HANA News Blog

Intel COD / SNC NUMA feature

Matthias Sander + Jens Gleichmann • 28. September 2023

Possible performance degradation or inconsistencies

Sub-NUMA Clustering by Frank Denneman

--Scroll down for English version


Es gibt ein Intel Feature mit dem Namen SNC bzw. COD. Dabei steht SNC nicht für Secure Network Connection, sondern für Sub-NUMA Clustering. COD steht hier nicht für ein Videospiel, sondern für Cluster-On-Die.


Diese Features werden bei neuerer Hardware eingesetzt und wird ca. seit zwei Jahren teilweise standardmäßig aktiviert (beispielsweise bei HPE).

Die Features dürfen unter keinen Umständen mit HANA verwendet werden. Neben Performanceproblemen können ebenso Inkonsistenzen auftreten!


Es wird ein phys. NUMA Node in zwei NUMA Nodes aufgeteilt. Bei einem unserer HANA Health Checks wird neben der Datenbank auch von Storage über Hypervisor und OS einiges überprüft. Bei einem Kunden, für den wir einen solchen Health Check durchgeführt haben, ist uns aufgefallen, dass der ESXi bei einem phys. 2 Sockelserver plötzlich 4 NUMA Nodes anzeigt.


Auf ESX Ebene kann dies mit dem folgenden Kommando überprüft werden:

echo "CPU Packages";vsish -e dir /hardware/cpu/packageList;echo "NUMA nodes"; vsish -e dir /hardware/cpuTopology/numa/nodes

Das Feature kann im BIOS deaktiviert werden (Beispiel HPE).

####################################################English#########################################################


There is an Intel feature called SNC or COD. SNC does not stand for Secure Network Connection, but for Sub-NUMA Clustering. COD does not stand for a video game, but for Cluster-On-Die.


These features are used in newer hardware and have been partially activated by default for around two years (for example at HPE).

The features may not be used with HANA under any circumstances. In addition to performance problems, inconsistencies can also occur!


A physical NUMA node is divided into two NUMA nodes. During one of our HANA health checks, not only the database but also the storage, hypervisor and OS are checked. We noticed at one customer that the ESXi suddenly shows 4 NUMA nodes on a physical 2 socket server.


At ESX level, this can be checked with the following command:

echo "CPU Packages";vsish -e dir /hardware/cpu/packageList;echo "NUMA nodes"; vsish -e dir /hardware/cpuTopology/numa/nodes

The feature can be deactivated in the BIOS (example HPE).

Source

2470289 - FAQ: SAP HANA Non-Uniform Memory Access (NUMA):

Is it possible to use sub-NUMA clustering for SAP HANA?

On BIOS/UEFI level it is possible to define more than one NUMA node per CPU socket (Intel Cluster-On-Die (COD) / sub-NUMA clustering technology). This scenario isn’t supported and can result in SAP HANA persistence corruptions (SAP Note 2116157).


2116157 - FAQ: SAP HANA Consistency Checks and Corruptions

"Unsupported activation of Intel Cluster-On-Die (COD) / sub-NUMA clustering technology"


Details:

Sub-NUMA Clustering by Frank Denneman


support Intel COD/SNC

SAP HANA News by XLC

Partitioning process
von Jens Gleichmann 26. November 2024
SAP HANA scaling and tuning with proper partitioning designs
HANA Roadmap
von Jens Gleichmann 21. November 2024
End of maintenance for HANA 2.0 SPS05 in 2025 - plan your upgrade path
SAP HANA 2.0 SPS08 Release for Customers
von Jens Gleichmann 21. November 2024
SAP HANA 2.0 SPS08 was released to customers at the end of November 2024. The release has 4 years of maintenance. In parallel, SPS05 will remain under maintenance until December 2025 and SPS07 until April 2028.
Feedback of customer projects and the labyrinth of the roles & responsibilities list
von Jens Gleichmann 18. November 2024
Feedback of customer projects and the labyrinth of the roles & responsibilities list
Performance degradation after upgrade to SPS07
von Jens Gleichmann 5. November 2024
With SPS06 and even stronger in SPS07 the HEX engine was pushed to be used more often. This results on the one hand side in easy scenario to perfect results with lower memory and CPU consumption ending up in faster response times. But in scenarios with FAE (for all entries) together with FDA (fast data access), it can result in bad performance. After some customers upgraded their first systems to SPS07 I recommended to wait for Rev. 73/74. But some started early with Rev. 71/72 and we had to troubleshoot many statement. If you have similar performance issues after the upgrade to SPS07 feel free to contact us! Our current recommendation is to use Rev. 74 with some workarounds. The performance degradation is extreme in systems like EWM and BW with high analytical workload.
HANA OS maintenance
von Jens Gleichmann 29. Oktober 2024
Please notice that when you want to run HANA 2.0 SPS07, you need defined OS levels. As you can see RHEL7 and SLES12 are not certified for SPS07. The SPS07 release of HANA is the basis for the S/4HANA release 2023 which is my recommended go-to release for the next years. Keep in mind that you have to go to SPS07 when you are running SPS06 because it will run out of maintenance end of 2023.
News for the hyperscaler AWS, GCP and MS Azure
von Jens Gleichmann 20. September 2024
news instances with SAPS, memory and CPU values in comparison
Unforeseen cloud cost increases
von Jens Gleichmann 13. September 2024
Unforeseen cloud cost increases - RedHat announced back in January this year that the costs for cloud partners will be changed effective April 1, 2024. They called it scalable pricing.
HANA 2.0 SPS08 Roadmap
von Jens Gleichmann 13. September 2024
SAP HANA 2.0 SPS08 Roadmap and features Q4 2024
SUSE maintenance
von Jens Gleichmann 16. August 2024
How to interpret the SUSE Lifecycle
more
Share by: