May 7-8, 2007ICVCI 2007 RTP Autonomic Approach to IT Infrastructure Management in a Virtual Computing Lab Environment H. Abdel SalamK. Maly R. MukkamalaM. Zubair Department of Computer Science, Old Dominion University D. Kaminsky IBM Research Triangle Park

May 7-8, 2007ICVCI 2007 RTP Outline Introduction –Problem Change Management Reservation Management –Motivation Approach –Autonomic (Policy Based) –Infrastructure Aware Policy Engine Integrated System –Loosely Coupled –Tightly Coupled Future Work

May 7-8, 2007ICVCI 2007 RTP IT Management Complexity Managing IT infrastructure for a Virtual Computing Laboratory (VCL) is a challenging task. We need to satisfy the requirements (often conflicting) for two major groups of users: Users of VCL High availability Flexibility in reservation (prioritization, etc.) VCL & IT Administrators High End User Satisfaction Maintain and upgrade the underlying IT infrastructure Minimal disruption Minimal administrative support

May 7-8, 2007ICVCI 2007 RTP Conflicting Requirements - Scenario A teacher has reserved a computer running an application for an entire Saturday. The application itself depends on a web server, and two databases, all running on different servers. Network resources to be used by teacher’s application along with the connectivity.

May 7-8, 2007ICVCI 2007 RTP Scenario…. A system administrator has received an upgrade request for a subset of routers that will need to be taken down and proposes to do so on that weekend Taking Routers C and D down would disable the application.

May 7-8, 2007ICVCI 2007 RTP How to Handle Conflicting Requirements? Small Scale Human Intervention: Before a reservation is committed, check for any conflict with the change management schedule (router upgrade) and vice versa. Identify conflict and resolve it. The resolution may imply changing one of the schedules to avoid conflict. Large Scale For large IT infrastructure and large number of VCL users, Human Intervention approach is not feasible. We need autonomic approach to IT Infrastructure Management in a Virtual Computing Lab Environment

May 7-8, 2007ICVCI 2007 RTP Approach Autonomic Approach: Minimal human intervention. How ? Policy based IT infrastructure management We need to address two interacting management tasks Change Management (Upgrade software, Security patch,..) Reservation Management ( Reserve a VCL resource )

May 7-8, 2007ICVCI 2007 RTP Infrastructure Aware Policy Based System Infrastructure-aware Policy Engine Policy language constructs to represent dependencies among applications, hardware and software XML representation of Infrastructure Our group has been working for the last one year on developing Infrastructure aware policy based system for change management

May 7-8, 2007ICVCI 2007 RTP Traditional Policy Engine Architecture Underlying Infrastructure Policy Deployer Policy Engine Policy1.xml Policy Evaluator If (Condition-1) Then Decision-1 If (Condition-2) Then Decision-2. If (Condition-k) Then Action-1 If (Condition-k+1) Then Action-2. Properties (Sensors): Property-1(Read-Only) Property-2(Read-Write). Getter-Setter Methods: Property-1.Get() Property-2.Set(value). Action Methods: Action-Handler-1() Action-Handler-2(). Policy2.xml Solicit Decision or Action Decision Action Conditions can contain Infrastructure independent operators & vocabularies only. Client (Managed Resource) Infrastructure information is limited to what the client offers to the engine through its properties or methods.

May 7-8, 2007ICVCI 2007 RTP Infrastructure Aware Policy Engine The policy language is able to support infrastructure dependent operators (e.g., IsAvailable, IsConnected)

May 7-8, 2007ICVCI 2007 RTP Infrastructure Aware Operators for Complex Dependencies Dependency between applications and hardware in a large IT infrastructure can be complex. OR Dependencies: Using of Database replicas to improve service availability. Access Type Dependencies: Many Applications offer different features based on the user privileges; these features have different dependency requirements

May 7-8, 2007ICVCI 2007 RTP Proposed Integrated System Combines reservation system with change management system Designed to work with legacy systems---existing reservation systems may be combined with a newly developed change management and vice versa. Conflicts between reservation requests and change management requests are resolved via resolution policies

May 7-8, 2007ICVCI 2007 RTP Model 1: Loosely Coupled

May 7-8, 2007ICVCI 2007 RTP Salient Features of Model 1 Works with legacy systems Reservation module (RM) makes scheduling decisions for reservation requests Critical reservations are converted into constraints and provided as feedback to change management module (CMM) CMM makes decisions for change management requests Critical scheduled changes are provided as feedback to RM Conflict resolution policies (not shown in the model) could be used to resolve conflicting schedules/ requirements between the two modules

May 7-8, 2007ICVCI 2007 RTP Model 2: Tightly Coupled

May 7-8, 2007ICVCI 2007 RTP Salient Features of Model 2 Both reservation requests and change management requests are handled by a single policy engine Relevant when developing systems from the start (rather than interfacing with a legacy system) Policy language should include constructs to express reservations as well as change management requests

May 7-8, 2007ICVCI 2007 RTP Future work Implement a simple reservation system Integrate it with the current change management system (using Model 1) Test the integrated system in a real VCL environment Deploy a VCL node at

May 7-8, 2007ICVCI 2007 RTP Supporting Slides

May 7-8, 2007ICVCI 2007 RTP Sample Policy

May 7-8, 2007ICVCI 2007 RTP Sample IT Management Policy

May 7-8, 2007ICVCI 2007 RTP Sample Infrastructure Representation