examples of system require high performance:
911 services.
health care services. ie need to claim the health insurance for surgery etc
We need to have qos in
performance
availablity
security
maintainability
response time=time taken by system + network + e commerce server
congestion time = time spend by the above three time on disk memory i/o
Throughput time
No requests processed per second
For a OLTp->> No of Transactions per second
Server--> sessions per second
Router->>KB per second/message per second
disk-->i/o requests processed per second
Cpu-Million instructions per second
FLOPS - Floating point instruction per second
throughput= min[offered work load/ server capacity];
through put increases with load and after the max load it starts to decrease again its called as thrashing
As load goes on too many processes compete for the resource like main memory.
There will be less space available on main memory as too many process compete for the resources.
So there will be overload such as time spent on cpu i/o cycles such as page fault.
Finally, the peformance suffers
availablity:
The system should be available and running.
The second with 99.99% may not be tolerable in applications such as 911 services and defense system
The reason for non available are failures and overloads
Limiting the no of users allowed to access the system is called admission control. If a particular
system exceeds the allowed no of request we can reject the new incoming request. The rejected request may be
served
scalablity:
The performance does not degrade with no of users or load
extensibility:
How sytem can cope uo with new functional requirement and performance requirements
system life cycle:
functional requirements
include also non functional requirement such as performance required for a book search system
design --> choose correct algorithm
development--> properly use iocalls
testing--> simulate virtual user and...