On 22/03/2016 4:02 PM, David Crayford wrote:
On 22/03/2016 1:17 AM, Tom Marchant wrote:
On Tue, 22 Mar 2016 00:16:52 +0800, David Crayford wrote:
z/OS:
DOC:/u/doc/src: >time iospeed
real 0m 1.15s
user 0m 0.62s
sys 0m 0.20s
Dell:
davcra01@cervidae:~$ time ./iospeed
real 0m0.254s
user 0m0.048s
sys 0m0.199s
I have two questions.
1. What do the above figures mean?
real = elapsed (wall) time
user = userspace (application) time
sys = system (kernel) time
Interestingly the numbers are even worse if I use QSAM instead of zFS.
I'm slightly surprised that zFS out-performs QSAM by such a
large margin but not shocked.
Program Name IOSPEEDQ hh:mm:ss.th
Step Name IOSPEED Elapsed Time 11.05
Procedure Step TCB CPU Time 00.41
Return Code 00 SRB CPU Time 00.17
Total I/O 33377 Total CPU Time 00.58
Service Units 9175
The blade server is a Dell PowerEdge E5-2640 SandyBridge-EP connected to
a direct access disk array using 12Gbit/s SAS.
We got 3 blades, the rack and 100TB of disk for $30,000!! Each blade
also has 10TB of internal disk, 256GB RAM. Now that's a steal!
And the core Hyper-V virtualization software is free! We consolidated
all of our existing Windows servers onto one blade and still have 80%
capacity free.
There is obviously quite a lot of suspend time when running the test
on our z/OS system. Otherwise the times are comparable.
2. What happens on z/OS when you put a heavy load on 320 channels all
at once?
I don't know we've only got 14 channels on the z114 at the lab :^) The
Linux server is a VM running under HyperV, which I must say is very
impressive tech.
I've never been much of a fan of Microsoft but the fail over and live
migration capabilities are excellent. Those x86 blades absolutely
thrash our zIIP for running
Java workloads.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
