Dedicated SQL for high-performance log analysis at scale - Simple Log Service

Why use Dedicated SQL

Limitations of standard queries

Standard queries have the following limitations for large-scale data:

Inaccurate results: Resource limits (time slices, I/O, and data volume) may prevent full data loading, affecting statistical accuracy.
Performance bottlenecks: A single shard supports only 400 MB of data, limiting TB-level log analysis and high-concurrency scenarios.
Resource competition: Multiple tenants share resources, causing contention.

Core values of Dedicated SQL

Enhanced mode: High performance and high concurrency

Enhanced mode targets real-time performance and high concurrency:

Performance: A single node processes up to 2 GB of data with up to 100 concurrent queries.
Elastic scaling: Storage and compute resources scale dynamically as needed.
Typical scenarios: API success rate alerts, real-time monitoring, and high-concurrency point queries.

Full precision mode

Full precision mode targets scenarios that require exact results:

Zero-error guarantee: Ensures that data is fully loaded by trading time for resources.
Dedicated resources: Ensures stable operations until the task is complete or a timeout occurs.
Typical scenarios: Financial reconciliation, security audits, ultra-long period analysis, and large-scale trend analysis.

Important

The maximum running time for an SQL query is 55 seconds, and the maximum concurrency is 5.

Aspect	Enhanced mode	Full precision mode
Core objective	Performance acceleration	Result accuracy
Resource policy	Shared resource pool and elastic scaling	Dedicated resource pool and time-for-accuracy tradeoff
Typical scenarios	Real-time monitoring and high-concurrency analysis	Rigorous analysis scenarios, such as financial reconciliation, security audits, ultra-long period analysis, and large-scale trend analysis.
Accuracy toleration	Limited errors are allowed	Zero errors are required

How Dedicated SQL works

SQL enhancement

Data in Simple Log Service is stored in shards. Each shard has limited processing capacity — if data volume is too large, queries may be truncated. Adding shards improves throughput but only for new data, and increases real-time consumption clients. SQL enhancement dynamically scales compute resources to improve analysis. Typical scenarios:

Real-time data analytics requiring high-performance analysis.
Long-period analysis, such as monthly data analytics.
Large-scale analysis of hundreds of billions of rows.
High-concurrency analysis (more than 15 concurrent SQL queries), such as multi-dimension report and point queries.

SQL complete accuracy

During large-scale analytics, data loading may be interrupted by:

Time slice exhaustion: Allocated time resources run out.
Data volume threshold: Total loaded data exceeds the limit.
Row count threshold: Loaded rows exceed the limit.
I/O threshold: Disk reads exceed the limit.

These interruptions can cause incomplete data loading, affecting result accuracy. SQL complete accuracy resolves these issues. Typical scenarios:

Business monitoring and alerting: Critical monitoring requires accurate analysis results.
Business operations analysis: Key metrics analysis involving revenue, finance, retention, and conversion.
Online data services: Serving external users with SQL results that must be fully accurate.

Billing information

Billed based on the CPU time used for SQL analysis, measured in core-hours. One core-hour represents the exclusive use of one CPU core for one hour. For more information, see Dedicated SQL billing examples.

Pay-as-you-go: Dedicated SQL fee = CPU time (hours) × Unit price per hour
Resource plan: Prepaid Plan, where usage is converted to and deducted in Cost Units (CUs).

Limits

Limit	General-purpose instance	Dedicated SQL
Limit	General-purpose instance	SQL enhancement	Full Precision
Concurrency	Up to 15 concurrent queries per project.	Up to 100 concurrent queries per project.	Up to 5 concurrent queries per project.
Data volume	Up to 400 MB per query (excluding cached data). Excess data is truncated with an incomplete query result marker.	Up to 2 GB per query (excluding cached data). Excess data is truncated with an incomplete query result marker.	Unlimited.
Enabling the mode	Enabled by default.	Enable via toggle. SQL enhancement.	Enable via toggle. SQL complete accuracy.
Fee	Free of charge.	Charged based on the actual CPU time used.	Charged based on the actual CPU time used.
Data effectiveness	Applies only to data written after the feature is enabled. To analyze historical data, you must reindex the data.	Applies only to data written after the feature is enabled. To analyze historical data, you must reindex the data.	Applies only to data written after the feature is enabled. To analyze historical data, you must reindex the data.
Return results	By default, a query returns up to 100 rows and 100 MB. Queries exceeding 100 MB return an error. To return more data, use the LIMIT clause.	By default, a query returns up to 100 rows and 100 MB. Queries exceeding 100 MB return an error. To return more data, use the LIMIT clause.	By default, a query returns up to 100 rows and 100 MB. Queries exceeding 100 MB return an error. To return more data, use the LIMIT clause.
Field value size	Default maximum field value length is 2 KB (2,048 bytes), configurable up to 16 KB (16,384 bytes). Content exceeding the limit is excluded from analysis and retrieval. Note To change the limit, set Maximum Length of Text Field. The updated setting applies only to incremental data. Create an index.	Default maximum field value length is 2 KB (2,048 bytes), configurable up to 16 KB (16,384 bytes). Content exceeding the limit is excluded from analysis and retrieval. Note To change the limit, set Maximum Length of Text Field. The updated setting applies only to incremental data. Create an index.	Default maximum field value length is 2 KB (2,048 bytes), configurable up to 16 KB (16,384 bytes). Content exceeding the limit is excluded from analysis and retrieval. Note To change the limit, set Maximum Length of Text Field. The updated setting applies only to incremental data. Create an index.
Timeout period	Maximum timeout: 55 seconds.	Maximum timeout: 55 seconds.	Maximum timeout: 55 seconds.
Number of bits for double-typed field values	Maximum 52 bits for double-typed field values. Floating-point numbers encoded with more than 52 bits lose precision.	Maximum 52 bits for double-typed field values. Floating-point numbers encoded with more than 52 bits lose precision.	Maximum 52 bits for double-typed field values. Floating-point numbers encoded with more than 52 bits lose precision.