0

I'm trying to use OpenDJ's searchrate command to measure LDAP's performance and response time to queries.

The output looks as follows:

    Throughput                            Response Time                       
   (ops/second)                           (milliseconds)                       
recent  average  recent  average  99.9%  99.99%  99.999%  err/sec  Entries/Srch
-------------------------------------------------------------------------------
2260.7   2262.8   6.219    6.219  413.842  476.425  476.817      0.0           1.5
3857.5   3178.3   4.149    4.777  352.825  476.404  500.017      0.0           0.8
5078.2   3753.2   2.825    3.978  360.940  460.154  500.017      0.0           1.0
4557.7   3934.5   3.411    3.830  352.480  455.638  500.017      0.0           1.0

It's unclear to me what each column represents. Please help by providing a clarification for each column meaning.

Ludovic Poitou
  • 4,788
  • 2
  • 21
  • 30
Doron Gold
  • 3,664
  • 4
  • 32
  • 44

1 Answers1

3

Looks like to me the header is pretty clear.

The first 2 columns represent throughput, in operations per second, one for the last period, and an average since the start.

The next 5 columns represent response times in milliseconds. Computed on last period, average since start, and then there are 3 percentiles that allows to understand the outliers. The value 352.480 means that 99.9% of requests average in 352ms response time, whereas 99.999% of the request average in 500.017ms. Percentiles are important to understand the distribution of response times.

Finally, the last 2 columns represent the number of errors per sec (should be 0) and the number of entries returned for each search (should be 1 or vary depending on the search and the data).

Ludovic Poitou
  • 4,788
  • 2
  • 21
  • 30
  • Thanks. Your explanation does make things clearer. However, one thing is still unclear to me: how can Entries/Srch return a non integer value? As I understand, it represents the average of the returned entries amount from the **same** query run multiple times. so how can the **same** query return different amount of entries? Also, can you please clarify, how many times the query is run per each output line (period)? – Doron Gold Jan 19 '15 at 09:41
  • 2
    The Entries/Search is computed counting the number of returned entries, and dividing by the number of Search Results. This gives the average number of entries returned per requests. But in the first couple of iterations, as the server and the client are warming up, the ratio between search results and search entries is unbalanced. There could also be searches that return more than 1 entries, thus making the average not an exact integer. This depends on the search query and parameters. Usually, you use searchrate with ranges or random filters, so the query is not always the same. – Ludovic Poitou Jan 19 '15 at 10:08