HP-UX HB v13.00 Ch-22 - Performance and Tuning

HP-UX Handbook – Rev 13.00 Page 7 (of 19)
Chapter 22 Performance & Tuning
October 29, 2013
NOTE:
Memory is allocated on sequential basis; it must be allocated in contiguous segments. Due to this
protocol, while there may be enough total memory free for a process , if there isn’t a large
enough contiguous segment , the call to memory will fail.
Whether in kernel or user mode, the system will allocate the first free segment that satisfies the
request.
The following parameters define the maximum individual segment size for the
resource:
All parameters are limited to quadrant boundaries for their bitness.
The amount of free contiguous space in a quadrant determines if the request
for these resources can be satisfied.
maxdsiz -the maximum allowed size for the Data segment of a 32-bit process
maxdsiz_64bit -the maximum allowed size for the Data segment of a 64-bit
process
maxssiz- the maximum allowed size for the kernel Stack segment of a 32-bit
process
maxssiz_64 bit- the maximum allowed size for the kernel Stack segment of a
64-bit process
maxtsiz- the maximum allowed size for the Text segment of a 32-bit process
maxtsiz_64bit- the maximum allowed size for the Text segment of a 64-bit
process
shmmax — maximum size (in bytes) for a System V shared memory segment
Text segments get first priority allocation in quadrant 1. Kernel Stack
segments get first priority allocation in quadrant 2
Buffer Cache/File Cache
Recent improvements to the buffer/file cache performance allow larger caches
to be used without the severe performance penalties seen in the past. In
11.31 the Dynamic Buffer Cache was replaced by the VxFS file cache. This
operates similarly and has the same 50% default value. The parameters
filecache_min and filecache_max define the range of this cache.
As with the buffer cache in earlier revisions, % rcache and %wcache should
be greater than 90% in a well tuned system . The cache parameters are
dynamic and require no reboot in 11.31
In modern systems , there are a number of factors to determine in sizing the caches.
 Is the system using a disk array / SAN that has onboard memory for caching read/ writes?
 Does the system use a database that contains an internal buffer pool ? (Oracle as an
example)
 Does the system use Online JFS direct mounts , bypassing the cache?
 Are the I/O’s larger than the cache threshold ( discovered_direct_iosz) ?