On 9/12/06, Bill Bitner <[EMAIL PROTECTED]> wrote:
It depends. From a response time perspective, the bulk of the
difference is dependent on the hardware. FICON/FCP to DS8000
both will likely give similar response time. Processor time,
particularly in a virtualized environment will vary significantly.
See http://www.vm.ibm.com/perf/reports/zvm/html/520lxd.html
In general on system z, FCP has three areas of potential
advantage:
1. avoids overhead of converting from block to eckd and
back to block oriented in the CU.
2. more I/Os can be executed in parallel (though PAV is
a method for ECKD to minimize this advantage)
3. more data can be moved in a single I/O command
Workloads are impacted differently by the above. There
are also a number of non-performance related
differences.
Bill Bitner - VM Performance Evaluation - IBM Endicott - 607-429-3286
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
Given the results from the IBM benchmark and the comments in this
paper, I am a bit confused as to when to use FCP and when to use ECKD.
Assuming an EMC Symm. disk subsystem, zVM and SLES, wouldn't I want
to keep my multipath simple by using ECKD and manage the I/O's via zVM
using mini-Disks?
Most of my I/O's are solve either in cache under Linux or cache under
zVM or cache in the I/O subsystem before I actually get to a device
(mostly Oracle applications). I do hit heavy writes but normally
off-hours when I don't care much.
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
