minimum latency calculations could be scewed by failures

[davidkel]

An example of test output where init fails all of it's tries

+--------+------+------+-----------------+-----------------+------------------+-----------------+------------------+
| Name   | Succ | Fail | Send Rate (TPS) | Max Latency (s) | Min Latency (s)  | Avg Latency (s) | Throughput (TPS) |
|--------|------|------|-----------------|-----------------|------------------|-----------------|------------------|
| init1  | 0    | 24   | 5.4             | 0.00            | 9007199254740.99 | -               | 5.4              |
|--------|------|------|-----------------|-----------------|------------------|-----------------|------------------|
| query1 | 24   | 0    | 7.2             | 0.82            | 0.15             | 0.60            | 6.8              |
|--------|------|------|-----------------|-----------------|------------------|-----------------|------------------|
| init2  | 0    | 24   | 5.4             | 0.00            | 9007199254740.99 | -               | 5.4              |
|--------|------|------|-----------------|-----------------|------------------|-----------------|------------------|
| query2 | 24   | 0    | 7.2             | 0.68            | 0.17             | 0.56            | 6.9              |
+--------+------+------+-----------------+-----------------+------------------+-----------------+------------------+

[aklenik]

I'm labeling this a bug. If we cannot calculate a metric, then it should be made explicit (both in code and in the report). The TransactionStatistics class current uses the Number.MAX_SAFE_INTEGER initial value for the min* attributes (for simpler implementation). Either that needs to be changed, or the reporting side has to check for that value (or if the corresponding counters are 0).