Skip to content

Latest commit

 

History

History
83 lines (61 loc) · 4.08 KB

2023-08 CrateDB 5.4.md

File metadata and controls

83 lines (61 loc) · 4.08 KB

Ingestion speed results CrateDB 5.4

This page documents ingest benchmark results from our one million rows/s blog post. While the above-linked article regularly gets updated to reflect most recent version of CrateDB, hardware architecture, etc., historic benchmark results are documented here.

Setup

  • CrateDB version: 5.4.0

  • EC2 instance type: m6in.4xlarge

  • Operating system: Amazon Linux 2023

  • Terraform config:

     module "cratedb-cluster" {
         source = "[email protected]:crate/crate-terraform.git//aws"
    
         region = "eu-west-1"
    
         vpc_id             = "vpc-..."
         subnet_ids         = ["subnet-..."]
         availability_zones = ["eu-west-1b"]
    
         ssh_keypair = "cratedb_terraform"
         ssh_access  = true
    
         instance_type         = "m6in.4xlarge"
         instance_architecture = "x86_64"
    
         # The size of the disk storing CrateDB's data directory
         disk_size_gb = 100
    
         disk_iops       = 5000
         disk_throughput = 400 # MiB/s
    
         # CrateDB-specific configuration
         crate = {
             # Java Heap size in GB available to CrateDB
             heap_size_gb = 40
    
             cluster_name = "cratedb-cluster"
    
             # The number of nodes the cluster will consist of
             cluster_size = 1 # increase to scale the cluster
    
             ssl_enable = true
         }
    
         enable_utility_vm      = true
         load_balancer_internal = true
    
         utility_vm = {
             instance_type         = "m6in.4xlarge"
             instance_architecture = "x86_64"
             disk_size_gb          = 50
         }
     }
    
     output "cratedb" {
         value     = module.cratedb-cluster
         sensitive = true
     }

Results

Without replication

Nodes Cluster size increase Throughput (rows/s) Throughput increase absolute (rows/s) Throughput increase relative Primary shards Processes Concurrent requests
1 293,147 16 1 16
2 +100% 493,723 +200,575 +68% 24 2 12
3 +50% 676,781 +183,058 +37% 32 3 11
4 +33% 856,480 +179,680 +27% 40 4 10
5 +25% 1,043,659 +187,198 +22% 48 5 9

With 1 replica

Nodes Cluster size increase Throughput (rows/s) Throughput increase absolute (rows/s) Throughput increase relative Primary shards Processes Concurrent requests
2 286,341 12 2 12
3 +50% 403,630 +117,288 +41% 16 3 11
4 +33% 510,086 +106,456 +26% 20 4 10
5 +25% 629,568 +119,482 +23% 24 5 9