We are using the new VNX Snapshot technology on our VNX 7500.
The new VNX snapshots put a different load on the array than the old Snapview Snapshots did.
We see that when the VNX snapshots delete the old snapshots on the array, that the lun iops go way up. We have been told that when the VNX snapshots delete old snapshots they are writing zeros to the silver of disk that the old snapshots occupied.
Does anyone else have experience with the new VNX snapshots on a VNX 7500? We are finding that the new VNX snapshots put a very heavy I/O load on the the thin pool. We are thinking about going back to Snap view snapshots becuase it will allow us to write the snapshots to the seperate RLP luns and therefore not cause slowness\loading on the primary luns.
I am looking to see if others have experience with the new VNX snapshot techonlogy and have developed ways to cope with the performance impact the technology has on the thin pool on the array.,
We have similare issuses with our VNX5300. EMC support indicates that we take snapthos to frequently. We take snapshots every 15 min for some LUNs. We see snaphots "hanging" around in destorying state a long time before they are actually removed. Support says this is working as designed so we have no other option then to move back to recoverpoint. Another point from support was that our pool has to many slow NL-SAS drives but we see the same slowness of deletion of snapshots in the SAS tier... If you need a pure SSD array to be able to run VNX snapshots then the product is useless.
So our conclusion is that the product has way more limitations then http://www.emc.com/collateral/software/white-papers/h10858-vnx-snapshots-wp.pdf tells you about. If you are thinking about using this product, test it carefully.. Our problems started to show themselfs after a month in production and no one at EMC or thier partners seems to "know" about these limitations.
05.32.000.5.011. No we dont have flash drives in our pool, only SAS and NL-SAS. We dont need SSD for our host IOPS and I find no documentation that we need it if we want to use VNX snapshots.
We only run block.
for thin LUNs and snapshots even as low as just two EFD drives typically helps performance.
This because these configs are more meta data intensive and from VNX OE 32 onwards we store the slices containing the meta data on flash drives (if there are any in the pool)
Until you can give me real proof or show me documentations regarding vnx snap and slices going back and forth in the pool I will have a hard time beliving that a couple of SSD solves the issue.
It's my understanding that the deletion process actual move around a lot of data (or writes zeros?) and it does it in the background as an low prio activity (?), which in turn takes time. This will affect how frequent you can take snapshots, if you want to do it like us every 15 minutes the system wont be able to delete snaps in time and therefore reach max snap per LUN.
Please tell me that I'm wrong!