change1

Author: msjuergent

articles/virtual-machines/workloads/sap/hana-vm-operations.md

@@ -92,7 +92,20 @@ Given that low storage latency is critical for DBMS systems, even as those, like
 
 Accumulating a number of Azure VHDs underneath a RAID, is accumulative from an IOPS and storage throughput side. So, if you put a RAID 0  over 3 x P30 Azure Premium Storage disks, it should give you three times the IOPS and three times the storage throughput of a single Azure Premium Storage P30 disk.
 
-Don't configure Premium Storage caching on the disks used for /hana/data and /hana/log. All the disks building those volumes should have caching of those disks set to 'None'.
+The caching recommendations below are assuming the these I/O characteristics and assumptions for SAP HANA that list like:
+
+- There hardly is any read workload against the HANA data files. Exceptions are large sized I/Os after restart of the HANA instance or reboot of the Azure VM when data is loaded into HANA. Another case of larger read I/Os against data files can be HANA database backups. As a result read caching mostly does not make sense since in most of the cases, all data file volumes need to be read completely.
+- Writing against the data files is experienced in bursts based by HANA savepoints and HANA crash recovery. Writing savepoints is asynchronous and are not holding up any user transactions. Writing data during crash recovery is performance critical in order to get the system responding fast again. However, crash recovery should be rather exceptional situations
+- There are hardly any reads from the HANA redo files. Exceptions are large I/Os when performing transaction log backups, crash recovery, or in the restart phase of a HANA instance.  
+- Main load against the SAP HANA redo log file is writes. Dependent on the nature of workload, you can have I/Os as small as 4 KB or in other cases I/O sizes of 1 MB or more. Write latency against the SAP HANA redo log is performance critical.
+- All writes need to be persisted on disk in a reliable fashion
+
+As a result of these observed I/O patterns by SAP HANA, the caching for the different volumes using Azure Premium Storage should be set like:
+
+- /hana/data - no caching
+- /hana/log - no caching - exception for M-Series (see later in this document)
+- /hana/shared - read caching
+
 
 Also keep the overall VM I/O throughput in mind when sizing or deciding for a VM. Overall VM storage throughput is documented in the article [Memory optimized virtual machine sizes](https://docs.microsoft.com/azure/virtual-machines/linux/sizes-memory).