You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a question regarding the system resource utilization in case when a single node CouchDB is heavy loaded.
I tried to find out what is the maximum CouchDB performance (especially max number of docs written per second) with a specific hardware/software configuration.
I ran bulk document creation tests using JMeter and Tsung loading tools with the following config (compacation was disabled during the tests).
Each JMeter node was producing specific load profile:
100 docs bulk write with doc size 10KBytes;
50 docs bulk write with doc size 21KBytes;
3 docs bulk write with doc size 1.2MBytes.
I was surprised with the results. It looks like CouchDB can't utilize all available resources. CPU core utilization ranged from 50% to 70%, RAM utilization - 16%, disk write bandwidth utilization - 10%, disk write IOPs utilization - 4%, network bandwidth utilization - 32%.
JMeter node CPUs were utilized around 30%
Could you please clarify, are there some architectural decisions in CouchDB which intentionally prevent full resource utilization?
Maybe some part of the resources is reserved for the CouchDB cluster mode data interchange?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hello friends!
I have a question regarding the system resource utilization in case when a single node CouchDB is heavy loaded.
I tried to find out what is the maximum CouchDB performance (especially max number of docs written per second) with a specific hardware/software configuration.
I ran bulk document creation tests using JMeter and Tsung loading tools with the following config (compacation was disabled during the tests).
CouchDB Server config
CPU: Intel Ice Lake 32 cores
RAM: 32 GB
Disk: 465 GB SSD (max bandwidth (read / write) - 550 MB/s / 410 MB/s)
Network: 10Gbit Ethernet
OS: Debian 11
CouchDB 3.2.2
JMeter / Tsung nodes config
CPU: Intel Ice Lake 4 cores
RAM: 16 GB
Disk: 64 GB SSD
Network: 10Gbit Ethernet
OS: Debian 11
Each JMeter node was producing specific load profile:
I was surprised with the results. It looks like CouchDB can't utilize all available resources. CPU core utilization ranged from 50% to 70%, RAM utilization - 16%, disk write bandwidth utilization - 10%, disk write IOPs utilization - 4%, network bandwidth utilization - 32%.
JMeter node CPUs were utilized around 30%
Could you please clarify, are there some architectural decisions in CouchDB which intentionally prevent full resource utilization?
Maybe some part of the resources is reserved for the CouchDB cluster mode data interchange?
Thanks in advance!
Beta Was this translation helpful? Give feedback.
All reactions