Atlassian uses cookies to improve your browsing experience, perform analytics and research, and conduct advertising. Accept all cookies to indicate that you agree to our use of cookies on your device.
Atlassian uses cookies to improve your browsing experience, perform analytics and research, and conduct advertising. Accept all cookies to indicate that you agree to our use of cookies on your device. Atlassian cookies and tracking notice, (opens new window)
/
PTF - Performance Testing of Exporting Circulation Logs
Circulation Log Export scales linearly up to 50K, indicating the inflection point begins around 100K records on both Central and Member tenants.
Export duration on Central tenant is ~10–15% faster than Member, but comparison is influenced by different dataset sizes (1M vs 7M logs).
1K–50K exports complete within ~4 sec–2.5 min and scale near-linearly.
Exports above 100K become heavy — 100K ~4-6 min, 200K takes ~14–18 min, and 300K ~27–31 min.
CICO is impacted by concurrent exports, but the impact of Circulation Log exports on CICO does not increase proportionally with export size.
Check-In/Check-Out response times increase (mostly under 20% on average), with the most notable spikes at the 95th percentile during 30K (37%) and 200K+ (27% - 64%) exports.
Test Runs
Test #
Test Conditions
Status
1
Central tenant – Export: 1K, 10K, 30K, 50K, 100K, 200K, 300K records
Completed
2
Member tenant – Export: 1K, 10K, 30K, 50K, 100K, 200K, 300K records
This table contains Circulation Log Export durations across multiple export sizes on the Central tenant. Export scales linearly up to 50K but degrades at 100K. 100K marks the inflection point where per-record cost begins accelerating non-linearly.
Export size
Central tenant
Duration
sec/1K
1K
3.9 sec
3.9 sec
10K
14.1 sec
1.4 sec
30K
44.4 sec
1.4 sec
50K
1 min 25 sec
1.7 sec
100K
4 min 33 sec
2.7 sec
200K
14 min 20 sec
4.3 sec
300K
27 min 16 sec
5.4 sec
Test №2
This table contains Circulation Log Export durations across multiple export sizes on the Member tenant. On member tenant follows the same pattern, the inflection point remains at 100K.
Export size
Member tenant
Duration
sec/1K
1K
18.8 sec
18.7 sec
10K
28.5 sec
2.8 sec
30K
1 min 17 sec
2.5 sec
50K
2 min 24 sec
2.8 sec
100K
5 min 51 sec
3.5 sec
200K
17 min 55 sec
5.3 sec
300K
30 min 52 sec
6.1 sec
Test №3
This table contains Circulation Log Export durations across multiple export sizes on the Member tenant with CICO background load (8 users). The average impact on CICO ~20%. However, pct95 reveals significant spikes: Check-Out Controller pct95 reaches 1874 ms during 30K export (+37%) and Check-In Controller pct95 hits 984 ms during 200K export (+64%). The impact of Circulation Log exports on CICO does not increase proportionally with export size.
The main CPU consumer is mod-data-export-worker. CPU utilization scales with export size, with sustained high usage observed during 200K–300K exports. Utilization was 31% for 1K exports, peaking at ~150% during the 300K export.
Service Memory Utilization
Memory utilization increases with higher export load and drops after each peak. No signs of memory accumulation or memory leak were observed. Overall memory behavior remains stable across both tests.
On the central tenant, higher peaks are observed (~75–80%) during large exports (200K–300K), for the member tenant remains (~55–60%).
DB CPU Utilization
CPU utilization shows clear scaling with export size in both tenants. At low load (1K–10K), CPU remains minimal (~9% on Central, ~21% on Member), while at high load (300K) it increases to ~18% on Central and up to ~40–48% on Member on spikes. Member tenant consistently shows higher % and more pronounced spikes. However, this comparison should be interpreted in context of different dataset sizes (Central ~1M logs vs Member ~7M logs).
DB Connections
Database connection counts remain stable - on average 560.
DB load
DB load during test 1 and test 2 sliced by SQL:
Resources Utilization for Test 3
Service CPU Utilization
With background CICO load, mod-data-export-worker CPU utilization patterns remain consistent. CICO-related modules do not show additional CPU consumption during exports.
Service Memory Utilization
Memory utilization under combined export + CICO load remained stable and comparable to the export-only.
DB CPU Utilization
Database metrics (CPU, connections, load) under concurrent export + CICO workload show slight increases compared to the export-only tests. At low load (1K–10K), CPU is ~35%, while at 300K it reaches ~29–52% with intermittent peaks.
Tenant Circulation Log Records Count cs00000int 1093310 cs00000int_0001 7148498
Methodology/Approach
Data preparation step:
Data was generated using the following query:
Prepared datasets of 1K, 10K, 30K, 50K, 100K, 200K, and 300K records for both Central and Member tenants.
Each dataset was tagged with a descriptive label indicating its size for easy filtering for export testing.
WITH bc AS (
SELECT
jsonb->>'barcode' AS barcode,
row_number() OVER () AS rn
FROM cs00000int_0001_mod_inventory_storage.item
WHERE NULLIF(TRIM(jsonb->>'barcode'), '') IS NOT NULL
LIMIT 500000
),
bc_cnt AS (
SELECT count(*) AS cnt FROM bc
),
gen AS (
SELECT
gen_random_uuid() AS new_id,
gs,
(gs % (SELECT cnt FROM bc_cnt)) + 1 AS idx
FROM generate_series(1, 300000) gs
)
INSERT INTO cs00000int_0001_mod_audit.circulation_logs (id, jsonb)
SELECT
g.new_id,
jsonb_set(
jsonb_set(
jsonb_set(
'{
"id": "",
"date": "",
"items": [
{
"itemId": "27efe92f-1b5c-5b32-b6c7-48e71539b52e",
"loanId": "0af86ff6-bc0f-4a62-a5ea-42976f1dc91f",
"holdingId": "e8cd4664-1b53-5cc2-a139-ee2bc8229eaa",
"instanceId": "243ad997-f24d-5f46-89a9-ff8b8b59c390",
"itemBarcode": ""
}
],
"action": "Checked out",
"object": "Loan",
"source": "EBSCOAdmin, EBSCOAdmin",
"linkToIds": {
"userId": "0001ef6e-216a-4436-8ee7-42121519afe6"
},
"description": "Type: Page. 300K",
"userBarcode": "813487757",
"servicePointId": "9d1b77e8-f02e-4b7f-b296-3f2042ddac54"
}'::jsonb,
'{id}',
to_jsonb(g.new_id::text)
),
'{date}',
to_jsonb(
to_char(
now() - (random() * interval '365 days'),
'YYYY-MM-DD"T"HH24:MI:SS.MS"Z"'
)
)
),
'{items,0,itemBarcode}',
to_jsonb(b.barcode)
)
FROM gen g
JOIN bc b
ON b.rn = g.idx;
Test Configuration:
Tested Circulation Log export on Central and Member tenants.
Executed exports of 1K, 10K, 30K, 50K, 100K, 200K, and 300K records were performed manually, using filtering by description to select the target dataset for export.
Ran background CICO workload (8 users) on the Member tenant during exports.