Setup for the z/VM large memory tests
To emulate a customer-like environment we configured our Linux systems, database server, the z/VM memory, and the OLTP workload software. We also configured our system to use CMMA and VMRM-CMM.
Linux Setup
As described in z/VM guest setup each Linux® system has three CPUs and 16 GB memory. The swap space has a size of 2 GB and resides on a DASD device.
Database server
We used a popular database product for Linux on System z® as a database server. The focus was on the memory management behavior of z/VM®. We felt any database product would provide a good example of high memory usage. Therefore, we describe the database setup only in general terms. The database workload used was a typical OLTP transactional workload.
The database server was setup to use asynchronous disk I/O and didn't bypass the Linux page cache for file system operations.
- Five physical volumes - 3390 Mod 9s
- One volume group created from the five physical volumes
- One logical volume defined from the volume group with five stripes and a stripe size of 32 K
- The total database size was 30 GB
The database cache size was 8500 MB, the total memory size allocated for all the database buffers and caches was 9152 MB. The database caches were sized in such a way that the system did not swap.
z/VM memory management overview
The z/VM system maps the guests' virtual memory into the real memory storage of the System z machine. If there are not enough real memory frames to contain all the required active guests' virtual memory pages, the active guests' virtual pages are moved to XSTOR.
Once XSTOR becomes full, the guests' pages are moved from XSTOR to DASD paging space.
Figure 1 provides a picture view of the z/VM memory management overview.
