Performance metric for CQL queries

Post Reply
Posts: 3
Joined: Sat Aug 02, 2014 6:14 am

Performance metric for CQL queries

Post by Alireza » Sat Aug 02, 2014 6:28 am

Does Odysseus has any metric reporting feature for CPU and memory consumption rate along with incoming event rate per each query?
For example, it could report CPU usage every second respecting to incoming events for Query_X. For example I want to determine Query_X in time t1 consume 100 messages and it uses 2000ms CPU. Again in t2, 120 msg, and 2400ms cpu, and so on.


Site Admin
Posts: 5
Joined: Mon Jul 28, 2014 4:58 pm

Re: Performance metric for CQL queries

Post by admin » Wed Aug 06, 2014 9:37 am

Currently, the only way to get these informations is to add further metadata.

Code: Select all

#METADATA TimeInterval
#METADATA SystemLoad
#METADATA Datarate
Remark: If you use #METADATA you always have to add #METADATA TimeInterval, so here only SystemLoad and Datarate are "new".

To activate, install feature:
- "Systemload Metadata Feature" to measure cpu and memory load
- "Datarate Systemload Metadata Feature" to measure datarate

We will provide a new operator to transfer the metadata to the payload to operators like aggregate can be used on the metadata values.

Remark: Every measurement slows down the system ;-)

User avatar
Marco Grawunder
Posts: 272
Joined: Tue Jul 29, 2014 10:29 am
Location: Oldenburg, Germany

Re: Performance metric for CQL queries

Post by Marco Grawunder » Thu Aug 07, 2014 2:14 pm

Now it possible to measure the values and export them to the output schema.

Code: Select all

#METADATA TimeInterval
#METADATA SystemLoad

#ODYSSEUS_PARAM systemLoadInterval 1000

#INPUT ${WORKSPACE}/Nexmark/StreamSourcesSimple.qry

out = SYSTEMLOADTOPAYLOAD({LOADNAME = 'local'}, person)
Remark: systemLoadInterval is the time in ms for the update rate of the measurement, default is 3000. Should not be set to values lower 500.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest