Commit 248524db authored by Thorsten Simons's avatar Thorsten Simons

1.3.1 - updated queries and documentation on tp()

parent d8342b76
......@@ -8,7 +8,7 @@ Built-in queries
$ hcprequestanalytics showqueries
available queries:
500_highest_throughput The 500 records with the highest throughput (KB/sec) for objects >= 1 Byte
500_highest_throughput The 500 records with the highest throughput (Bytes/sec) for objects >= 1 Byte
500_largest The records with the 500 largest requests
500_worst_latency The records with the 500 worst latencies
clientip No. of records per client IP address
......@@ -24,7 +24,7 @@ Built-in queries
node_req No. of records per request per node
node_req_httpcode No. of records per http code per request per node
percentile_req No. of records per request analysis, including percentiles for size and latency
percentile_throughput_kb No. of records per request, with percentiles on throughput (KB/sec) for objects >= 10MB
percentile_throughput_b No. of records per request, with percentiles on throughput (Bytes/sec) for objects >= 10MB
req No. of records per request
req_httpcode No. of records per http code per request
req_httpcode_node No. of records per node per http code per request
......@@ -75,12 +75,9 @@ You can check the available queries, including the additional ones::
$ hcprequestanalytics -d dbfile.db -a addqueries showqueries
available queries:
500_highest_throughput The 500 records with the highest throughput (KB/sec) for objects >= 1 Byte
500_highest_throughput The 500 records with the highest throughput (Bytes/sec) for objects >= 1 Byte
500_largest The records with the 500 largest requests
500_worst_latency The records with the 500 worst latencies
add_count count all records
add_node_req_http node-per-request-per-httpcode analysis
add_req_count count records per request
clientip No. of records per client IP address
clientip_httpcode No. of records per http code per client IP address
clientip_request_httpcode No. of records per http code per request per client IP address
......@@ -94,7 +91,7 @@ You can check the available queries, including the additional ones::
node_req No. of records per request per node
node_req_httpcode No. of records per http code per request per node
percentile_req No. of records per request analysis, including percentiles for size and latency
percentile_throughput_kb No. of records per request, with percentiles on throughput (KB/sec) for objects >= 10MB
percentile_throughput_b No. of records per request, with percentiles on throughput (Bytes/sec) for objects >= 10MB
req No. of records per request
req_httpcode No. of records per http code per request
req_httpcode_node No. of records per node per http code per request
......@@ -144,8 +141,16 @@ latency INT the internal latency needed to fullfil the request
=============== ======= ========================================================
Non-standard aggregate functions
--------------------------------
Non-standard SQL functions
--------------------------
* ``tp(size, latency)``
Calculates the throughput (in bytes/second) from an objects size and the
internal latency.
Non-standard SQL aggregate functions
------------------------------------
* ``percentile(column, float)``
......
......@@ -223,21 +223,21 @@ query : SELECT * from
order by 'Bytes/sec' desc limit 500;
freeze pane : E5
[percentile_throughput_kb]
comment : No. of records per request, with percentiles on throughput (KB/sec) for objects >= 10MB
[percentile_throughput_b]
comment : No. of records per request, with percentiles on throughput (Bytes/sec) for objects >= 10MB
query : SELECT request,
count(*), min(size), avg(size), max(size),
percentile(tp(size, latency)), 10) as 'pctl-10 (KB/sec)',
percentile(tp(size, latency))/1024, 20) as 'pctl-20 (B/sec)',
percentile(tp(size, latency))/1024, 30) as 'pctl-30 (B/sec)',
percentile(tp(size, latency))/1024, 40) as 'pctl-40 (B/sec)',
percentile(tp(size, latency))/1024, 50) as 'pctl-50 (B/sec)',
percentile(tp(size, latency))/1024, 60) as 'pctl-60 (B/sec)',
percentile(tp(size, latency))/1024, 70) as 'pctl-70 (B/sec)',
percentile(tp(size, latency))/1024, 80) as 'pctl-80 (B/sec)',
percentile(tp(size, latency))/1024, 90) as 'pctl-90 (B/sec)',
percentile(tp(size, latency))/1024, 95) as 'pctl-95 (B/sec)',
percentile(tp(size, latency))/1024, 99) as 'pctl-99 (B/sec)',
percentile(tp(size, latency))/1024, 99.9) as 'pctl-99.9 (B/sec)'
percentile(tp(size, latency), 10) as 'pctl-10 (B/sec)',
percentile(tp(size, latency), 20) as 'pctl-20 (B/sec)',
percentile(tp(size, latency), 30) as 'pctl-30 (B/sec)',
percentile(tp(size, latency), 40) as 'pctl-40 (B/sec)',
percentile(tp(size, latency), 50) as 'pctl-50 (B/sec)',
percentile(tp(size, latency), 60) as 'pctl-60 (B/sec)',
percentile(tp(size, latency), 70) as 'pctl-70 (B/sec)',
percentile(tp(size, latency), 80) as 'pctl-80 (B/sec)',
percentile(tp(size, latency), 90) as 'pctl-90 (B/sec)',
percentile(tp(size, latency), 95) as 'pctl-95 (B/sec)',
percentile(tp(size, latency), 99) as 'pctl-99 (B/sec)',
percentile(tp(size, latency), 99.9) as 'pctl-99.9 (B/sec)'
FROM logrecs where size >= 10048576 and latency > 500 GROUP BY request
freeze pane : C5
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment