For banks and insurers, the ability to leverage the cloud, process payments at high volumes in near real time, and scale quickly is essential. However, for many of these financial services institutions, the systems that support core processes remain largely dominated by monolithic architectures, which is no longer a sustainable option. Large, single-node IT architectures cannot keep up with the increasing amount of data and ever-changing business demands. Many organizations are, therefore, making the transition to the cloud to take advantage of more agile operations, accelerated innovation, and lower costs. However, this raises the question of whether or not distributed configurations in the cloud can cope with those heavy workloads while maintaining the robustness and elasticity required. And this is exactly what SAP and Google Cloud set out to determine.
SAP and Google Cloud recently teamed up to create a proof of concept at the request of their joint customer, PayPal, who has adopted Google Cloud as their core IT partner. The goal of the project was to obtain performance data to demonstrate the scalability of
SAP S/4HANA for financial products subledger, running SAP HANA on Google Cloud with a scale-out approach.
SAP S/4HANA for financial projects subledger is a cloud-optimized, comprehensive subledger for companies within the financial services industries, such as banks, insurers, and re-insurers. Organizations traditionally run this solution on a single-node architecture that only supports scale-up scenarios. However, scaling up in high-demand environments, can lead to oversized monolithic systems, with increased risk due to a single point of failure.
On the other hand, scaling out means adding new instances to the system that run in parallel. This option enables organizations to take advantage of multiple distributed processing units in cloud environments to more flexibly scale their systems on demand.
At the core of SAP S/4HANA for financial projects subledger is SAP HANA, SAP’s in-memory database. Designed to improve the handling of mixed workloads and for massive parallel processing of large data sets, a prerequisite for this test configuration, SAP HANA supports transactions (OLTP) and analytics (OLAP) on a single platform. This means that the testing could be done in a simplified configuration and the test team could run reports on live transactional data in real time.
The financial products subledger capability is designed with a new, scalable data model that enables two-level range-range partitioning. This model supports advanced data tiering technologies and made the use of SAP HANA native storage extension possible. This smart data tiering capability has been added for scale-out deployments of SAP HANA starting with
SAP HANA 2.0 SPS 05.
It distinguishes between “hot” data maintained in memory for immediate access and “warm” data placed in storage for lower-priority access. This results in a faster solution with a lower data footprint that can also scale as demand grows. For example, during the test run, the relevant database tables showed a 78% reduction in memory footprint (from 18 TB to just 4 TB) while only adding 2 TB of data volume to the disk when the subledger posting documents for the fiscal year 2020 were removed from memory.
Over the course of six months, using more than 20 TB of real financial data, the teams from SAP and Google Cloud pushed the architecture to the extremes, scaling the system and infrastructure out as workloads increased.
The results of this testing were very promising and demonstrated the feasibility of this approach. For OLTP processing, the engineers achieved end-to-end runtimes of 44 minutes for a full business day, which is a reduction of 63% compared with the performance levels for a scale-up system similar to what PayPal is currently using.
To determine whether the system could also scale for peaks in data volume that typically occur during events such as Black Friday and Cyber Monday, engineers progressively increased loads from 40 million transactions to 160 million, quadrupling the number of line items per posting date from 500m to 2bn. Under these conditions, the system responded, as expected, with slightly more than linear performance, while CPU utilization rates were well below capacity.
By adjusting parallelization parameters, engineers reduced the runtime by a factor of 40, from 20 minutes down to 30 seconds compared with the performance levels for a reference scale-up system of similar size.
Throughout the entire test, the system demonstrated rock-solid robustness. The configuration ran actively for 24 weeks, processing nonstop for multiple business days per calendar day without hardware or operating system outages. The team was able to maximize system uptime and minimize maintenance downtime. For example, a unique combination of high-memory VMs and features, such as live migration on Google Cloud, allowed highly critical SAP HANA instances to stay up and running during scheduled host maintenance or while applying security patches.
To test the impact on maintenance procedures, engineers optimized the system for redistributing data in parallel, increasing the SAP HANA cluster from 8 to 10 nodes. The result was a four-hour maintenance window, a reduction of 50% when compared with the sequentially managed process that involves redistributing table partitions and balancing memory across all nodes.
While this proof of concept focuses on a high-volume scenario for payments on financial instruments, similar benefits can be expected elsewhere, for example, when processing loans, securities, or insurance contracts in financial products subledger. Running SAP HANA in the cloud with a scale-out architecture that minimizes costs and increases flexibility helps organizations quickly respond to changing market conditions.
Over the past ten years, we have seen incredible innovation with SAP HANA. This joint proof of concept for PayPal using Google Cloud for SAP S/4HANA for financial products subledger is just the latest example of how SAP HANA enables our customers to derive business value from their data.
Please also make sure to check out the blog
Google Cloud and SAP demonstrate massive scalability for financial services customers by June Yang, VP and General Manager, Compute, Google Cloud, or the read the paper
Increasing Scalability and Performance for Even the Heaviest Financial Workloads SAP S/4HANA for Fin... for more information!
