Skip to content

Records service#

The Records object holds time-series data such as message flow in a table as shown below.

Start timestamp ... End timestamp
0.0 ... 0.1
1.0 ... 1.1
2.0 ... NaN
3.0 ... 2.1
... ... ...

The start timestamp column contains the system time at the starting point. The end timestamp column contains the system time at the end point. Both columns indicate the system time, and if there is no corresponding value for start, the value is NaN.

For the intermediate columns, the time at the intermediate point between starting point and end one is represented.

This table representation can be used for various measurement targets such as callbacks, nodes, and paths.

The rows of the table are visualized as lines in a message flow diagram.

The above table is stored by Records object.

The followings are processing using Records object.

See also

See also

Period#

Period is metrics defined as the elapsed time between two occurrence of cyclic events. Difference between two neighboring timestamp on the same column is defined as period.

period_n = t_{n} -t_{n-1}

Example#

Input

Start timestamp
0.0
1.0
2.0
3.0
...

Output

Timestamp Period
0.0 1.0
1.0 1.0
2.0 1.0
... ...

See also

Frequency#

Frequency is defined as the number of events that occur in one second.

Example#

Input

Start timestamp
0.0
0.1
0.5
1.2
1.3
2.3
...

Output

Timestamp Period
0.0 3.0
1.0 2.0
... ...

Latency#

Latency is length of time from preceding event to afterward one. It is defined as difference from preceding timestamp to afterward on the same row.

latency_n = t^{end}_{n} - t^{start}_{n}

Example#

Input

Start timestamp End timestamp
0.0 0.1
1.0 1.1
2.0 NaN
3.0 3.1
... ...

Output

Start timestamp Latency
0.0 0.1
1.0 0.1
3.0 0.1
... ...

Response Time#

Response time is amount of time it takes for a system to respond to an input.

As shown above, latency can be calculated if a table is constructed. Latency is defined as delay from input to any output. This calculation is very simple, but it is not used for evaluating response. If multiple output depends on a single input, latency is not regarded as response time. Moreover, for example, if a sensor is driven at 10 Hz, a latency of up to 100 ms should be considered.

CARET serves best-case response time and worst-case response time. The former is roughly equal to latency from an input to a corresponding initial output. CARET provides the latter because it also takes into account the latency until times wake up. For example, this latency is equivalent to a sensor operating at 10 Hz taking a 100 ms delay at maximum.

Example#

Input

Start timestamp End timestamp
0.0 0.1
1.0 1.1
2.0 NaN
3.0 3.2
4.0 4.3
... ...

Intermediate
Create intermediate data for when the interval in Start timestamp [0.0, 4.0] maps to End timestamp.

Start timestamp End timestamp
[0.0, 1.0) 1.1
[1.0, 3.0) 3.2
[3.0, 4.0) 4.3

Output

Start timestamp Best-case response time Worst-case response time
1.0 0.1 (1.1 - 1.0) 1.1 (1.1 - 0.0)
3.0 0.2 (3.2 - 3.0) 2.3 (3.2 - 1.0)
... ... ...

Note that best case response time is equal to Latency, except for the cumbersome cases listed later.

Worst-case response time also counts as response time in the case of a drop.

Visualize response time#

Latency vs Response time

Of the pseudo two message flow diagram as shown above, the upper figure shows latency and the lower explains response time. The upper figure includes cumbersome cases as follows;

  • Cases where multiple latencies are defined for a single output from multiple inputs
  • Message dropping
  • Crossing
  • Branching

Some of these can be pessimistically large when it comes to latency.

The diagram above describes the table as follows

Start timestamp End timestamp
0.0 2.0
0.5 2.5
2.0 3.5
3.0 3.5
4.0 5.0
4.5 7.0
5.5 6.0
5.5 6.5

After extracting only the best case flows as shown in the lower figure, the best case is calculated as the latency of the flow and the worst-case as the latency is regarded as delay from the previous input to the output.

According to those definitions, the response time is calculated as below.

Start timestamp Min response time Max response time
0.0 2.0 2.5
0.5 0.5 3.0
3.0 1.0 2.0
4.0 0.5 2.0

See also