After 6 months of diligent testing at the SAP Co-innovation Lab Silicon Valley (COILsv), a team of engineers from SAP and Pure Storage validated a set of best practices of using Pure Storage FlashStack for SAP HANA dynamic tiering and finally released their findings in a white paper.
Why It Matters
HANA is an in-memory data platform. The value of in-memory computing is proven with many live customers. Knowing that RAM is a relatively expensive resource, HANA provides tools to help our customers lower cost by moving data to storages based on data temperature. No question that hot data shall reside in main memory; when done appropriately, warm and cold data may be moved to various storage devices without comprising performance.
Using extended storage host to offload warm data from HANA, Dynamic tiering allows data-aging rules to be defined by administrators, and to be executed automatically at scheduled intervals.
The Tested System Landscape
Quite some storage vendors are certified for HANA. Each storage vendor may have a set of best practices for enabling HANA dynamic tiering. To show the benefits of using Pure Storage, the team built up a complete HANA system at COILsv consisting of a combined stack of hardware (storage, network, and compute) and software (Cisco UCS Manager, SAP HANA SPS11, and SAP HANA dynamic tiering; SAP BusinessObjects Design Studio; SAP HANA Cloud Platform, smart data streaming; SAP BusinessObjects Lumira, server version for teams; and Pure Storage GUI). As shown in the diagram below, the main components of hardware used include:
- Network: Cisco Unified Computing System (UCS) fabric interconnect 6120XP for external and internal connectivity of IP and fiber channel (FC) network.
- Storage: Pure Storage FlashArray//m20 with fiber channel connectivity
- Compute: Cisco UCS B250 M2 Server
The Tested Use Case
IoT solutions often are paired with large amount of data to be collected and analyzed. The team implemented a simple IoT application that demonstrates a coal mining company leveraging IoT applications for a ground-level alert system. Coal mines are complex work environments where safety of workers is of utmost priority; a number of precautions have to be in place to avoid any mishaps. To aid this process, various networked sensors monitor the environment closely, reporting any deviation from an ideal situation.
As illustrated in the following diagram, the sensor data is collected and sent to HANA, monitored regularly with Web applications and servlets. Alerts are triggered when the system senses smoke, fire, earthquake, equipment failure, and so on.
The Key Findings
After many trial and errors, the team spent a lot of time documenting the best practices of using FlashStack to achieve the maximum benefit in the following key areas:
- Data reduction
- Accelerated warm data readiness
- Dashboard queries
- Cost and performance analysis
More than 15 pages of the paper are used in examining various settings and summarizing findings. I don’t want to duplicate the effort here.
For anyone who is interested in using SAP HANA dynamic tiering, the paper is definitely a good reference to check out. Without further ado, here is the link http://go.sap.com/documents/2016/12/b437d5d2-997c-0010-82c7-eda71af511fa.html.