270 likes | 293 Views
CompSci 296.2 Self-Managing Systems. Shivnath Babu. Motivation. Systems are becoming hard to manage Increasing size (both software and hardware). Clients. WAN. Web server. Application servers. Database servers. Motivation. Clients. WAN. WAN. Web server. Application servers.
E N D
CompSci 296.2 Self-Managing Systems Shivnath Babu
Motivation • Systems are becoming hard to manage • Increasing size (both software and hardware)
Clients WAN Web server Application servers Database servers Motivation
Clients WAN WAN Web server Application servers Database servers WAN WAN Motivation WAN
Motivation • Systems are becoming hard to manage • Increasing size (both software and hardware) • Increasing heterogeneity (e.g., Grid systems) • 24 x 7 operation • 5 nines availability (system is down at most 5 minutes and 15 seconds per year)
Downtime Costs (per Hour) • Brokerage operations $6,450,000 • Credit card authorization $2,600,000 • Ebay (1 outage 22 hours) $225,000 • Amazon.com $180,000 • Package shipping services $150,000 • Home shopping channel $113,000 • Catalog sales center $90,000 • Airline reservation center $89,000 • Cellular service activation $41,000 • On-line network fees $25,000 • ATM service fees $14,000 Sources: InternetWeek 4/3/2000 + Fibre Channel: A Comprehensive Introduction, R. Kembel 2000, p.8. ”...based on a survey done by Contingency Planning Research."
Motivation • System administration cost is increasing
Motivation • System administration cost is increasing • Recently, $1 storage $9 administration cost [Fujitsu] • Up to 75% of overall database ownership cost is for administration [Aberdeen] • Up to 80% of Information Technology (IT) budgets spent on maintenance [McKinsey]
Motivation • System administration time & effort is increasing
Causes of system crashes Other 18% Systemmanagement 53% % of System Crashes Softwarefailure 18% 10% Hardwarefailure Time (1985-1993) Motivation • System administration time & effort is increasing • >40% of computer system outages caused by operator/administrator error
Global Storage Service Site Failures Unknown Hardware 0% 9% 41% 28% SW Human Network 22%
Motivation • System administration time & effort is increasing • >40% of computer system outages caused by operator error • System is too difficult to understand • Decisions need to be made quickly, under pressure • Not enough well-trained operators • Changes are frequent • E.g., workload, hardware, people, data
The Real Problem … The obstacle is complexity … Dealing with it is the single most important challenge facing the IT industry Paul Horn, Director of Research, IBM
The Solution • Let the system deal with the complexity of management • Computer-science-wide push towards Self-Managing Systems • IBM calls this new field Autonomic Computing
Autonomic Computing (IBM) Computer systems that can regulate themselves much in the same way as our autonomic nervous system regulates and protects our bodies Paul Horn, Director of Research, IBM
Autonomic Nervous System • Tells you heart how fast to beat, checks your blood’s sugar and oxygen levels, and controls your pupils so the right amount of light reaches your eyes as you read these words, monitors your temperature and adjusts your blood flow and skin functions to keep it at 98.6ºF • Is autonomic -you can make a mad dash for the train without having to calculate how much faster to breathe and pump your heart, or if you’ll need that little dose of adrenaline to make it through the doors before they close
What will we do in this class? • Read research papers • Listen to guest lectures • Goal of the class: Give structure to this field, e.g., • Concretely defining problems that arise in this setting • Identifying algorithms and techniques useful in this domain • Proposing guidelines for designers of future systems and software • Semester-long project
Outline • Part 1: Motivating Factors, Problems, and Applications • From Internet services, database management, computational grids, weather analysis and prediction, oil reservoir optimization, and others • Part 2: Algorithms and Techniques • Control theory, machine learning, performance modeling, stochastic optimization, massive data management, data integration, building blocks in systems, and others • Part 3: Putting everything together, implications, and future work
Evaluation • Class participation 25% • Project 75%
Resources • Google keywords • Autonomic computing • Self-managing systems • IBM autonomic computing web page • IBM Journal special issue on autonomic computing • Berkeley ROC project
In the next class • Read an overview paper on self-managing systems • Summary of work in this area • Sample projects