Seminar: Hochskalierbare Betriebliche Datenbankanwendungen

1
Adaptive Service Management

• Seminar Hochskalierbare Betriebliche
  Datenbankanwendungen

2
Agenda

• Introduction
• Aspects of Adaptive Infrastructure
• Models and Approaches
• Summary

3
Introduction

• Motivation
• Objectives concerning SOA Administration
• Automatic management in an infrastructure
• AutoGlobe

4
Motivation (1)

• TCO (total cost of ownership) of IT increasingly
  dominated by people cost (operational,
  administration)
• Distributed application services in network (SOA)
• Complexity

5
Motivation (2) - Dealing with Complexity

• When does which service cause which load?
• How are services integrated?
• How much load is caused by an additional user in
  the system?
• How to recognize an exceptional situation?
• How to respond to the exceptional situation?
• (Expensive) Administrators are required!

6
Objectives concerning SOA Administration
7
Objectives concerning SOA Administration
?
1
2
?
?
?
?
High-scaled hardware architecture
?
?
3
4
?
?
?
?
Assign services to available computing resources
in an optimized way
Automatic administration
8
Automatic management in an infrastructure

• Self-organizing, autonomic, automatic
  Self-management capabilities including
• self-configuration,
• self-optimization,
• self-healing,
• and self-protection.
• Allocate services in an optimized way, according
  to requirements, automatically.

9
AutoGlobe (1)

• Research prototype of Adaptive Computing
  Infrastructure (http//www-db.in.tum.de/research/p
  rojects/AutoGlobe/)

10
AutoGlobe (2)

• Conceals the complexity of IT-Management
• Virtualizes, monitors, pools hardware
• Uses a fuzzy-logic based controller to monitors
  running services and performs appropriate actions
  in an exceptional situation.

11
Aspects of Adaptive Infrastruture

• Storage Virtualization
• High Scalability
• Service Virtualization
• Quality of Service
• Service Level Agreement (SLA)
• Automatic Administration

12
Storage Virtualization

• Definition Logical abstraction of physical
  storage systems.
• Hides complexity of physical storage devices and
  their specific requirements from management view.
• Example SAN (Storage Area Network) Linking
  servers to disk arrays, tape backup systems,
  other devices.

13
High Scalability

• Help customers reduce costs while simplifying
  change.
• Example
• Blade server environment
• (an infrastructure in a box
• proven to save time, power
• and money in data centers).

14
Service Virtualization

• Problem Uncertainty of the reliability and
  usability of resources or services invoked by an
  application.
• Service Virtualization request for each
  virtualized service transferred dynamically to
  most suitable service instance.

15
Quality of Service

• Ensuring Quality of Service in database
  application
• Example
• QoS in query processing Is the result set
  sufficiently large?
• Related to the large number of requests to
  services in a system.

16
Service Level Agreement (SLA)

• Formal agreement between service provider and a
  customer.
• Imposes obligations on the service provider
  regarding the service level of the provided
  services.

17
Automatic Administration

• Includes the autonomic and self-organizing
  capabilities.
• Conclusion Important!

18
Models and Approaches

• General Architecure
• Load Analysis and Static Resource Management
• Dynamic Resource Management
• Adaptive SLA Enforcement

19
General Architecture (1)

• Composed of
• Static resource management
• Dynamic resource management
• Adaptive SLA enforcement
• A service model is required!

20
General Architecture (2)

• Service Model

21
Load Analysis and Static Resource Management

• Service Load Patterns
• Static Resource Management

22
Service Load Patterns

• Consists of 3 phases
• Preprocessing Server and service instances
  monitored, the load logged to an archive.
• Analysis Extraction of load pattern assuming
  that the pattern is periodic.
• Classification Separating time series with
  periodical characteristics from those without.

23
Static Resource Management (1)

• Allocation of services (with various load
  pattern) onto available computing resources
  greedy heuristics.
• Allocation of Non-Scale-Out Capable vs Scale-Out
  Capable Services.

24
Static Resource Management (2)
25
Dynamic Resource Management

• Achieving QoS
• Architecture of Dynamic Controller
• Fuzzy Controller
• Dynamic Load Balancing Load Forecasting
• Dynamic Load Balancing Hints

26
Achieving QoS

• Analytical Performance Model determines the
  required resource capacities needed to accomplish
  given service level objectives.
• Monitoring Based Systems continuous monitoring
  of various performance metrics.

27
Architecture of Dynamic Controller

• Monitors and Advisor Modules
• Monitoring System
• Fuzzy Controller
• Load Data Archive

28
Fuzzy Controller (1)

• Based on fuzzy logic.
• Controller has a set of rules analyze the rules,
  take an appropriate action.
• The second controller chooses the suitable host
  for a service instance if required.

29
Fuzzy Controller (2)

• IF cpuLoad IS high AND (performanceIndex IS low
  OR
• performanceIndex IS medium) THEN scaleUp IS
  applicable
• IF cpuLoad IS high AND performanceIndex IS high
  THEN
• scaleOut IS applicable

30
Dynamic Load Balancing Load Forecasting

• Reacting dinamically on exceptional situations
  need a period of time.
• Historic load data analyzing recognizes pattern.
• Use proactive
• controller.

31
Dynamic Load Balancing Hints

• Based on hints specified by administrators.
• Example specify the necessary number of service
  instances running during the course of a working
  day.

32
Interaction between Static and Dynamic Resource
Management

• DRM monitors services and servers, stores the
  load data.
• The historic load data is evaluated, service
  specific load patterns are determined.
• SRM uses the knowledge to optimize the allocation
  and gives administrators recommendations.

33
Adaptive SLA Enforcement

• Quality of Service is more relevant for a
  customer than the optimal utilization of system.
• Another important parameter Service Level
  Agreement (response time, etc)
• Technique Scheduling request, adaptive
  penalization.

34
Summary

• To balance the objectives concerning SOA
  administration Adaptive Service Management.
• Covering of aspects in adaptive infrastructure.
• Approaches Static, Dynamic, Adaptive SLA
  Enforcement.

35
Thank you for your attention!Questions