Hologres Serverless Computing lets you run large SQL workloads using on-demand, serverless compute resources. You no longer need to pre-allocate fixed compute capacity. Each task gets its own isolated serverless environment, ensuring no interference between jobs and consistent performance.
Serverless Computing is available for commercial use starting July 1, 2024, at 00:00 (UTC+8). For pricing details, see Billing overview.
Why use Serverless Computing
Traditionally, you must reserve sufficient compute power to handle peak workloads, even if those large jobs run only occasionally. This leads to two major issues:
Wasted resources: Reserved capacity remains idle during off-peak hours.
Performance bottlenecks: During execution, multiple jobs compete for limited resources, leading to instability and failures.
Serverless Computing solves both problems. It runs big SQL jobs in a serverless resource pool, completely separate from your instance's exclusive resources. This prevents resource contention and underutilization, delivering greater stability, reliability, and cost efficiency.
How it works
Serverless Computing automatically provisions isolated compute resources for each eligible SQL job. These resources are:
Ephemeral: Allocated only when a job runs.
Scalable: Can scale up to three times the capacity of your instance's dedicated resources.
Isolated: Not shared with other tasks, ensuring predictable performance.
Once the job completes, the resources are released—so you only pay for what you use.
When to use Serverless Computing
Running large SQL jobs reliably:
The problem: Your instance has limited dedicated resources. Importing or querying large datasets often results in Out-of-Memory (OOM) errors.
Benefits of Serverless Computing:
Leverage up to 3x the compute power of your instance.
Run resource-intensive jobs in isolation.
Improve job success rates and protect instance stability.
Reducing idle resources and saving money:
The problem: You've reserved high computing capacity for occasional large jobs, but they run less than 16 hours per day—leaving resources idle for most of the time.
Benefits of Serverless Computing:
Downsize your reserved dedicated resources.
Use serverless compute only when needed.
Pay only for actual usage—no more paying for idle capacity.
Handling peak workloads flexibly:
Benefits of Serverless Computing:
Route specific users or query queues to use serverless resources.
Reduce load on your exclusive instance during high-traffic periods.
Serverless Computing is not a direct replacement for your instance's exclusive resources. They serve different purposes:
Item | Serverless Computing | Exclusive resources |
Resource sharing |
|
|
Real-time workloads | Not supported. Includes real-time data ingestion (via Flink, DataWorks, Holo Client) and immediate point queries using Fixed Plan. | Required for real-time operations. |