Dmitry Trofimenko
09/11/2026, 9:55 AMRocky
09/11/2026, 9:56 AMm5, c5, r5, c6i).
* Frontend (FE): Responsible for query planning and metadata.
* Spec: 8 vCPU, 16–32 GB RAM.
* AWS Instance: m5.2xlarge or c5.2xlarge.
* Backend (BE): Responsible for data storage and query execution.
* Spec: 16 vCPU, 64 GB RAM.
* AWS Instance: m5.4xlarge or r5.4xlarge.
* Storage: SSD is strongly recommended (EBS gp3).
* FE: ~100-200 GB for metadata.
* BE: Depends on your data volume. (Formula: Raw data size × 3 replicas ÷ 3:1 compression ratio).
* Network: 10 Gbps Ethernet is recommended for node-to-node communication.
Minimal Requirements (Developer/POC)
For testing, you can run a single-node deployment (1 FE and 1 BE on the same machine).
* Spec: Minimum 8 vCPU and 32 GB RAM.
* AWS Instance: m5.2xlarge.
──────────
2. Estimated Monthly Price (Self-Managed on EC2)
Prices estimated based on us-east-1 (N. Virginia) On-Demand rates.
Component | Quantity | Instance Type | Monthly Cost (Per Unit) | Total Monthly Cost
FE Nodes | 3 | m5.2xlarge | ~$280 | ~$840
BE Nodes | 3 | m5.4xlarge | ~$560 | ~$1,680
EBS Storage | 2 TB | gp3 | $0.08/GB | ~$160
Total (HA) | | | | ~$2,680 / month
How to reduce costs:
* Reserved Instances/Savings Plans: Can reduce the compute cost by 30–60%, bringing the total closer to $1,100–$1,600/month.
* Spot Instances: BE nodes can run on Spot instances for up to 90% savings, though this is risky for production unless you use a "Shared-Data" architecture (where data is persisted in S3).
* Shared-Data Architecture: By using S3 for primary storage and CN (Compute Nodes) instead of BE, you can scale compute down when not in use and save significantly on EBS costs.
3. StarRocks Managed Service (CelerData Cloud)
If you prefer not to manage the infrastructure:
* Free Developer Tier: Available for small-scale testing.
* BYOC (Bring Your Own Cloud): You pay for the AWS resources in your account plus a management fee (contact sales for pricing; enterprise support typically starts around $1,200/month).
For specific sizing advice tailored to your data volume and query concurrency, you can check the official StarRocks cluster planning guide.
References
• deployment/deployment_prerequisites.md
• deployment/plan_cluster.md
• faq/Others.md
• faq/Deploy_faq.md
• deployment/environment_configurations.mdDmitry Trofimenko
09/11/2026, 9:58 AMRocky
09/11/2026, 9:58 AMm5.2xlarge or c5.2xlarge.
* RAM: 16 GB (Absolute bare minimum).
_ _Note:* While the system can start with less (e.g., 8GB in Docker), you will quickly hit "Memory Limit Exceeded" errors during query execution or metadata operations. 32 GB is recommended for a smoother dev experience.
* Storage: 20 GB+ of SSD space.
* OS: Linux (Ubuntu 18.04+, CentOS 7+, or Debian 10+).
2. Mandatory Configuration (Single Node)
By default, StarRocks expects 3 replicas for high availability. If you are using only one BE node, you must change the default replication setting, or your tables will remain in an "unhealthy" state and queries will fail.
In your fe.conf file, add:
properties
default_replication_num = 1
3. Docker-Specific Requirements
If you are using Docker for your dev setup:
* Docker Desktop Memory: Ensure you have allocated at least 8GB - 12GB of RAM to the Docker engine itself in settings.
* Disk Space: At least 20GB free for the StarRocks images and basic data.
4. Port Requirements
Ensure these common ports are open on your dev machine:
* 9030: MySQL client connection (FE).
* 8030: FE HTTP interface (Web UI).
* 8040: BE HTTP interface.
Summary Table for Dev/POC
Resource | Minimum | Recommended
vCPU | 8 (with AVX2) | 16
RAM | 16 GB | 32 GB
Disk | 20 GB SSD | 100 GB SSD
Deployment | 1 FE + 1 BE | 1 FE + 1 BE
Important: If you use an AWS instance that does not support AVX2 (like older m4 or t2 instances), the Backend (BE) will fail to start or will crash immediately. Stick to m5, c5, r5 or newer generations.
References
• faq/Others.md
• deployment/plan_cluster.md
• deployment/deploy_manually.md
• developers/development-environment/dev-env-setup.md
References
• faq/Others.md
• deployment/plan_cluster.md
• faq/Deploy_faq.md
• deployment/deploy_manually.md
• administration/management/BE_configuration.md
• quick_start/shared-data.md