Running a 24x7 Internet Data Centre

1
Running a 24x7 Internet Data Centre
Dave Moskovitz Technical Director The Web
Limited May 2001
2
Agenda

• Availability measures in context
• Defining your objectives
• Factors affecting availability
• Service monitoring
• Turning your web site into a 24x7 operation
• Adding a web site to your Data Centre
• Which bits to outsource?
• Summary
• Questions

3
Availability measures in context

• End-to-end availability
• Component uptime
• MTBF Mean time between failures
• MTTR Mean time to repair
• Scheduled vs unscheduled outages
• Performance
• User experience

4
Cost of downtime

• Lost revenue
• Emergency support
• Productivity
• Damaged reputation
• Financial performance

5
Defining objectives

• You cant control what you cant measure
• Set SLOs or SLAs
• Measure and report regularly

6
Availability factors

• Environment
• Network
• Server hardware
• Software
• Business processes

7
Availability factors environment

• Structural
• Physical security
• Power
• Air conditioning
• Emergency services

8
Availability factors network

• Internet connectivity
• Upstream connectivity / ISP peering
• Internal network configuration
• Capacity / Load balancing
• Network security

9
Availability factorsserver hardware

• Capacity / Security
• System redundancy / Hot standby
• Component redundancy
• Multiple CPU
• ECC memory
• Power supplies
• Network adaptors
• RAID
• Big cost difference in approach!

10
Availability factors software

• Capacity
• Security
• Support / Serviceability
• Simplicity
• Documentation
• Backups
• Reliance on external services / suppliers

11
Availability factors business processes

• Change control
• Fault reporting / escalation / resolution
• Knowledge base
• Key staff and suppliers
• Procedural documentation
• Company-wide disaster recovery

12
The Weak Link
Your 24x7 Application is only as strong as its
weakest element
13
Monitoring

• End-to-end monitoring is critical!
• - Use Keynote for performance and availability
  reporting
• Component monitoring
• Heartbeat monitors
• Set up alarms well
• Regular reporting

14
Making your web site available 24x7

• Most sites are already 24x7 except when theyre
  down or performing poorly!
• Start with end-to-end monitoring
• Identify risks and their associated costs
• Spend money proportionally to risk definitely
  cover the easy wins
• Outsource where it makes sense

15
Adding a web site to your 24x7 data centre

• Good! You already have lots of process
• Its a bad bad world out there what else are
  you putting at risk?
• Choose your ISPs (and other vendors) carefully
• Start with end-to-end monitoring
• Identify risks and their associated costs
• Outsource where it makes sense

16
Which bits to outsource?

• End-to-end monitoring
• Call centre
• Network management
• Software maintenance
• The Whole Enchilada?

17
Summary

• End-to-end availability is THE key measure
• Identify risks and their associated costs
• Expect the unexpected
• Review risks on a regular basis
• Spend money proportionally to risk
• Outsource where it makes sense