core
Video Optimization Traffic
& Gateways Offload
Market For Network Optimization Products Growing by 25% CAGR
Strong growth in DPI, Web & Video Optimization
Network Optimization Needs Blended Approach
…No ‘One Size Fits All’ Solution
8
9. 3G Networks: Efficiency…
Protocol Analysis
Signaling Correlation across signaling & bearer
Probe Application & QoS KPI awareness
Radio Access Network Packet Core PCRF
User Control
Equipment Plane
Femto
User
Equipment
RAN SGSN GGSN PCEF
Attributes KPI
Layer 7 Awareness Voice Data
Tapped vs Bump in Wire Call Sessions
Temporary vs Permanent Installations Bandwidth QoS/latency etc.
Network
Probe
9
10. 3G Networks: Capacity…
Radio Access Network Packet Core PCRF
RF & Transport
User Control …bandwidth mgmt
Equipment Plane
Femto
User
Equipment
RAN SGSN GGSN PCEF
Transport RF & Transport
…direct around core network …transcoding
…local content access …local content re-direct
…tailored packages
Offloading Video
Optimization
10
11. New LTE Networks
Radio Access Network Packet Core PCRF
How Differ ?
User MME RF probe planning
Equipment
Policy ?
HeNB SON
…then optimize
User
Equipment
EnodeB SGW PGW PCEF
Offloading Video
Optimization
11
12. Poll Question #1
Which do you consider the most important network
optimization tool? (Select all that may apply)
a. Stand-alone DPI
b. Video optimization
c. Local content caching & CDN
d. DPI capable network probe (L4-7)
e. 3G offload
f. Small cell / wifi offload
g. SON (Self Organizing Network)
12
13. Role of Stateful Load Balancer
Load Balancer Value Adds
Scalability: how to scale up
and down? Does it need re-
Extended architecting?
Scalability Product
Life Extended life: bridge the
performance/throughput gap
before move to next generation
Fault
Topology
tolerance Topology Hiding: Hide internal
Hiding
details (blades/servers) from
peers
Fault tolerance: Redirect flow
to new active element
13
14. Load Balancer - 1
DPI Filtering Transport Optimization
Load Balancer - 2
DPI Filtering Transport Optimization
Flows 1 - 10
Video Optimization Blade – 1
Transcoding Trans-rating User awareness
Flows 51-55
Flows 11 - 50
Video Optimization Blade – 2
ATCA Platform
Stateful Load Balancer
Transcoding Trans-rating
… User awareness
Video optimization gateway
Video Optimization Blade – n
Transcoding Trans-rating User awareness
blades
HTTP/video traffic
non–video traffic to be
Transport offloading:
DPI based Filtering: Stop
14
passed to video processing
of TCP connection that carry
Offload handling/optimization
16. Stateful Load Balancer
Extended life cycle using load balancer
Scalability: Upfront load balancer
to scale existing network
probes/monitoring box
I/O: LB should be able to support
large amount of I/O’s
Protocol awareness: LB should
understand all of wireless protocols
and their transports
Distributed load balancing: LB,
itself, should be scalable to support
variable number of backend
application servers
16
17. Stateful Load Balancers
Carrier Cloud Load Balancers
Scalability: Upfront load balancer
to being efficiency of scale in
carrier cloud
Fault tolerance: Fault tolerance of
LB itself becomes one of the most
critical aspect in the carrier cloud.
This includes fault tolerance at
platform, I/O, blades and
application level
Protocol awareness: LB should
understand all of wireless protocols
and their transports
17
18. Load Balancer: Key Characteristics
ATCA Platform for load balancer
High number of I/O: A
heterogeneous platform like ATCA
offer variety of I/O solution including
centralized and distributed
Specialized processing: State of
art packet processing and switching
technologies (XLP, OcteonII, NP4,
Trident) for common function
offload
Fault tolerance: Carrier grade
ATCA provides redundancy at I/O,
platform elements, backplane and
blade level
18
19. LTE Network Overview
Home
Subscriber
AF IMS
Application
Server Network
Function
Mobility
Management Policy &
Entity Charging
Rules
Function
Internet
eNodeB Serving PDN Policy & Charging
UE
Gateway Gateway Enforcement Function
Web
Radio S1 S5
Email Bearer Bearer Bearer
Service
Data Flows
Voice
Packet
Filters
19
20. Main Drivers of Signaling Traffic
What is driving the signaling traffic in a LTE network?
• Data Usage
– Multiple connected devices: smartphones, tablets, notebooks, smart
cameras, M2M etc.
• Small Cells
– Future networks will be heterogeneous i.e. a combination of macro
and small cells (femto, pico, micro etc.). The evolved CN has to
manage a lot more base stations than the legacy networks.
• IMS
– VoIP based call control network in LTE provides rich communications
capabilities not limited to just voice conversation
• Policy & Charging Control (PCC)
– Policy and service based charging plays a key role in the LTE
networks. This is causing a tremendous increase in Diameter
signaling traffic with in the EPC that needs to be managed. 20
21. S1-flex Architecture
• S1-flex architecture uses many-
to-many network architecture
between eNBs and MMEs for
load balancing and redundancy
purpose. eNB selects a MME on
UE registration based on the
current resource utilization at
each of MMEs in the pool area
Pool Area
• MME uses a similar logic to MME MME MME
select a S-GW from a pool of S-
GWs serving the UE area
S1
• Possible to re-direct UEs to new
MMEs in case of overload at one eNB eNB
of the MME in the pool area
• S1-flex is a pre-requisite for the
Network Sharing architecture
discussed in the next chart 21
22. Network Sharing
• This optimization technique
involves sharing RAN and CN
resources among multiple
service providers. It is possible
to share just RAN (MOCN) or
both RAN and CN nodes (GWCN)
• Multiple PLMN-id(s) are
broadcasted on the air interface.
UE selects a candidate PLMN
and RAN assigns the CN node
based on resource utilization
and current loading of the
shared CN elements.
• Applicable to both 3G and LTE
networks
22
23. LIPA
Local IP Access (LIPA)
• An offload GW is co-located with
the small cell (HeNB/HNB) and
routes the data destined for
home/enterprise network
appropriately bypassing the EPC
• UE uses standard signaling
methods as a regular EPS bearer
to setup the LIPA tunnel
• LIPA can be enabled on per
APN/UE basis
23
24. SIPTO
Selected IP Traffic Off-load
(SIPTO)
• Network uses DNS or other
mechanisms to select a GW in
close proximity to the UE’s point
of attachment to the access
network and offload the traffic
from there
• Option to enable off-load on a
per UE/APN basis
• Applicable to both small cell &
macro networks providing E-
UTRAN/UTRAN access
24
25. Multi-mode Small Cells (3G/LTE/WiFi)
Non-3GPP Access
• Use commonly deployed WiFi
access points to offload traffic to
the Internet.
• Access Network Discovery and
Selection Function (ANDSF)
helps UE in selecting the
appropriate access network
based on Operator policies.
• Architecture standardized by
3GPP so inter-access mobility is
covered.
25
26. Misc. Network Optimization Techniques
• Co-located SGW, PGW and
GGSN nodes. This can improve
the packet latency by eliminating
one of the nodes in the data
path. Mainly a deployment
decision governed by network
topology i.e. ratio of SGWs to
PGWs/GGSNs
• Similar colocation is possible for
control plane nodes i.e. MME &
SGSN. Allows for reduction of
signaling traffic during inter-RAT
(3G<->LTE) mobility
• Direct tunnel architecture for
UTRAN. This uses a direct
connection from RNC to S-
GW/GGSN bypassing the SGSN
and thus improving packet
26
latency
27. Policy and Charging Control
Architecture (PCC)
Subscription
Profile AF
Repository
(SPR) Online
Sp Rx
Charging
System (OCS)
Policy and Charging Rules
Function (PCRF)
Gxx Gx
BBERF PCEF
Gy
Gz Offline
Charging
System
Gateway (OFCS)
27
28. Diameter Routing Agent - DRA
Diameter is extensively used as an AAA protocol in the DB, charging,
and policy domains of EPC and contributes to majority of signaling
traffic load in the 4G networks.
Scalability demands multiple PCRF(s)/HSS(s) and Charging
DRA helps with routing, load balancing and session management
of traffic flowing between these Diameter entities.
DRA ensures:
all Diameter sessions established for a given EPS connection
reach the same PCRF when multiple and separately addressable
PCRFs have been deployed in a Diameter realm
A DRA can also incorporate SLF functionality to locate HSS for a
IMS UE when multiple HSSs are deployed.
A DRA can be implemented as a re-direct or a proxy agent.
28
32. Poll Question #2
How do you see the opportunity for Software Defined
Networks (SDN) in your organization?
a. Revolutionizes how we architect our networks
b. Has real potential but SDN needs to mature as a
technology before it will be of use in a live network
c. Interesting for some niche functions within the network
but will be restricted to a limited set of element types
d. Irrelevant as we are already able to manage networks in a
way that suites our needs
32
33. What is a Software Defined Network?
There are two commonly accepted defining attributes
of a Software Defined Network (SDN):
• Decouples network management elements from the packet
forwarding entities. Network intelligence & supporting
protocols are maintained independently of the network nodes
which actually handle the through traffic.
• Provides an API so that application developers can have their
own applications directly configure that part of the overall
network infrastructure which delivers packets on their behalf.
Although not part of the ‘common definitions’ the
abstraction of the various network elements down into
a single virtual switch is seen as an important benefit
of a SDN.
33
34. Why All the Excitement?
Possibility for single virtual switch/router image
• Facilitates rapid and consistent deployment of new rules across
the entire network
Independence to go beyond vendor provided features
• Network architects can leverage more of the capabilities of the
underlying hardware elements in their network
Promotes innovation
• Separation of function allows both network element hardware
and switch management suite vendors to break into the market
Allows applications control over their network
• Applications get similar control over their ‘virtual slice’ of the
network as they have over their virtual server environment
34
35. Challenges in growing SDN into
Carrier Networks
Topology Management
• Most of today’s SDN management s/w deals well with flat full
mesh network infrastructures – not dynamic hierarchies.
Policy Policing
• How to control how much network resources a management
agent can reserve?
Security
• How to prevent the creation of illicit portals?
• How can a network entity spot a rogue rule?
Traffic Management
• While good at creating traffic flow classification rules not so
good at defining traffic scheduling characteristics
35
36. Opportunities
De-coupled control & data plane
• allows for independent network scaling
Standardise network management strategy
• while keep flexible hardware choices
Allows innovative network appliances to be created
• Powerful APIs open up market to much wider developer pool
• Applications with integrated control over their network can
deliver better services
• Provides opportunity to support the less common (or even
proprietary) routing/forwarding protocols on Common Off The
Shelf (COTS) network devices
36
37. Basic Model of a Network Appliance
New packets arriving enable additional detail to be extracted from flow
…approx 10% packets
HTTP GMAIL Metadata
…Username
…Email title
Server Load
Open Flow …Content
State Machine
User Application
API Add new entry by default…
Application adds …or wait for application Buffered
table entry & rule Packets
Apply Rule
e.g. put into correct
priority queue
37
39. Q&A
Contact us!
Karl Wale Prashant Sharma
karl.wale@radisys.com prashant.sharma@radisys.com
James Radley Dikshit Sawhney
james.radley@radisys.com dikshit.sawhney@radisys.com
~Please fill out our short survey~
THANK YOU FOR ATTENDING!
To register for upcoming webinars:
http://go.radisys.com/optimizing.html
39