180 likes | 301 Views
Policy-Based QoS Management with XML. Ntanzi Carrilho Neco Ventura Department of Electrical Engineering University of Cape Town. Presentation Outline. Introduction Related Work Problem Description Research Objectives Proposed PBNM framework Performance Study Conclusions. Introduction.
E N D
Policy-Based QoS Management with XML Ntanzi Carrilho Neco Ventura Department of Electrical Engineering University of Cape Town
Presentation Outline • Introduction • Related Work • Problem Description • Research Objectives • Proposed PBNM framework • Performance Study • Conclusions
Introduction • What is Policy-based QoS Management? • Rules governing the state of network • Automates execution of management objectives • Simplifies administration • Decentralises network management • Enable seamless QoS management
Related Work • PBNM Architectures • IETF Policy Framework • 3GPP Policy Control Architecture (IMS) • Policy Information/Data Models • DENng (TMF) • PCIM/PCIMe (IETF/DMTF) • QPIM (IETF/DMTF) • DiffServ PIB (IETF) • QoS Policy Provisioning Protocols • COPS-PR / SPPI (IETF)
Problem Description • Wide range of policy representation formats involved in current PBNM solutions • Unnecessary complexity to the structure of systems • Manageability of PBNM systems • Traditional standards and protocols are complex but inefficient (COPS-PR and SPPI) • SPPI complex but not flexible • COPS-PR messages are not human readable • Limited availability of parsers capable of processing them • Increased development and maintenance costs
Research Objectives • Use XML to develop a PBNM system • XML is a widely deployed standard supported by a large number of applications • Human and machine readable format • Widely available tools to process XML • Allows the description of complex structured data • Easy to learn, easy to process • Objectives: • Create an XML-based QoS Policy Model • Use XML in entire policy life cycle • Use XML-based protocol for QoS policy provisioning – IETF Network Configuration protocol (NETCONF) • Study the impact of design choices on performance
Proposed QoS Policy Model • Independent of any specific device • Independent of QoS mechanisms • Independent of device technology • Dependent of QoS mechanisms • Dependent of device technology • Dependent of QoS mechanisms
Proposed QoS Policy Model “User Ntanzi receives high QoS for internet traffic” If (IP = 10.128.1.91) and (port = 80) then (high QoS actions)
Proposed QoS Policy Model • Network-Level Policy example <PolicyGroup Name="edge-policiesGroup" PolicyRoles="edge"> <PolicyRule Name="EFedge-policyRule"> <PolicyCondition> ... </Policycondition> <CompoundPolicyAction Name="EFedge-actions"> <PolicyActionInPolicyAction> <xi:include href="/db/QoSPolicies/ReusablePolicies/ QoSPolicyPoliceAction.xml#xpointer(//QoSPolicyPoliceAction[ @Name='EFedge- policer'])"/> </PolicyActionInPolicyAction> <...> ... </...><!-- Other PolicyActions --> </CompoundPolicyAction> </PolicyRule> <...> ... </...><!-- Other PolicyRules --> </PolicyGroup>
Proposed QoS Policy Model • Device-Level Policy example (DiffServ) <frwk-pib:IpFilterTableContainer> <frwk-pib:IpFilterTable BaseFilterPrid="3" IpFilterDscp="101110"/> </frwk-pib:IpFilterTableContainer> ... <ds-pib:ClfrElementTableContainer> <ds-pib:ClfrElementTable ClfrElementPrid="1"> <ds-pib:ClfrElementSpecific Pointer="//frwk-pib:IpFilterTable[@BaseFilterPrid=3]"/> <ds-pib:ClfrElementNext Pointer="//ds-pib:MeterTable[@MeterPrid=1]"/> </ds-pib:ClfrElementTable> </ds-pib:ClfrElementTableContainer> <ds-pib:MeterTableContainer> <ds-pib:MeterTable MeterPrid="1"> ... </ds-pib:MeterTable> </ds-pib:MeterTableContainer>
Performance Evaluation Setup • Performance parameters • Network usage • Impact of compression • System resource usage • Processing delays • Memory consumed
Performance Results • Network usage
Performance Results • System resource usage at PEP Processing delays
Performance Results • System resource usage at PEP Dynamic memory
Conclusions • Benefits brought by the use of XML • Reduced development costs • Simplify the structure of Policy models • Added flexibility • Open up possibilities of integration • Proposed system was successfully implemented and validated • Performance results are satisfactory