Unsolved
This post is more than 5 years old
48 Posts
0
1248
June 10th, 2008 03:00
General question - DFW/CFW
Hi,
In a mainframe systems from historical point of view there are two methods to improve I/O operations:
1. Cache Fast Write (CFW)
2. DASD Fast Write (DSW)
In original IBM implementation we can use CFW for temporary data sets, and we don¿t need use NVS cache (battery backed up). The data is still in cache until jobs or step has been ended.
DASD Fast Write required NVS cache. Data in cache has been destaged asynchronous to the disks.
Now, in DMX-3 we have fully mirrored and battery protected all cache.
So, what is the difference between DFW and CFW in DMX implementation ?
I guess, that should be because EMC recommend switch of only CFW
Can I switch off (using IDCAMS) both: DFW and CFW or it is not possible like in IBM ESS and DS8xxx ?
Even If EMC switch off CFW on DMX, I still observe CFW activities (from z/OS RMF reports). Is it normal ? Does EMC DMX-3 ¿deceive¿ z/OS ?
Marcin
In a mainframe systems from historical point of view there are two methods to improve I/O operations:
1. Cache Fast Write (CFW)
2. DASD Fast Write (DSW)
In original IBM implementation we can use CFW for temporary data sets, and we don¿t need use NVS cache (battery backed up). The data is still in cache until jobs or step has been ended.
DASD Fast Write required NVS cache. Data in cache has been destaged asynchronous to the disks.
Now, in DMX-3 we have fully mirrored and battery protected all cache.
So, what is the difference between DFW and CFW in DMX implementation ?
I guess, that should be because EMC recommend switch of only CFW
Can I switch off (using IDCAMS) both: DFW and CFW or it is not possible like in IBM ESS and DS8xxx ?
Even If EMC switch off CFW on DMX, I still observe CFW activities (from z/OS RMF reports). Is it normal ? Does EMC DMX-3 ¿deceive¿ z/OS ?
Marcin
No Events found!
Notach
21 Posts
0
June 13th, 2008 00:00
a long time ago there was a 3990-6 where CFW & DFW were implemented. Cache was very small and NVS was used too as you said.
So today DMX is gerenated as 2105 ControlUNIT (as shark ESS IBM dasd) and cache is big and protected. ALL IDCAMS commands to manage CFW & DFW exist but i Think they have no effect on DMX 3 an only EMC can disable or enable CFW by apply a new binfile on each DMX3. Today we have a call with EMC to discuss this point since the ETA emc186911.
So untill now we don't see any malfunction since ths ETA on our DMX.
have a nice day
Message was edited by:
Notach
Notach
21 Posts
0
June 13th, 2008 02:00
apparenty only binfile can disable CFW. I ask EMC technical and if this problem occcurs we can loose one or more Chpid on Control Unit.
So, we have Autoswap to prevent a complete shutdown of Lpars .
But...we apply newbinfile asap too.
have a nice day
Davidy8s
154 Posts
0
June 13th, 2008 05:00
If the Benevolent Mainframe Moderator could conceive of a way to link the knowledgebase entry herein, he would do so.
Dave Yates
EMC TS3
Benevolent Host S/W & Mainframe forum Moderator.
xe2sdc
2 Intern
•
2.8K Posts
0
June 13th, 2008 07:00
-s-
MarkF4
39 Posts
0
June 13th, 2008 07:00
Mark Foreman
Another Moderator
Davidy8s
154 Posts
0
June 13th, 2008 09:00
more then benevolent
need a good moderator you'll have to look elsewhere
Hey! Benevolent is good! Who said anything about helpful?
You guys are playing "good moderator / bad moderator" on me or something.
Just so the search engine(s) might actually pick it up:
This thread displays
How to hyperlink directly to a primus solution from a forum posting
How to hyperlink directly to a knowledgebase entry from a forum posting
Dave Yates
EMC TS3
Benevolent - and now helpful - Host S/W & Mainframe forum moderator who may never reach Stefano's number of posts...
xe2sdc
2 Intern
•
2.8K Posts
0
June 13th, 2008 15:00
Yes .. now you are a benevolent and helpful host !!
BTW if you want to reach my number of posts you have simply to find a lot of newbies needing assistence. They'll put a lot of questions and you'll have to give a lot of answers ...
I suspect that MF isn't a place for newbies
Message was edited by:
Stefano Del Corno