Day 20

1
Day 20
2
Network/System Administrator

• Plenty of jobs
• Not likely to be outsourced because you need
  physical access to the machines
• Challenging job
• New problems every day, no 2 days alike
• Things tend to grow not shrink
• Always have things to plan for

3
Network vs System Admin

• System Admin typically deals with just the
  desktop computers
• Network Admins typically deal with both the
  network and the servers
• Often these terms are not that cleanly cut.

4
Your job

• In general any job has the following traits
• You should be providing the company as much
  return as you take.
• Sometimes things are difficult to quantify
• Easy to identify things like
• Customers
• Money
• Customer Satisfaction
• If not youll likely find yourself on the
  chopping block

5
Projects

• Generally companies break things up into projects
• Easier to track
• Time
• Expense
• Progress
• Plus people feel like they are getting somewhere
  if a project ends and a new one begins

6
Project Cycle

• Decide what you want to do
• Draw up Requirements Specification
• Design how you are going to do it
• Draw up a Design Specification
• Do it
• Implement your design
• Test and verify your implementation against your
  Requirements

7
If not gt Failure

• If you try to make changes to the requirements
  mid way, your design will probably break
• Overruns in time and cost are typically caused by
  Feature creep

8
System Development Life Cycle

• Planning
• Analysis
• Design
• Implementation
• Maintenance
• Wash, Rinse and Repeat.

9
Planning

• How much of something do you need?
• Do you need 10mb networking? 100mb? Gig
  ethernet?
• Do you need 1 web server? 10?, 100?
• Do you need the 40 switch, or the 5,000 switch
• Overbuy?
• Costs more
• Underbuy
• You may lose your job in a failure
• Or the system may fail

10
You arethe weakest link!

• The weakest link takes down the system
• Often spend all the money on the wrong things
• You need all parts of the system to be of equal
  capability otherwise it all runs as slow as its
  slowest part
• E.g. You buy an OC48 externally, gig networking,
  gig network cards, and have a single hard drive
  in a web server. The hard drive can only
  transfer at 30MB at best.
• Often sexiness determines which is a problem

11
Sometimes you have to guess

• The Aquarium just opened
• Their website failed for 4 days, why?
• Sometimes you can calculate
• If you know that your biggest webpage is 100k,
  and you guess 500 people will hit it per second,
  you can now calculate things

12
Planning

• Which to use (Min, Avg, Max)?
• Is it possible?
• Technically?
• In your area?
• In the time you have?
• With the money you have?
• Will it pay for itself
• Consider productivity,
• customer satisfaction,
• cost of maintenance
• Uptime payback

13
Uptime

• Calculated by measuring how much you are down
• Sometimes you get this as a guarantee
• Take with a grain of salt. All it means is
  theyll refund your money, little comfort if you
  lose your biggest customer
• 4 9s
• Means 99.9999 uptime
• This means you have 8.6 seconds of down time per
  day, 4.5 minutes per month.
• Some industries are regulated
• Nuclear power plants often have 9 or 10 9s
  requirements.
• Mean time between failure

14
Baseline study

• Monitor what you are doing now
• This will help you know how things will look in
  the future
• Also lets you see non obvious trends

15
SNMP

• Allows you to monitor a network
• You can see utilization of resources
• Network traffic
• CPU usage
• Memory usage
• Disk usage
• You can monitor services
• When sendmail fails it can trap
• Traps can be caught and generate
• Emails, pages, phone calls etc.