The latest Monitiq agent bundle for HANA gives you detailed metrics memory usage within HANA systems. We aim to provide the necessary metrics for performance and capacity planning. If you think we are missing anything please let me know.
All of the instances of “GiB” refer to Gigabytes or 1024 MiB
HANA memory usage metrics collected on a per instance basis. Use this display to review the overall usage of memory across the cluster on a per instance basis.
HANA memory usage metrics collected per instance, by cluster node. Use this display to compare the distribution of data between nodes in the cluster, and assess capacity on a node by node basis.
Row Memory in use GiB (RvC/Node)
The amount of memory in consumed by row store tables. Row store tables need to be in memory before HANA will restart
Column Memory in use (loaded) GiB (RvC/Node)
The amount of memory consumed by column store tables.
Column store tables are loaded into memory on demand and unloaded when there is contention for memory. A production system should have all of its column stores in memory. See also Column memory Unloaded GiB (RvC/Node)
Memory Size in delta GiB (RvC/Node)
When data is loaded into HANA it is loaded into delta table space and is then converted into Column store. You only see significant values here during large loads of data.
Memory Size in History GiB (RvC/Node)
Current memory consumption in “History Column” tables. Tables with session type HISTORY support time travel queries. Time travel queries are queries against historical states of the database.
Column memory Unloaded GiB (RvC/Node)
The total size of Column storage that is not currently in memory. This could be because the database has recently been started and the data has not yet been requested, or the system is oversubscribed and has forcibly unloaded data to created working space. When unloads happen the hanadbcsunload (unld) monitor will register unloads, and you will also see HANA statistic server events picked up by the hanadbevents (evnt) monitor. The level of commitment can be assessed by reviewing the current “Row and column commitment as percent of allocation limit“ as discussed below.
Current allocated memory GiB (RvC) / HANA total GB (Node)
This is the total footprint of memory used by this HANA instance including all row and column storage in memory and computational memory (i.e. memory used by HANA processes that is not directly associated with data storage). In the stacked 24×7 graph the computational memory is split out from the total allocated and is referred to as Temporary Computations GiB.
Temporary Computations GiB (RvC)
The memory used by HANA processes that is not directly associated with data storage, i.e. the total memory used by HANA processes less the column, row, delta and history table store values.
Global Allocation limit GiB (RvC)
The total allowable memory the HANA instance is permitted to consume across the cluster.
Node Allocation Limit GiB (Node)
The total allowable memory the HANA instance is permitted to consume on each node in the cluster
Row and Column loaded as a percentage of total allocation (RvC)
The sum of row and column storage as a percentage of the current memory usage for this instance. This metric indicates the proportion of the memory currently consumed by data storage rather than computational use.
Row and column commitment as percent of allocation limit (RvC)
The sum of row and column storage as a percentage of the total memory permitted to be used by this instance for this instance. This metric gives a measure of the capacity of the system. Current recommendations from SAP state that column + row should be below 50% of the available memory. This metric allows you to track the capacity of your system.
Row Column Percent of Global Allocation (Node)
The sum of row and column storage on the specified cluster node and instance as a percentage of the total allocation limit specified cluster node and instance. Current recommendations from SAP state that column + row should be below 50% of the available memory. This metric allows you to track the capacity of each node in a HANA cluster.