[talwgx]

How far does a ClickHouse Cloud service's compute fall when the rows written into its logs table fall, above the smallest size, and how long does the autoscaler take? The setup was one Scale-tier service in AWS us-east-1 running ClickHouse 26.6 with 2 replicas, vertical autoscaling from 8 to 32 GiB per replica, and the ClickStack otel_logs schema under seven queries a minute. Two arms differ only in the rows written:

• Arm A writes every record of the 215 MB OTel replay at 70,000 rows/s.
• Arm B writes the 65.3% the Receiver (log10x/edge-10x:1.1.153) kept, at 45,716 rows/s.

The run lasted from 2026-11-02 19:48 to 2026-11-05 03:47 UTC. Every figure is in clickhouse-cloud-elasticity/RESULT.md with the file and field it was read from.

Result

The same cut gave a step down or no change, depending on the size the service started from.

Pair	Arm A's size	Units, A to B	Ratio	computeCHC per 15 min, A to B	Cut to step down done
1	the autoscaler's	4.0 to 2.997	0.749	0.298 to 0.224	4.33 h
2	the autoscaler's, where pair 1 left it	3.0 to 3.0	1.0	0.224 to 0.224	no step in 10.25 h
3	16 GiB by config, released at the cut	4.0 to 3.0	0.750	0.298 to 0.224	4.21 h
4	16 GiB by config, released at the cut	4.0 to 3.0	0.750	0.298 to 0.224	2.91 h

• Path dependence. At 4 units, arm B's busiest replica reads 0.337 to 0.351 of its CPU, below the autoscaler's 37.5% low watermark. At 3 units both arms sit inside the 37.5% to 75% band, where the autoscaler takes no action.
• Dollars. usageCost computeCHC reads $1.196/h against $0.896/h (ratio 0.749). Unit-hours × $0.2985, the published Scale compute rate per unit-hour in AWS us-east-1 (clickhouse.com/pricing, read 2026-11-05, results/raw/pricing-quotes.md), reads $1.194/h against $0.8955/h (ratio 0.750). The postings match the rate: a quarter-hour is 0.2985 at 4 units and 0.2239 at 3.
• Elasticity. The fall in compute over the fall in rows is 0.721 to 0.723 in pairs 1, 3 and 4.
• Scale-down lag. From the cut to the step down being done took 4.33, 4.21 and 2.91 h. The new replicas booted 4.20, 4.07 and 2.78 h after the cut.
• Write share W. (insert + merge) / (insert + merge + select) cores reads 0.65 to 0.67 as the mean over replicas. On the busiest replica it reads 0.47 to 0.51 in arm A and 0.54 to 0.55 in arm B.
• Memory. About 2 GiB used per replica, of 12 to 16 GiB.
• Floor. 2 units, $0.60/h.

Not shown

• A production query mix.
• The collector path live: the feed was pre-split.
• More than one region, schema or service shape.

Not measured, each reported as read in RESULT.md:

• the wire parse of real inserts
• the probe's steady-state select share
• which replica the recommender reads
• why pair 4 stepped 1.3 h sooner

Credit and teardown

• Spent. $42.61 of the $300 credit, against the run's $200 cap (creditBalances amountSpent, unchanged from 11-05 04:25Z to 05:35Z).
• chce-elasticity's own usageCost. totalCHC 41.714, of which computeCHC 41.680.
• Deleted. chce-elasticity was stopped at 03:44:15Z and deleted at 03:46:57Z. A GET then returned 404 (results/raw/teardown.jsonl).
• Kept. "My first service" was not deleted.

Committed files carry the runner's address only as <runner-ip> and no key or password.

Not merged; for Tal.

🤖 Generated with Claude Code