AWS-to-Azure Replication

Benchmark statistics are based on product replication engine version 6.0 test results.
Product versions 9.0 and above deliver substantially better performance.

AWS-to-Azure Replication Scenario


Geo-replication scenario:

Moderate ~100,000 inserts per minute into 4 tables, each containing 500+ million record, into a database (size ~200GB) hosted on an AWS Virginia based RDS R3.2Xlarge instance, geo-replicated asynchronously to Azure SQL type P1. Data in transit is encrypted.

Database size: ~200GB

Database size increment: ~50MB/min (~70GB/day)

In oder to conduct the benchmark in an environment closer to a real production configuration, an additional load of ~250,000/min inserts into another database (size 750GB) on same R3.2Xlarge RDS is executed. The 750GB RDS database is geo-replicated to another AWS region. This replication as well other 2 smaller databases replications are handled on the same M4.XLarge CLOUDBASIC RDS AlwaysOn Server instance.

[AWS-to-Azure Geo-replication lag: < 1min]

Virginia RDS R3.2XLarge (Primary; ~100,000 inserts per minute into 4 tables)
Resource Utilization Statistics

Benchmark RDS R3.2XLarge

Azure P1(Read-Replica; ~100,000 inserts per minute geo-mirrored with with <1min lag)
Resource Utilization Statistics

AWS-to-Azure Benchmark: Azure SQL P1 CPU

Database Size: ~200GB

AWS-to-Azure Replication Benchmark DB-Size

RDS AlwaysOn/Geo-Replicate Dashboard

AWS-to-Azure Replication Benchmark Dashboard

RDS AlwaysOn/Geo-Replicate Continuous Synchronization Logs, divided into ~1 min time slots:

Number of changes per time slot: ~100,000
Geo-Replication lag: <1min

RDS AlwaysOn/Geo-Replicate Seeding (initial replication) Statistics

Note: For AWS to Azure replications, the seeding phase takes substantial amount of time. Once it transitions to continuous replication it would need to catch up with the pending changes which had occurred during the seeding phase. Then it will transition to continuous replication with a low lag (lag may vary based on database structure, parallel replication execution on same RDS AlwaysOn server, connectivity between AWS and Azure etc.)

<< Back to Benchmarks