SigNoz Resources Planning

This guide provides a comprehensive overview of the core components, sizing guidelines, resource specifications, and architectural considerations crucial for scaling and capacity planning when deploying SigNoz. Use this information to plan the right deployment size and ensure your SigNoz installation is optimized for both present and future growth.

Components Overview

A SigNoz deployment consists of several key components, each with specific resource requirements. The following sections detail the configuration requirements for each component.

The following table outlines the resource requirements for all SigNoz components:

Notes:

• Collectors: Scale with processing needs. More processors or pipelines will require more CPU. In very high-scale environments or if ClickHouse is throttling, allocate more memory to Collectors.
• ClickHouse: Ensure storage capacity is sized according to your retention policy and expected data volume. Plan for growth over time.
• SigNoz Core: Resource requirements vary based on component functionality and usage.
• PostgreSQL: Size according to metadata volume and query patterns. Managed database services (AWS RDS, GCP Cloud SQL, Azure Database for PostgreSQL) are recommended
• ZooKeeper: Required by ClickHouse for cluster operation; should have stable resource allocation.

Instance Types

These are the node types we suggest from various cloud providers. Please see the relevant specifications in the provider's documentation.

A family fixes the chip and the CPU-to-memory ratio; capacity comes from the size you pick inside it. Choose a size that meets or exceeds the per-component CPU and memory in the table above. The smallest size in each of these families falls short of what its workload needs: t3.nano gives you 2 vCPUs and 0.5 GiB, against the 4 vCPUs and 16 GiB each Collector needs.

For AWS:

• General Purpose workloads (Collectors, SigNoz Core, PostgreSQL, ZooKeeper): Any General Purpose machine available in your region that belongs to T3 instance family and above for Intel chips and T4g instance family and above for ARM chips.
• Compute-optimized workloads (ClickHouse): Any Compute Optimized machine available in your region that belongs to C5 instance family and above for Intel chips and C6g or C7g instance family and above for ARM chips.
• Smallest sizes that meet the table: t3.xlarge or t4g.xlarge (4 vCPUs, 16 GiB) for General Purpose workloads, and c5.4xlarge, c6g.4xlarge, or c7g.4xlarge (16 vCPUs, 32 GiB) for ClickHouse.

For GCP:

• General Purpose workloads (Collectors, SigNoz Core, PostgreSQL, ZooKeeper): Any General Purpose machine available in your region that belongs to E2 instance family and above.
• Compute-optimized workloads (ClickHouse): Any machine available in your region that belongs to C3 or C3D machine series and above. GCP documents C3 and C3D under its general purpose machine family, so look for them there rather than under compute-optimized, which covers C2, C2D, H3, and H4D.
• Smallest sizes that meet the table: e2-standard-4 (4 vCPUs, 16 GiB) for General Purpose workloads, and c3-highcpu-22 (22 vCPUs, 44 GiB) for ClickHouse. C3 has no 16-vCPU shape, jumping from 8 to 22 vCPUs, and c3-standard-8 carries only 8 vCPUs. GCP labels these same figures GB in its machine type tables, defining 1 GB as 1,024 MB, so the numbers match the GiB values used here.

For Azure:

• General Purpose workloads (Collectors, SigNoz Core, PostgreSQL, ZooKeeper): Any General Purpose machine available in your region that belongs to the Dsv5 size series and above for Intel chips, Dasv5 and above for AMD chips, or Dpsv5 and above for Arm chips.
• Compute-optimized workloads (ClickHouse): Any Compute Optimized machine available in your region that belongs to the Fsv2 size series and above for Intel chips, or Fasv6 and above for AMD chips. Azure's compute-optimized F family has no Arm-based series, so use a general purpose Dpsv5 or later size if you want ClickHouse on Arm.
• Smallest sizes that meet the table: Standard_D4s_v5, Standard_D4as_v5, or Standard_D4ps_v5 (4 vCPUs, 16 GiB) for General Purpose workloads, and Standard_F16s_v2 (16 vCPUs, 32 GiB) or Standard_F16as_v6 (16 vCPUs, 64 GiB) for ClickHouse.