COMP9334: Capacity Planning for Computer Systems and Networks

1
COMP9334 Capacity Planningfor Computer Systems
and Networks

• Week 1 Introduction
• Lecturer Dr Bruce Howarth

2
Introduction

• This subject
• Why performance is important
• Overview of capacity planning
• Review Web protocols

3
What is capacity planning?

• Determine desired service levels
• Measure actual workload
• Measure actual performance
• Tune as necessary
• Create performance models
• Predict workload growth
• Predict capacity required to maintain service
  levels

4
Comments

• All these steps are hard to do
• All have serious costs
• Tools, time, gaining expertise
• Lots of politics and negotiation
• Lots of compromises and tradeoffs

5
Related activities

• Software performance engineering
• Designing performance into applications
• Better than designing it out
• Technology assessment
• Examining new and coming technologies
• Consider how they affect ability to deliver
  required services

6
Why worry about performance?

• Slow systems lose money
• Employees waste time waiting
• Customers will not buy if website too slow
• Theyll go elsewhere
• Performance is hard to control
• Especially in Web applications
• Many delays are outside your organization

7
Why worry (cont.)?

• Systems are complex, hard to understand
• N-tier systems
• Multiple servers networked together
• Complex storage subsystems
• RAID, SANs, etc
• Can be hard to identify cause of performance
  problems

8
Why worry (cont.)?

• Throwing hardware may not work
• Especially if its the wrong hardware
• E.g., a faster CPU wont solve a disk problem
• Or the software uses slow algorithms
• E.g., linear search vs index
• Extra hardware can have hidden costs
• Software license fees, power, cabling ...

9
Performance Examples

• Menascé gives typical cases in Ch. 1
• Lost revenue from an overloaded Web server
• Sizing a distributed client-server system
• Uses the tools described later
• The problems are quite realistic

10
System Architectures - review

• We assume you know about
• Basic computer organization
• Memory hierarchy, virtual memory
• Process scheduling
• Round robin, context switch
• Process priority structures
• Basic concepts of client-server systems
• Some of this material is reviewed in Menascé

11
Communications basics - review

• We assume you know about
• Basic data transfer
• LANs and WANs
• Protocol concepts
• TCP/IP, HTTP, and similar
• Much of this material is reviewed in Menascé

12
Not in Menascé

• Storage area networks
• Switched ethernet
• IEEE 802.11 (wireless ethernet)
• (But same models as basic IEEE802)
• What else?

13
Protocols and performance

• Menascé describes how TCP takes a while to
  initiate a connection
• Also how HTTP adds overhead to exchanges
• You need to understand how these work and the
  magnitudes of the effects

14
Web service protocols

• Menascé introduces the likely future of Web
  services
• SOAP - simple object access protocol
• WSDL - web services description language
• UDDI - universal description, discovery and
  integration