PostgreSQL: AWS RDS vs. Aurora vs. Google Cloud SQL
Basic Features
|
AWS RDS |
AWS Aurora |
Google Cloud SQL |
| Latest Version |
10.1
9.6 installed in benchmark
|
9.6 |
9.6 |
| Secure Connection (SSL,encrypted) |
YES |
YES |
YES |
| Scalable (scale storage and instance types) |
YES |
YES |
YES |
| High Available and Durable |
YES |
YES |
YES |
| Managed |
YES |
YES |
YES |
| Easy to Administer |
YES |
YES |
YES |
Pricing
Note: the following table lists Single zone pricing. The pricing of instance and storage will simple
double when using RDS with multi-AZ, using Aurora with a replica, and using Google SQL with high
availability.
|
AWS RDS |
AWS Aurora |
Google Cloud SQL |
| 2vcpu+15.25G instance |
$0.25 /hour |
$0.29 /hour |
$0.189 /hour
This price is calculated of 2vcpus+15.25G for comparison purpose. But in reality the
maximum ram is 13G for 2vcpus.
|
| Storage |
$0.125 /GB |
$0.10 /GB |
$0.17 /GB |
| Disk I/O |
Free 3 IOPS /GB storage |
$0.20 /1 million requests |
Free 30 IOPS /GB storage |
Benchmarks
Conditions
|
AWS RDS |
AWS Aurora |
Google Cloud SQL |
| Instance type |
r4.large(2 vCPUs + 15.25G) |
r4.large(2 vCPUs + 15.25G) |
High memory(2 vCPUs + 13G) |
| Region |
us-west-2a |
us-west-2a |
us-west1 |
| Client Side (running pgbench) |
EC2 instance in us-west-2a |
EC2 instance in us-west-2a |
Google Compute in us-west1 |
| Installed PG version |
9.6 |
9.6 |
9.6 |
| Storage Encryption |
enabled |
enabled |
enabled |
| Multi-AZ/ Replication/ High-availability |
Disabled |
Disabled |
Disabled |
Benchmark details
Import data:
pgbench -i -s 10 test
AWS RDS:
[email protected]:~$ pgbench -c 10 -j 10 -t 500 test -h testrds.*.us-west-2.rds.amazonaws.com -U test
Password:
starting vacuum...end.
transaction type:
scaling factor: 10
query mode: simple
number of clients: 10
number of threads: 10
number of transactions per client: 500
number of transactions actually processed: 5000/5000
latency average = 5.745 ms
tps = 1740.77029 (including connections establishing)
tps = 1774.006224 (excluding connections establishing)
AWS Aurora:
[email protected]:~$ pgbench -c 10 -j 10 -t 500 test -h test.*.us-west-2.rds.amazonaws.com -U test
Password:
starting vacuum...end.
transaction type:
scaling factor: 10
query mode: simple
number of clients: 10
number of threads: 10
number of transactions per client: 500
number of transactions actually processed: 5000/5000
latency average = 13.045 ms
tps = 766.564305 (including connections establishing)
tps = 774.188096 (excluding connections establishing)
Google Cloud SQL:
[email protected]:~$ pgbench -c 10 -j 10 -t 500 test -h 35.203.*.* -U postgres
Password:
starting vacuum...end.
transaction type:
scaling factor: 10
query mode: simple
number of clients: 10
number of threads: 10
number of transactions per client: 500
number of transactions actually processed: 5000/5000
latency average = 7.084 ms
tps = 1411.595534 (including connections establishing)
tps = 1422.435397 (excluding connections establishing)