Performance Measures
In this section, we will look at three performance measures which are commonly used as a criteria in System design.
- IOPS
- TPS
- Availability (in nines)
IOPS:
IOPS defined as 'Input Output operations per second' is a performance metric for measuring the rate of data transfer. This is generally used as a criterion to measure data transfer in storage devices and network devices. In any system, the overall performance is related to or affected by, the rate of data transfer in low level physical devices. Generally IOPS is measured in MB/sec or GB/sec.
When we are selecting storage device for a system, one of the criteria could be selecting a device which offers higher IOPS.
TPS:
TPS defined as 'Transactions per second' is a performance metric for measuring the amount of atomic transactions which the system can handle. It provides a measure of the traffic that the system can handle. It could be measured as an average or peak traffic value.
To calculate TPS, we would need to take into account the capacity of the application server, the database and any other system involved in an end to end transaction.
Availability (in nines):
Availability of a system refers to a state of the system wherein it successfully responds to requests. It represents a SLA (Service Level Agreement), which is guaranteed by the system.
Availability of a system could be represented as uptime or as downtime.
Uptime is represented as a percentage of the time duration in a year for which the system is Up or available.
Downtime is represented as a percentage of the time duration in a year for which the system is down or unavailable.
Thus, if a system is up for 90% in a year, then the downtime period for the system in one year is 10% or 36.5 days. For a month, the downtime is 72 hours. 90% uptime is also known as "One Nine".
On similar terms, 99% uptime is known as "Two nines", 99.9% is known as "Three nines" and so on.