Comparisons
Comparisons betweenDatabend Cloud and Snowflake
We are dedicated to building the world's most powerful data cloud, determination to provide exceptional service
TPC-H Benchmark
The TPC-H benchmark is a standard for evaluating decision support systems, focusing on complex queries and data maintenance. In this analysis, we compare Databend Cloud with Snowflake using the TPC-H SF100(SF1 = 6 Million Rows) dataset, encompassing 100GB of data and approximately 600 million rows across 22 queries.
Benchmark Environment
Edition
Standard
Standard
Warehouse Size
Small
Small
vCPU
16
16
Price
$2/hour
$4/hour
AWS Region
us-east-2
us-east-2
Storage
AWS S3
AWS S3
Data Loading Benchmark
The total elapsed time is 446 seconds, costing $0.25.About 36% faster
0100s200s300s400s500s
Query Benchmark: Cold Run
The data warehouse was suspended and resumed before executing the queries.
The total elapsed time is 166 seconds, costing $0.09.About 20% faster
11.703 s
8.036 s
4.524 s
3.786 s
8.908 s
6.04 s
8.108 s
4.462 s
9.202 s
7.014 s
1.237 s
3.234 s
9.082 s
7.345 s
10.886 s
8.976 s
18.152 s
13.34 s
13.525 s
12.891 s
2.582 s
2.183 s
10.099 s
8.839 s
13.458 s
7.206 s
8.001 s
4.612 s
8.737 s
4.621 s
4.864 s
1.645 s
5.363 s
14.315 s
19.971 s
12.058 s
9.893 s
12.579 s
8.538 s
8.836 s
16.439 s
12.27 s
3.744 s
1.926 s
Query Benchmark: Hot Run
The data warehouse is not suspended, local disk cache is used.
The total elapsed time is 124 seconds, costing $0.07.About 10% faster
8.934 s
7.568 s
3.018 s
3.125 s
6.089 s
5.234 s
4.914 s
3.392 s
5.8 s
4.857 s
0.891 s
2.142 s
5.381 s
4.389 s
5.724 s
5.887 s
10.283 s
9.621 s
10.368 s
8.524 s
1.165 s
1.364 s
7.052 s
5.352 s
12.829 s
6.18 s
3.288 s
2.725 s
3.475 s
2.748 s
4.094 s
1.124 s
4.203 s
13.757 s
18.583 s
11.63 s
3.888 s
7.881 s
6.379 s
5.797 s
10.287 s
9.806 s
1.573 s
1.122 s
Architectures
We are on a quest to build a better
Open Source
Open source (Core components)
Proprietary
Language
Rust (Fast & memory safe)
C++
SQL Support
Separation of Computing and Storage
Task
In Seconds
In Minutes
Warehouse Auto-suspending
In seconds
In Minutes
Storage Format
Parquet - open data format
Accessible to downstream applications
Legacy micro-partition
NOT accessible to downstream applications
Semi-Structured Data
Query Result Caching
Enterprise Grade Encryption
Scalability
Unlimited and so on
Scheduling
Kubernetes pod-based
Virtual machines-based
Elasticity
Compute node unit
Cluster unit; No choice of node size
Costs
More technologies, less costs.
Transparent Pricing
Cloud Service isn't transparent pricing.
Pricing
50% of Snowflake's Pricing
100%
Costs
Economic
Per-second billing
Provisioning compute resources in seconds
Expensive
Per-minute billing
Provisioning compute resources in minutes