Data and in-memory databases go together like peanut butter and chocolate. More data can lead to better insights; cutting out the latency from disk I/O can drastically increase application performance to yield business insights more quickly. Running databases and analytics in memory can also reduce costs: a Forrester study commissioned by SAP in 2014 indicated that simply moving large-scale enterprise resource planning (ERP) workloads into memory saved 15 percent on hardware costs (and more still on software and administration).[i] So what’s not to like?
Well, the amounts of memory involved can be a challenge. As the size of in-memory database workloads grows, obtaining enough memory to accommodate those workloads can become a problem. One option is to scale out and run applications across clustered nodes. This works well for some workloads, like SAP Business Warehouse (SAP BW) on the SAP HANA platform and SAP BW/4HANA. But in some cases, scaling up is the only way to go; for workloads like SAP Business Suite on the SAP HANA platform and SAP S/4HANA, more memory on a single node is the only tenable solution for dealing with more data.
In-memory computing and analysis at large scales is expensive, so any variability in price can have consequences for projects going forward. Against this background, Prowess recently investigated scale-up SAP HANA solutions involving 12 TB or more of memory and compared contiguous-memory SAP HANA platform–certified solutions from Hewlett Packard Enterprise (HPE) and Lenovo to evaluate the cost of acquisition and scalability:
- HPE ConvergedSystem 900 for SAP HANA Scale-up Configurations
- HPE Integrity MC990 X Server
- Lenovo Scalable Solution for SAP HANA
We found that the Lenovo Scalable Solution for SAP HANA, which uses a unique implementation of software-defined memory, provided both the lowest acquisition cost of the solutions tested and the most flexibility for scaling up memory after purchase.
The Lenovo Scalable Solution for SAP HANA is centered on the Lenovo System x3950 X6 server, which provides the computational muscle and, on its own, functions as an 8-TB SAP HANA solution. However, the Lenovo solution can also make use of Lenovo System x3850 X6 servers as memory-donor nodes (Lenovo’s name for them). In this configuration, the Lenovo System x3950 X6 server provides 6 TB of memory and each Lenovo System x3850 X6 server contributes an additional 3 TB. By adding memory-donor nodes in this fashion, the Lenovo solution can scale up to 36 TB.
The Lenovo Scalable Solution for SAP HANA is less expensive up front. For example, a 12-TB configuration is 21–35 percent less expensive than a 12-TB HPE ConvergedSystem 900 for SAP HANA appliance. And an 18-TB Lenovo Scalable Solution for SAP HANA is 25 percent less expensive than a 16-TB HPE ConvergedSystem 900 for SAP HANA appliance (Figure 1).[ii]
Figure 1. Comparison of the up-front costs for the Lenovo Scalable Solution for SAP HANA (12-TB and 16-TB configurations) versus HPE ConvergedSystem 900 for SAP HANA Scale-up Configurations (12S/12 TB, 16S/12 TB, and 16S/16 TB appliances)
Flexible in-field memory upgrades
Not only did the Lenovo Scalable Solution for SAP HANA have the lowest costs to acquire and operate, but it also provided the most flexibility for scaling up memory after purchase. In-field memory upgrades are possible with the HPE ConvergedSystem 900, but only up to 12 TB of memory.[iii] In-field memory upgrades are not available for HPE Integrity MC990 X Server scale-up configurations certified for the SAP HANA platform.
In contrast, not only does the Lenovo Scalable Solution for SAP HANA enable large contiguous-memory configurations (up to 36 TB), the modular approach to memory scalability in the Lenovo solution also means that customers do not have to confront the dilemma of either buying an SAP HANA appliance that they might soon outgrow or paying for additional memory headroom that they might not ever use.
Acquisition cost isn’t everything, but it can matter Acquisition cost is by no means definitive for any IT project, and the total cost of ownership for large-scale, contiguous-memory SAP HANA solutions can be significant. However, acquisition cost does provide a starting point for further analysis and for comparing the figures that will be the first things to catch purchasing managers’ eyes.
[i] Forrester. “Projected Cost Analysis of SAP HANA.” Commissioned by SAP. April 2014. Study summarized at the SAP news site: https://news.sap.com/forrester-study-savings-potential-sap-hana/.
[ii] $1,345,292 for the Lenovo Scalable Solution for the SAP HANA platform versus $1,630,748, $1,816,414, and $1,929,150 for the HPE ConvergedSystem 900 12S/12 TB, 16S/12 TB, and 16S/16 TB configurations, respectively. Prices based on figures received by Prowess Consulting from Lenovo and HPE.
[iii] HPE. “HPE ConvergedSystem 900 for SAP HANA Scale-up Configurations.” www.hpe.com/us/en/product-catalog/detail/pip.specifications.hpe-convergedsystem-900-for-sap-hana-scale-up-configurations.7311041.html.