From 10 Parallel request (of 10 sequential request) the client can't always connect and we see time out error (succes ratio <100%)
Sequential request are fired faster than actual responses so from DB perspective they are almost parallel request as well
Database probably already become bottleneck with 2 threads, effectively firening a total of 20 call very quickly. Its know that the DB connection pool/internal will slow down from 12 or more 'parallel' request
Get 1000 nodes in Parallel with varying thread count
In this test, 1000 requests are sent using curl, but with varying thread count (using --parallel-max option).
which would expand into a series of 16 requests to:
http://example.com/archive1996/vol1.html
http://example.com/archive1996/vol2.html
...
http://example.com/archive1999/vol4.html
Results
Threads
Time (s)
Speedup
Comments
1
140.4
1.0
2
71.6
2.0
2 threads is 2x faster than 1 thread
3
48.5
2.9
4
37.2
3.8
5
31.0
4.5
6
26.6
5.3
7
23.8
5.9
8
21.6
6.5
9
20.0
7.0
10
18.7
7.5
10 threads is 7.5x faster than 1 thread
11
17.7
7.9
12
16.8
8.4
There are exactly 12 CPU cores (logical) on test machine
13
16.7
8.4
14
16.7
8.4
15
16.8
8.4
20
16.8
8.4
30
16.7
8.4
40
16.8
8.4
50
16.7
8.4
Graphs
Image Added
Observations
There were no failures during the tests (e.g. timeouts or refused connections).
Performance increases nearly linearly with increasing thread count, up to the number of CPU cores.
Performance stops increasing when the number of threads equals the number of CPU cores (expected).
Verbose statistics show that each individual request takes around 0.14 seconds, regardless of thread count (but with multiple CPU cores, requests are really done in parallel).
Data sheets
...
View file
name
CpsPerformance.xlsx
height
250
View file
name
performanceTest.zip
height
250
View file
name
Performance test.postman_collection.json
height
250
Test scripts overview
- performanceTest.sh Get 1000 times single large object from thousands of devices (1000, 2000, ..., 10000) and create metric after each run - performanceRootTest.sh Get 10 times the whole data tree as one object from thousands of devices (1000, 2000, ..., 10000) and create metric after each run - parallelGetRequestTest.sh Get one devices parallel from a database with 10000 devices, executed 10 times sequential
- buildup.sh Create the dataspace, create the schemaset, create the anchor and create the root node - owb-msa221.zip The schemaset for the tests - outNode.json The input for the root node creation - createThousandNode.sh Helper script for the database creation - innerNode.json The input for the sub node creation - createMetric.sh Helper script for metric creation