Quite right. You have to consider the whole stack, from storage/networking to the hipervisor, guest OS, database and the applications. If you only consider that a Xeon is cheaper than an IFL it is anybody's guess what the final functional outcome for the customer will be. ________________________________________ Van: Linux on 390 Port <[email protected]> namens Sergey Korzhevsky <[email protected]> Verzonden: woensdag 3 februari 2016 09:44 Aan: [email protected] Onderwerp: Re: TCO comparison Intel vs EC12 IFLs
Just to add 5 cents. Besides hardware, you also changed database which may have HUGE impact on performance. 03-02-2016 01:28:09 "Harder, Pieter" wrote: >In fact the disk system didn't change with our migration. The VMWare farm was >run against the same IBM DS8000 the z used. Most of the DS8000 space was used >for open systems anyway. If memory serves me right before the DS8000 went out >the door we had 48 disk sets in it and only 6 were used for the z. And the x86 >world was connected by the same Cisco SAN switches that were used for >connecting the open storage on the z. But a world of difference in resulting >performance. > >best regards, > >Pieter Harder >________________________________________ >Van: Linux on 390 Port <[email protected]> namens Philipp Kern ><[email protected]> >Verzonden: dinsdag 2 februari 2016 22:24 >Aan: [email protected] >Onderwerp: Re: TCO comparison Intel vs EC12 IFLs > >On 2016-02-02 14:54, Harder, Pieter wrote: >> I think there is another factor involved. On x86 you usually run >> database engines etc with very large caches to avoid going to disk. On >> z this is much less of a penalty because of the powerful I/O >> subsystem. I have some experience with this as we moved our SAP >> systems from z to x86 as part of a strategic directive to focus on one >> platform. We ended up running the same systems with about ten times >> the amount of memory on the databases to get like performance. Both >> platforms were virtualized (Linux on zVM with DB2 versus Windows on >> VMWare with MSSQL). > >But is it really the capability of the I/O subsystem or is it just that >the storage racks needed for the z are much more beefy than the Netapps >you usually get for "distributed" virtualization? > >I do buy that there's more bandwidth available / less latency, although >a lot of that depends on how you connect the storage on x86. > >Kind regards and thanks >Philipp Kern > >---------------------------------------------------------------------- >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 >---------------------------------------------------------------------- >For more information on Linux on System z, visit >http://wiki.linuxvm.org/ > >---------------------------------------------------------------------- >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 >---------------------------------------------------------------------- >For more information on Linux on System z, visit >http://wiki.linuxvm.org/ WBR, Sergey ---------------------------------------------------------------------- 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 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/ ---------------------------------------------------------------------- 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 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
