Multi-Tier SLA Management and Standards

1. Multi-Tier SLA Management and Standards QoS/SLA WorkshopOGF 28, Munich, March 17, 2010 Ramin Yahyapour

2. Example: SLA@SOI Project • Vision • A business-ready service-oriented infrastructure empowering the service economy in a flexible and dependable way. • Business-readiness requires • predictability & dependability  prerequisite for acceptance & uptake of (new) services • holistic SLA management  transparent IT management • automated negotiation  dynamic, scalable service consumption • Impact on the knowledge economy • decreased time to market for new services • increased productivity and competitiveness • lower entry barriers, especially for SMEs

3. Technical Motivation • Service Consumer • dynamic demand for complex business solutions at low costs Flexible usage Business Services Visionof SLA@SOI A business-readyservice-oriented infrastructure empowering theservice economy in a flexible anddependable way • Software Provider • SOAs provide un-precedented flexibility Engineering of predictable services • Service Provider • service economy requiresdependable services Automated SLA negotiation and management • Infrastructure Provider • virtualization technologiesallow for adaptive SOIs SLA enforcement via adaptive infrastructures

4. SLA@SOI Project • Duration • June 2008 – May 2011 • 12 Partners • 6 industrial, 1 SME, 3 academic, 2 research centres • 7 countries: Austria, Germany, Ireland, Italy, Slovenia, Spain, United Kingdom • Budget • 15.2 M€ • Info • http://www.sla-at-soi.eu

5. Envisioned Interaction Customer Service Provider BusinessUse Service Demand Contracting/Sales Procurement SOA Infrastructure Provider Software Provider virtual SOI physical

6. Envisioned Interaction Customer Service Provider BusinessAssessment BusinessUse Service Demand SLA Contracting/Sales SLA (Re-)Negotiation SLA Orchestration/Transformation/Aggregation Monitoring, Arbitration Procurement BusinessAssessment SOA Service Demand Forecasting Infrastructure Provider Software Provider Provisioning virtual Resource Consumption Forecasting Mapping SOI physical

7. Envisioned Interaction Customer Service Provider BusinessAssessment BusinessUse Service Demand SLA Contracting/Sales SLA (Re-)Negotiation SLA Orchestration/Transformation/Aggregation Monitoring, Arbitration Procurement BusinessAssessment SOA Service Demand Forecasting Monitoring Adjustment Alerting Infrastructure Provider Software Provider Provisioning virtual Resource Consumption Forecasting Mapping SOI physical

8. Project innovation • SLA management framework • harmonizing perspectives of relevant stakeholders (software/service/infrastructure provider and customer) • standards for SLA specification and negotiation & systematic multi-layer SLA management (planning, optimization, and provisioning), monitoring and accounting • guaranteed QoS in a dynamic and end-to-end fashion via consistent SLA handling across IT stack • Adaptive SLA-aware infrastructures • standardized interfaces for adaptive infrastructures with harmonized access to different virtualization technologies. • advanced technologies for SLA enforcement on infrastructure level • efficient resource usage w/ reliable SLA enforcement at infrastructure level • Engineering methods for predictable service-oriented systems • modelling techniques and prediction tools for SOA and SOI components • Business management suite for e-contracting • covers complete business lifecycle of a service provisioning/delivery

9. Example real world use-case ... • Multiple service types hosted • Communications, • Social networking, media, entertainment, • Enterprise. • On a heterogeneous infrastructure : • Data centres, grid, • Wired, wireless. • With associated diversity of : • Margin and profitability, • Customer expectations and satisfaction requirements, • SLA ‘levels’ – platinum … bronze … etc, • Workload patterns and scheduling / provisioning requirements.

10. Additionally .. planned unforeseen • Constraints • Internal governance • Logging for billing, • Auditability, • Licensed product management. • Internal efficiency • Managed resource consumption, • Utility cost minimisation, • Operational cost minimisation, • And third party resources and services. • External events • ‘Slashdot effect’, • Resource failure, • SLA adjustments.

11. Service Level Agreements Complex provisioning challenge • Constraints expressed in Service Level Objectives • ‘Make composite resource available between 09:00 and 11:30.’ • ‘Complete processing of a daily data transformation by 14:00.’ • ‘Sustain average throughout of 3000 transactions/hour.’ • ‘Peak throughput of 150 transactions/minute for up to 10 minutes.’ • Statistical constraints • ‘deliver 99.98% availability’ • ‘ensure <1% transaction timeouts’ • ‘ensure 90% transactions complete within 1sec.’

12. SLA Lifecycle view Diagram : SLA@SOI internal reference • Normal lifecycle • Negotiation  planning/optimisation  provisioning  execution • Monitoring provides a feedback loop for adjustment at run time.

13. Negotiation - provisioning phase Internal policies and constraints Customer facing SLA Third party SLA #1 Third party SLA #2 Internal resources • Implications from service level objectives – • Selection and configuration of individual resources based on performance and cost. • Statistical likelihood of SLA compliance with end customer. • Calculation and expression of SLOs, negotiation of SLAs with third party providers. • Final orchestration and provision of service to end customer.

14. Run time phase Diagram : SLA@SOI Use Case Specification - Enterprise IT. • Reactive adjustment • Recovery from failure. • Adjustment – • Full re-provisioning of any part of the service chain followed by re-orchestration. • Proactive adjustment • Predictive analysis of monitoring data. • Adjustment – • Dynamic resource allocation to any part of the service chain. • Full re-provisioning of any part of the service chain followed by re-orchestration.

15. Experiences • QoS/SLA Management are crucial in many scenarios • Current trend in different areas • Fueled by Cloud scenarios: • SLAs for services, platforms, infrastructures,compute, software, licenses, data, network • Interoperability is a key requirement • While single QoS/SLA are current goals, • However, immediately combination of SLAs in a complete stack will become important (Offering a SaaS which requires several infrastructure or other components • Data center management moves from system management to service management. • A single service is not individually management but needs to be evaluated in global view: SLAs seem to be the common trend to do so. 15

16. Standards • Interoperability requires open standards. • Existing core: • WS-Agreement • OCCI • OGSA-BES • Open challenges: • Multi-layer, multi-tier SLA management • Linking SLAs, Translation of SLAs • Extending of RM systems by SLA features 16

17. Current State • WS-Agreement • Specification GFD.107 available for a long time; many implementations • Urgent need for microspecs for common SLAs • Extension for multi-round negotiation (current activity) • JSDL • Specification GFD.056,GFD.111 well adopted in the community • Extensions available • DRMAA • Programming API for the submission and control of jobs to one or more Distributed Resource Management (DRM) systems (GFD.022,GFD.130). • OCCI • Management of Cloud Systems • Supporting SLAs as input (by SLA@SOI and RESERVOIR projects) 17

18. OCCI and SLA@SOI • SLA@SOI & OCCI have common goals • “Standardisedinterfaces for adaptive infrastructures with harmonized access to different virtualization technologies.” • OCCI is means for SLA@SOI to: • Embed the importance and concepts of SLAs • Influence infrastructural standards • OCCI will be used in-part to SLA-enable infrastructure management • WP A4 (Adaptive SLA-aware Infrastructure Management) is the interested work package • In development. Will implement an OCCI compliant interface • In terms of the SLA@SOI Framework this means: • Supply OCCI-specific SLA@SOI Framework plugins

19. WS-Agreement and SLA@SOI • SLA@SOI is interested in and will contribute a negotiation extension to WS-Agreement • As part of the architecture update, a rule-based protocol orchestration engine has been designed • Results to come • SLA@SOI-specific SLA negotiation requirements (deliverable input) • Evaluation of negotiation protocols (simulation & report) • Definition of the protocol • Protocol implementation (software) • Implementation of the orchestration engine 19