Signaling load evaluations for policy-driven cognitive management architectures

1
Signaling load evaluations for policy-driven
cognitive management architectures
Kostas TsagkarisM. Akezidou, A. Galani, P.
Demestichas Telecommunication Networks
Integrated Services LabDept of Digital Systems,
University of Piraeus, Piraeus, Greece
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
2
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

3
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
4
Motivation, problem area

• Significant increase in user demands, mainly
  stemming from the wireless and mobile domains
• Heterogeneous wireless networks with great levels
  of complexity
• Multiple, collaborating Radio Access Networks
  (RANs)
• Able at operating a plethora of diverse Radio
  Access Technologies (RATs)
• Variant types of Mobile Terminals (MTs), with the
  ability to choose among various supported
  RAN/RATs
• Both devices and networks with dynamic spectrum
  access capabilities

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
5
Motivation, problem area

• How to optimally select and use spectrum and
  radio resources?
• Stress for developing mechanisms to confront the
  challenges and to leverage the opportunities
  posed by such a versatile radio environment.
• Solution adaptive and flexible management
  paradigms that are able to dynamically manage
  network elements and terminals, ensuring the
  great availability and efficient usage of
  spectrum and other radio resources

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
6
Motivation, problem area (cont.)

• A Policy-driven Cognitive Management Architecture
  is proposed as a solution to the above problems
• destined to the optimized management of future
  wireless networks operating in versatile radio
  environments
• comprises a generic amalgamation of a set of
  architectures that have been proposed for
  managing future networks
• A variant structure of the Functional
  Architecture (FA) that has been developed by the
  E3 European project
• for introducing reconfigurable, cognitive Systems
  in the B3G world
• for improving the utilization of spectrum and
  radio resources
• More importantly, a variant of the architecture
  that has been actually elaborated within the
  Working Group 3 (WG3) of the ETSI Reconfigurable
  Radio Systems Technical Committee (ETSI RRS TC)
• In May 2009, the ETSI Board has approved the
  respective technical report as ETSI TR 102.682
  Functional Architecture (FA) for the Management
  and Control of Reconfigurable Radio Systems

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
7
Motivation, problem area (cont.)

• However little work on the performance
  assessment of the above cognitive management
  architectures so far
• This work places focus on the evaluation of
  signaling loads that the cognitive management
  architecture will burden to the network that it
  operates in

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
8
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

9
Overview of considered cognitive management
architecture

• DSNPM
• manages and reconfigures the network elements
• detects the new elements
• provide the essential configuration information
  of the managed RATs to MTs
• derivates policies for the managed MTs
• calculates metrics on spectrum utilization

• Functional entities and interfaces

• DSM
• assigns operating frequencies to RATs
• detects opportunities for sharing or trading
  spectrum with other network operators (NOs)

• JRRM
• is responsible for jointly managing the radio
  resources belonging to heterogeneous RATs
• performs functionalities such as MT access
  selection, neighborhood information provision,
  and QoS/bandwidth allocation/admission control

• CCM
• implements the decisions of DSNPM and JRRM in
  network elements and JRRM in MTs
• responsible for all the stages of reconfiguration
  and all the possible related actions (e.g. SW
  downloads)

• RAT
• Encapsulates the set of network and/or terminal
  elements and resources that are bound to a
  specific radio access technology and are subject
  to reconfiguration

10
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

11
Signaling load evaluation methodology

• Our purpose is to calculate the signaling loads
  in the interfaces of the architecture
• Assuming mapping of the functional interfaces to
  actual system interfaces e.g. ones defined in
  3GPP
• The interfaces are first defined in terms of
  elementary procedures
• Every single operation/scenario in the considered
  architecture is supposed to be built from a set
  of elementary procedures taking place in the
  interfaces of the architecture (thus the term
  elementary)
• Calculations can be then done by characterizing
  the signaling loads that are needed to carry out
  this set of elementary procedures

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
12
Signaling load evaluation methodology (cont.)

1. Determination of procedures in the considered
   scenario(s) Every scenario in question can be
   seen as a constitution built from a set of
   elementary procedures taking place in the
   interfaces of the architecture.
2. Determination of messages Determination of
   messages that flow in-between the peer entities
   of the interfaces and are used to carry out the
   procedures
3. Determination of parameters Determination of the
   set of parameters needed to be conveyed within
   each of the messages defined in the previous step
4. Message length calculation Calculation of the
   length of every message considered as deriving
   from the summation of its constituent parameters
   length values.

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
13
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

14
Application to an indicative scenario

• Application of the above defined methodology to
  an indicative scenario, namely New Spectrum
  Assignment
• New frequencies are disposed by a regulator to a
  network operator (NO)
• The NO want to operate the new frequencies for
  specific Radio Access Technologies (RATs)
• The scenario includes processes and exchange of
  messages that are necessary for informing about
  and accommodating the new spectrum to the network
  (network elements and terminals)

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
15
1. Determination of procedures

• 10 procedures in total

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
16
2. Determination of messages (cont.)

• 10 procedures in total
• 15 messages

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
17
2. Determination of messages (cont.)

1. DSM first requests updated spectrum utilization
   metrics from DSNPM (msgSpectrumUsageRequest)
2. DSNPM then requests from n-JRRM, updated context
   information (msgContextInfoRequest_MJ)
3. n-JRRM forwards the context information related
   requests to the respective entities in MTs,
   namely m-JRRMs, as well (msg ContextInfoRequest_J
   J-TN)
4. The context information is sent from MTs to
   n-JRRM (msg ContextInfoResponse_JJ-TN)

1
2
3
4
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
18
2. Determination of messages (cont.)

1. n-JRRM requests from RATs, context and current
   configuration information, as well (msg
   ContextConfigurationInfoRequest)
2. This context configuration information is sent
   to n-JRRM (msgContextConfigurationInfoResponse)
3. The total information is forwarded to DSNPM (msg
   ContextInfoResponse_MJ)
4. Based on this information DSNPM calculates the
   current spectrum usage metrics and forwards them
   to DSM (msg SpectrumUsageResponse)
5. DSM implements algorithms that result to the new
   spectrum assignment and sends the new directives
   for final allocation of the frequencies to DSNPM
   (msgSpectrumAssignmentRequest)

5
6
7
8
9
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
19
2. Determination of messages (cont.)

1. DSNPM decides for the final allocation of
   frequencies and informs a number of RATs (n-CCM)
   for reconfiguration (msg ReconfigurationRequest_M
   C)
2. n-CCM entities determine the specific moment of
   reconfiguration by forwarding the DSNPMs
   information to the RATs (msg ReconfigurationReque
   st_CR) and implement the reconfiguration process

10
11
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
20
2. Determination of messages (cont.)

1. n-CCM supervises the reconfiguration process and
   when it is informed by RATs for the successful
   processs effectuation (msg ReconfigurationRespon
   se),
2. it informs DSNPM (msg ReconfigurationExecutionNot
   ification)
3. DSNPM sets out a context notification procedure
   in order to inform MTs for the updated context
   information in the network (msg
   NetworkContextNotification)
4. Finally, DSNPM derives new policies and sends
   them to MTs via n-JRRM (msg Policy)

12
13
14
15
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
21
3. Determination of parameters

• Next step describe the type and number of
  parameters that each of the messages must convey
  for satisfying the purpose for which the
  architecture has been designed for i.e. spectrum
  and radio resource usage optimization
• Mainly based on the authors view and experience,
  albeit in alignment with the respective
  functionality in each of the functional entities
• Consider msg ContextInfoResponse_JJ-TN as an
  example. It may comprise information such as
• MT identification number
• Reconfiguration capability
• Current configuration of its interfaces
• Geographic coordinates at a time instance
• Mobility profile
• Requested services
• Signal measurements
• etc

4
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
22
4. Message length calculation

• A more formal description of the parameters and
  accordingly of the messages and procedures seems
  to be of prominent importance
• Use of Abstract Syntax Notation One (ASN.1) in
  order to describe the syntax of the messages
  conveyed between the interfaces of the
  architecture in a formal way
• ASN.1 is a standardised specification language
  that describes data structures for representing,
  encoding, transmitting, and decoding data
• Calculation of the length of every message as a
  summation of its constituent parameters length
  values
• Calculations are done on top of and independently
  of any specific transport protocol

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
23
4. Message length calculation (cont.)
Procedure
Message
Parameters
Interface
Calculated Load w/o encoding
Calculated Load assuming BER encoding
Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
24
4. Message length calculation (cont.)

• Interesting note 1
• The evaluation work reveals that the loads obey
  some generic formulas comprising a combination of
  both variable and constant parts
• The variable part reveals dependency on
  parameters, which
• are specific to each procedure e.g. number of
  requested frequency bands f
• or more generic ones e.g. such as the number of
  mobile terminals m (like in ContextInfoResponse_JJ
  -TN)

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
25
4. Message length calculation (cont.)

• Interesting note 2
• One of the most interesting contributions of this
  work is the formalization of the Policy message
• Policies are derived by the network (a policy
  derivation function in DSNPM) and are
  communicated to the user mobile terminals for
  guiding their selection of proper radio resources
  (radio resource selection policies)
• Policies are formatted (using ASN.1) so that they
  apply to the whole set (broadcast-like) or a
  sub-set (multicast-like) of MTs.
• They are formatted as rules adhering to the well
  known Event-Condition-Action (ECA) type
• ON ltEventgt IF ltConditiongt THEN ltActiongt
• the event part specifies the signal that triggers
  the invocation of the rule (e.g. changes in
  spectrum)
• the condition part is a logical test that if
  evaluates to true (e.g. MTs located at specific
  areas and use specific type of service), causes
  the action to be carried out,
• the action part consists of the actual execution
  of the modification/update (e.g. change operating
  RAT/freq etc.)
• Calculation of the Policy message length by
  applying the same methodology

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
26
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
27
Test cases and results

• Test case 1 Generic evaluations
• Process of evaluations of the signaling load by
  assuming a generic situation

Number of active mobile terminals 50
Number of Flexible Base Stations (FBSs) 3
Number of RATs 2
Total produced signaling load (in bytes)
Air and core signaling loads (in bytes)
Signaling load (in bytes) per procedure
Signaling load (in bytes) per interface
28
Test cases and results (cont.)

• Test case 2 Scalability issues
• How the signaling load evolves in function to the
  number of RATs, of FBSs and of MTs in the managed
  network?

Evolution of signaling load vs number of FBSs
Evolution of signaling load vs number of RATs
Evolution of signaling load vs number of MTs
29
Test cases and results (cont.)

• Test case 3 Signaling delays
• The objective is to give some evidence on the
  delay that the management operations will suffer
  as a result of the transmission of the produced
  management signaling information
• Assumed wired and wireless links offering 100Mbps
  and 70Mbps of capacity, respectively (i.e. S1-MME
  and air interfaces in 3GPP LTE)

Evolution of signaling delay vs number of FBSs
Evolution of signaling delay (in ms) vs number of
RATs
Evolution of signaling delay vs number of MTs
30
Outline

• Motivation, problem area
• Overview of considered cognitive management
  architecture
• Signaling load evaluation methodology
• Application to an indicative scenario
• Test cases and results
• Conclusions and future work

31
Conclusions and future work

• Conclusions
• The heterogeneity and versatility of future
  wireless networks resulted in a cognitive
  management architecture for offering optimized
  management
• A methodology for evaluating signaling loads in
  this management architecture was presented
• Results that were obtained from the application
  of the methodology to an indicative scenario were
  presented and showed that the management
  architecture will not aggravate the overall
  network operation
• Future work
• Handling the complete set of scenarios
• Identification of the periodicity of specific
  procedures-messages in order to give insight on
  the expected load that will regularly appear in
  the managed network (already done)
• Mapping to existing transport protocols that
  currently used for signaling purposes
• Setup an experimentation platform for validating
  the proposed methodology
• already done based on Java Agents DEvelopment
  (JADE) framework

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
32
Further reading

• K.Tsagkaris, M.Akezidou, A.Galani, P.Demestichas,
  Evaluation of Signalling Loads in a Cognitive
  Network Management Architecture, submitted,
  available under request
• A.Galani, K.Tsagkaris, P.Demestichas,
  Information Flow for Optimized Management of
  Spectrum and Radio Resources in Cognitive B3G
  wireless networks, Journal of Network and
  Systems Management, Springer, Vol. 18, Issue 2,
  pp. 125-149, June 2010
• A.Galani, K.Tsagkaris, N.Koutsouris,
  P.Demestichas, Design and Assessment of
  Functional Architecture for Optimized Spectrum
  and Radio Resource Management in Heterogeneous
  Wireless Networks, International Journal of
  Network Management, Wiley, to appear
• K.Tsagkaris, N.Koutsouris, A.Galani,
  P.Demestichas, Performance Assessment of a
  Spectrum and Radio Resource Management
  Architecture for Heterogeneous Wireless
  Networks, in Proc. Future Network Mobile
  Summit 2010, Florence, Italy, 16th -18th June
  2010
• K.Nolte, A.Kaloxylos, K.Tsagkaris et al. The E3
  architecture Enabling future cellular networks
  with cognitive and self-x capabilities,
  International Journal of Network Management,
  Wiley, to appear
• G.Dimitrakopoulos, P.Demestichas, A.Saatsakis,
  K.Tsagkaris, A.Galani, J.Gebert, K.Nolte,
  Functional Architecture for the Management and
  Control of Reconfigurable Radio Systems,
  Vehicular Technology Magazine, IEEE, vol.4, no.1,
  pp.42-48, March 2009
• ETSI TR 102.682 Reconfigurable Radio Systems
  (RRS) Functional Architecture (FA) for the
  Management and Control of Reconfigurable Radio
  Systems, V1.1.1 (2009-07)

Wireless Track Session 2 BROADNETS 2010 27
October 2010 October 25-27, Athens, Greece
33
Acknowledgment

• This work was performed in the project E³
  (www.ict-e3.eu) which has received research
  funding from the Community's Seventh Framework
  programme. The work is evolved in the context of
  UniverSelf and OneFIT (www.ict-onefit.eu
  lthttp//www.ict-onefit.eu/gt) Projects. This paper
  reflects only the authors' views and the
  Community is not liable for any use that may be
  made of the information contained therein.

34

• Thank you
• Questions?

Contact details Dr. Kostas Tsagkaris University
of Piraeus Department of Digital Systems E-mail
ktsagk_at_unipi.gr