This post is more than 5 years old
148 Posts
0
3143
Question regarding Write pending slots
Hi All,
below is the out from symcfg -sid xxxx list -v output. Can someone explain what is the diffrence between below terms
1. # of Available Cache Slots
2. Max # of System Write Pending Slots
3. Max # of DA Write Pending Slots
4. Max # of Device Write Pending Slots
| Cache Size (Mirrored) | : 262144 (MB) |
| # of Available Cache Slots | : 1720537 |
Max # of System Write Pending Slots : 1377773
| Max # of DA Write Pending Slots | : 688886 |
Max # of Device Write Pending Slots : 68885
| Replication Cache Usage (Percent) | : | N/A |
Thanks,
Quincy561
1.3K Posts
0
January 17th, 2012 10:00
#1 is the total number of 64k slots available for user data (DMX3 and above) DMX2 and below used 32k slots.
#2 is the total number of these slots that can be written to but not yet destaged (Write Pending), usually 80% of the total slots
#3 Not used in VMAX. I don't think it is used in DMX3 or 4 either. Even if it was, it was rarely envoked (see #2 and #4)
#4 The total number of slots for a single Symmetrix device that can be write pending. Usually 5% of the writable cache (#3)
anoopcr
148 Posts
0
January 17th, 2012 10:00
So if we have few devices which reaching its max device WP,that will impact overall system WP
Quincy561
1.3K Posts
0
January 17th, 2012 10:00
If the device WP limit was 20,000 slots and the system WP limit was 400,000 , and two devices had reached the limit it would not impact the writes of other devices, just the devices that reached the limits. The system WP count would be 40,000 in this example.
Quincy561
1.3K Posts
0
January 17th, 2012 11:00
If a device reaches WP limit, or the system reaches WP limit, no further writes are allowed until a slot is destaged and the WP flag is cleared. This will increase the write response time.
anoopcr
148 Posts
0
January 17th, 2012 11:00
What my understand is when write pending crosses 80%, it has to wait for a free slot that inturn result in increase in response time
Is that correct ?
Quincy561
1.3K Posts
0
March 14th, 2012 19:00
Yes, #4 is 5% of #2, my bad.
SKT2
2 Intern
2 Intern
•
1.3K Posts
0
March 14th, 2012 19:00
#4 is 5% of #2; isn't it??
SKT2
2 Intern
2 Intern
•
1.3K Posts
0
March 15th, 2012 17:00
also in addition there is a stage called "priority destage" which happens at 50%(of #2) line or a logical volume level at 100%((#4) on the WP slots.At this stage the frame starts destaging aggressivly and at the same time starts to throttle WR from the front end host resulting in elongation of resposne time for host writes. You will start noticing RD miss times increasing since DA will be busier now with writes (even though WR and RD miss have same priority)