Carbon footprint of an in-memory applicance

[Martin_Lauer]

Hi,

can you make any estimation on the carbon footprint, or performance/watt ratio, of an in-memory database (or applicance)?

Keeping all data in memory, i.e. online, could require more power than keeping only the data with frequent access and putting all other data on discs, i.e. offline.

[Former Member]

No news ?

[former_member93896]

It's an interesting question. I will just add my personal view.

First, in-memory databases use disks (or SSD) to persist data. There can also be a distinction between hot (always in-memory) and cold data (only read from disk to memory when needed). So just looking at "more memory uses more power" is not going to cut it.

Secondly, the a single SAP In-memory Appliance might be able to replace several traditional servers. For example, the next version could be used as a database server for SAP NetWeaver BW and replace a BW Accelerator appliance. Future versions of the in-memory appliance will consolidate database and application servers. Overall you will need less hardware i.e. consume less power.

Finally, power consumption of hardware components like CPUs and memory is basically being reduced with every new hardware generation. Since the in-memory appliance uses very current hardware (for example, Intel CPUs), it's save to assume that you get pretty energy efficient solutions.

In summary, I would say an in-memory solution has a significantly smaller footprint than comparable other solutions.

Regards,

Marc

SAP Techology RIG

[lbreddemann]

Hi there!

Concerning the usage of SSDs: this is only required/recommended for the LOG filesystem, but not for the DATA filesystem.

The persisted state of the HANA tables will mainly be stored in the DATA filesystem which could be a shared local filesystem, a filer, a NAS/SAN storage, NFS... whatever the hardware partner thinks of.

SAP does not specifiy the type of storage but certain performance KPIs that need to be fulfilled by it.

Concerning the carbon footprint during operations:

The processors used by HANA allow to automatically raise the speed during high load and to switch off cores during less load as well.

Since more application data can be processed with less technical data movement (e.g. due to compression and cpu-cache-aligned data structures) quite more CPU cycles should contribute to the actual business data processing - which in turn leads to the assumption that HANA will be more efficient concerning energy usage and thereby carbon footprint.

However, the currently tested use-cases for HANA are significantly different from those of the classic DBMS platforms.

Therefore, comparing those right now wouldn't lead to proper results.

regards,

Lars

[Former Member]

GRC team to be in picture to analyze the sustainability.

Regards,

Rajesh