DSQL Experiment Notes
Back to the experiment list
E010CompletedP0

Cost of the same workload and selection criteria

What costs and constraints come with DSQL's performance and convenience advantages?

Issue #10 Measurement plan Measured 2026-09-29

Results summary

This experiment calculated how the cost of DSQL and existing services changes with the shape of the load, for the same workload and the same latency target. As a minimum scope (MVP), we did not launch any new AWS resources; instead, we multiplied values measured in E002 (throughput, DPU, ACU) and E009 (idle behavior) by Seoul Region On-Demand prices. The workload is the E002 order mix (40% product lookup, 30% order history, 20% order creation, 10% cancellation).

DSQL is billed in DPUs for the amount of work it processes, and this workload used 0.029–0.032 DPU per request. At the Seoul price ($10 per million DPUs), that is about $0.31 per million requests, or about $1.12 to sustain 1,000 TPS for one hour. By contrast, RDS Multi-AZ (db.r6g.xlarge, gp3 400 GiB) costs about $1.22 per hour regardless of load and handled up to 5,839 TPS within the SLO. Therefore, DSQL is cheaper when average load is below about 1,100 TPS, and a fixed instance is cheaper when load above that level continues. Aurora Provisioned (writer+reader, about $1.25 per hour + I/O) shows a similar break-even point.

The difference grows depending on the shape of the load. For a constant 1,000 TPS load over 24 hours, the daily cost was similar: DSQL about $27 and RDS Multi-AZ about $29. For a service that works at 1,000 TPS for only 8 hours and is idle for 16, DSQL cost about $9 and RDS about $29, so DSQL was about one third. Aurora Serverless v2 (A2) with a read replica scales the writer and reader together, costing about $3.4 per hour at 1,000 TPS, and because of its minimum capacity it is billed during idle time too, for about $53 per day. Using 0 ACU auto-pause reduces this to about $27, but in E009 the first connection failed after exceeding 15 seconds.

Production readiness

On cost alone, DSQL favors “services with few or irregular requests” and is at a disadvantage for “services with high load all day.” For this workload the break-even point was an average of about 1,100 TPS. Below that, DSQL is cheaper because the cost of keeping a fixed instance running disappears and high availability is included by default. Conversely, if thousands of TPS are processed continuously, the DPU charges that scale with request count exceed the cost of a fixed instance. Large aggregations are also billed by the amount processed (E012), so services with heavy analytical workloads need a separate cost calculation. This calculation is an estimate based on one workload mix, On-Demand prices, and measurements from a single run.

Question

What costs and constraints come with DSQL’s performance and convenience advantages?

Test conditions

  • Additional AWS runs: None. Calculated from the measurements and prices below.
  • Sources of measurements:
    • DSQL DPU per request: in the 2,400–26,285 TPS cells of the second E002 measurement, the per-minute sum of CloudWatch TotalDPU ÷ number of attempted requests = 0.029–0.032 (0.031 used in calculations).
    • A2 ACU: average writer and reader ACU in the cells of the first E002 measurement. 5 each at 100–400 TPS, 12 each at 1,600 TPS, 28 each at 5,657 TPS, and 32 each (the maximum) at 11,559 TPS and above.
    • Highest throughput passing the SLO for each service: E002 (R1 5,839, A1 3,375, A2 11,559, DSQL 24,541 TPS or more).
    • Idle behavior: E009 (A2’s first connection after auto-pause failed after exceeding 15 seconds; DSQL responded immediately).
  • Prices (Seoul, On-Demand, USD): DSQL $10 per million DPUs; RDS PostgreSQL db.r6g.xlarge Multi-AZ $1.079/h + two copies of gp3 400 GiB about $0.144/h; Aurora db.r6g.xlarge Standard $0.627/h (2 instances, writer and reader); Aurora Serverless v2 Standard $0.20/ACU-h; Aurora I/O $0.24 per million. Checked on 2026-09-27 (RDS, Aurora) and from the 2026-09-11 published price list (DSQL).
  • Not included: storage (Seoul DSQL $0.40/GB-month, Aurora $0.12/GB-month, RDS gp3 $0.131/GB-month. With this run’s 5 GiB of data, DSQL storage is small at about $0.07 per day, but when data grows to several TB, the fact that DSQL’s storage price is about 3.3 times Aurora’s becomes significant. DSQL price source: AWS Price List API, 2026-09-11 publication), Aurora I/O (not measured separately per load level), backups, networking, load generators, and commitment discounts and credits.
  • Deviations from the plan: Instead of actually running 24-hour, 8+16-hour, and spike patterns, we converted them using measured unit costs. Instead of cost per million successful operations, we calculated hourly costs matched to load durations.

Performance results

Hourly cost by load

Average load DSQL RDS Multi-AZ (R1) Aurora Provisioned (A1) Aurora Serverless v2 (A2)
100 TPS $0.11 $1.22 $1.25 + I/O $2.0 (10 ACU)
400 TPS $0.45 $1.22 $1.25 + I/O $2.0 (10 ACU)
1,000 TPS $1.12 $1.22 $1.25 + I/O about $3.4 (about 17 ACU, interpolated)
1,600 TPS $1.79 $1.22 $1.25 + I/O $4.8 (24 ACU)
5,657 TPS $6.3 $1.22 Exceeds SLO $11.2 (56 ACU)
11,559 TPS $12.9 Exceeds SLO Exceeds SLO $12.8 (64 ACU, maximum)
24,541 TPS $27.4 Exceeds SLO Exceeds SLO Exceeds SLO
  • DSQL cost = load (TPS) × 3,600 × 0.031 DPU × $10 / 1 million. DPU per request drops slightly as load increases, so this slightly overestimates cost at high load.
  • “Exceeds SLO” marks configurations that could not handle that load within the latency target in E002. A larger instance could handle it, but we did not measure that cost.
  • A2 is configured with the reader following the writer at the same capacity for HA. Without the reader, A2’s cost is about half, but there is no standby replica for failover. For DSQL and R1, multi-AZ high availability is included in the price.

Daily cost scenarios

Load shape DSQL R1 A1 A2 Notes
1,000 TPS × 24 hours about $27 about $29 about $30 + I/O about $82 Constant load
1,000 TPS × 8 hours + 16 hours idle about $9 about $29 about $30 + I/O about $53 A2 at minimum 4 ACU × 2 when idle
Same as above, A2 auto-pause about $9 - - about $27 A2 first connection exceeds 15 seconds (E009)
1,000 TPS × 23 hours + 10,000 TPS × 1 hour about $37 Cannot handle Cannot handle about $91 1-hour spike
  • Break-even: the average load at which DSQL costs the same as R1 is $1.22 ÷ ($0.31 × 3,600 / 1 million) = about 1,100 TPS.

Development and operations

  • DSQL requires no choosing of capacity or setting of minimum and maximum capacity, so cost forecasting is as simple as “number of requests × DPU per request.” However, DPU per request depends on the workload’s queries and must be measured with the actual workload.
  • DSQL did not require growing instances as load increased. In E002 the fixed instances could not handle the 256-connection load within the SLO, so a larger instance would have to be chosen to handle that load.
  • Human working time is covered separately in E011 and is not added to this table as a monetary amount.

Cost

This report itself did not create any AWS resources. For reference, the experiment costs for this entire study are as follows.

Run Scope Cost Basis
E001 SQL compatibility about $0.08 Cost Explorer billed amount
E004 Contention and consistency about $2.07 Cost Explorer billed amount
E002 first measurement Pilot and exploration of four configurations $48.57 Cost Explorer billed amount
E002 second measurement + E003, E008, E012 DSQL only $18.54 Cost Explorer charge (estimate about $18.6)
E005, E006, E007, E009 Small-scale D1 and A2 $5.07 Cost Explorer charge (estimate about $5.1)
Shared by both 2026-09-29 runs Cross-AZ transfer, EBS, public IPv4 $0.65 Cost Explorer charge

Conclusions and limitations

  • DSQL’s cost is proportional to the number of requests, while a fixed instance’s cost is proportional to time. For this workload, the two lines cross at an average of about 1,100 TPS.
  • DSQL has the advantage for services with long idle periods, services with irregular load, and small services that need high availability. For services with sustained high load, fixed instances are cheaper.
  • DSQL storage costs $0.40 per GB-month, about 3.3 times Aurora’s ($0.12), so services with large data must compare storage costs separately.
  • Limitations: this is an estimate that linearly scales the unit cost of one workload mix. Aurora I/O, storage, backups, networking, and commitment discounts were not included, and the load patterns were not actually run for 24 hours. The A2 value at 1,000 TPS is interpolated between measurement points.

Cleanup record

  • This report did not create new AWS resources. For the three cited measurement runs (e002-20260928t033404z-e1b1, e002-20260929t081139z-4dce, e002-20260929t082035z-afbe), each report confirmed 0 remaining resources (2026-09-28 12:07:44, 2026-09-29 11:10:30, and 10:20:27 UTC).