Use Azure Data Factory to Bring SAP Data to Cloud (Round 2)
/Since our last blog posted 2 years ago, we received multiple requests to perform a comparison with Azure Data Factory’s “SAP Table” connector, in the context of bringing SAP data to cloud. We did some tests and would like to share our finding.
One-Line Summary
AecorSoft Table connector through ADF SSIS IR is multiple times faster and significantly more scalable than ADF SAP Table connector.
Test Setup
SAP Source: IDES EhP7 with SQL Server 2012R2
OS: Windows Server 2012R2 Hyper-V VM
Hardware: 16GB RAM in VM, 16 v-core of Xeon E5-2670 v1 2.6GHz
ADF SSIS IR: Standard_D2_v3 2 Core(s), 8192MB, 1 node
Destination: Azure Blob
Test Case 1 - BKPF
112 columns, 584,448 rows
AecorSoft and ADF SSIS IR: 1m 34s
ADF SAP Table via Selfhosted IR: 4m 44s
Quite note: the network to Azure is a bottleneck in our test setup. On this old workstation, if the destination is local CSV file, AecorSoft would only require around 23 seconds to complete. Therefore, about 1 minute is spent on the data transfer to Azure.
As the network limitation applies to both scenarios, we would live with this result for now. If such bottleneck is eliminated, the performance difference would be much bigger.
Test Case 2 - BSEG
354 columns, 1,678,379 rows
AecorSoft and ADF SSIS IR: 17m 7s
ADF SAP Table via Selfhosted IR: FAILED. It caused page allocation error short dump on SAP!
Looking at the statistics, the network bandwidth issue is still very concerning. For those customers who have Express Route, the bottleneck should be much mitigated.