Upload
vutram
View
219
Download
1
Embed Size (px)
Citation preview
Budapest, 26-28 October 2016
BIG DATA INTERPRETATION AND CHALLENGES IN MOBILE NETWORK TESTING Benjamin Teke and Zoltán Elzer© All rights reserved
Testing challengesMOBILE TELEPHONY NETWORK
2
• Network elements
• Separate software instances
© All rights reserved
• Network
• Connection of network elements
• Communication according to
standards
I WANT TO
MAKE A
CALL
YOU HAVE
A CALL? „What is in the network?”
Testing challengesCOMPLEXITY GROWTH
3 © All rights reserved
2G
3G
User Data Management domain2G / 3G network domain
MSC
RNCNodeB
BSCBTS
2G/3GUser
SGSN
GGSN
Internet
4G network domain
eNodeB
4GUser
MME PCRF
P-CSCF
I/S-CSCF
E-CSCF/EATF
MMTel AS
SCC AS
MGWMRFP
MGCF
BGCF
MRFC
TrGW
IBCF
BGF
Presence Server
AAA
ePDGInternet
Internet
4G (LTE)
Voice over 4G (VoLTE)
Voice over WiFi
SGW PGW
HLR/VLR
HSS/SLF
More complex
network
architecture
= MORE TESTING
IP Multimedia
Subsystem
(IMS) domain
Testing challengesTEST LEVELS IN MOBILE NETWORK DEVELOPMENT
4 © All rights reserved
Solution
test
Domain test
Product test
Software module test
• Focus on end-to-end
network
• Communication between
network functions
• Check interworking and
end-to-end behaviors
Testing challengesSOLUTION TEST
5 © All rights reserved
Mobile
networkSimulators
Traffic capturing
and data collectionUser equipment
Test & operation
framework
Signaling
messages
Load generators
Beyond testingAFTER TEST EXECUTION
6 © All rights reserved
Test
executionMonitoring
the network
Troubleshoot /
Repair
Collecting dataReal time/offline
analysis
• Automation of steps is required!
Automated analysis, troubleshoot
and repair based on collected data
Focus
Beyond testingDATA COLLECTION
7 © All rights reserved
Network element reported
session records
• MB to GB / day
Network element reported
performance data
• MB to GB / day
Network element internal logs
• MB to GB / day
Traffic capturing
• Multiple TBs /day
SUM data size: „Small” VoLTE network = ~1-2 TByte/day
Example: 300+ application level messages per VoLTE call
Beyond testingDATA COLLECTION – TRAFFIC CAPTURING
8 © All rights reserved
• Monitoring
• Various points of the network
• Collect signaling messages
• The gathered data
• Linear list of messages
• Not certainly in the expected
order
User Data Management domain2G / 3G network domain
MSC
RNCNodeB
BSCBTS
2G/3GUser
HLR/VLR
SGSN
GGSN
Internet
4G network domain
eNodeB
4GUser
MME PCRF
SGW PGW
HSS/SLF
IP Multimedia Subsystem (IMS) domain
P-CSCF
I/S-CSCF
E-CSCF/EAT
F
MMTel ASSCC AS
MGWMRFP
MGCF
BGCF
MRFC
TrGW
IBCF
BGF
Presence Server
AAA
ePDGInternet
Internet
„How do we get the data?”
Beyond testingCORRELATION – THE TRADITIONAL METHOD
9 © All rights reserved
• FIND is the keyword
• Incoming data
• No structure
• Flows can overlap
• Troubleshooting
• Time consuming
• Prone to errors
10 © All rights reserved
Beyond testingCORRELATION – STRUCTURING THE DATA
• Correlated data
• Assembled flows
• With grouping and
filtering options
• Placed in the network
Beyond testingCORRELATION – SIMPLIFIED WORKFLOW
11 © All rights reserved
Intra-protocol
correlation
Inter-protocol
correlation
Forming the
input
Connect flows
to end-user
User Data Management domain2G / 3G network domain
MSC
RNCNode
B
BSCBTS
2G/3G
User
HLR/VLR
SGSN
GGSN
Internet
4G network domain
eNodeB
4GUser
MMEPCR
F
SGW PGW
HSS/SLF
IP Multimedia Subsystem (IMS) domain
P-CSC
F
I/S-CSC
F
E-CSCF/EATF
MMTel AS
SCC AS
MGW
MRFP
MGCF
BGCF
MRFC
TrGW
IBCF
BGF
Presence
Server
AAA
ePDG
Internet
Internet
Beyond testingCORRELATION – BIG DATA PROCESSING
• The tasks can be a separate services
• Services can run parallel or pipelined parallel
• Multiple instances can be started
• Real-time processing is achievable
• Micro-batch processing
• Same algorithm for offline and online processing
12 © All rights reserved
”Information must be correlated”
Intra-protocol
correlation
Inter-protocol
correlation
Forming the
input
Connect flows
to end-user
Beyond testingBIG DATA FRAMEWORKS – WHAT DO WE USE?
• Own platform
• Helps to develop, test, maintain applications using Big Data
technologies
• Gathers open source frameworks
• Used frameworks
• Processing: Hadoop MapReduce, Apache Storm, Apache Spark
• Messaging: ZeroMQ, Apache Kafka
• Misc.: Hadoop HDFS, Apache HBase, Apache Phoenix
13 © All rights reserved
14 © All rights reserved
Example
Takeaway
15 © All rights reserved
Focus on end-to-end
system behavior and drill
down to the details
Automation of data
collection, analysis and
troubleshooting
Common solution for
offline and online analysis
Multi-stage correlation to
parallelly process diverse
data
QUESTIONS & ANSWERThank you for your attention.
© All rights reserved
COME AND SEE OUR DEMO