Ste
125 Posts
Quote from Ste on July 1, 2022, 4:43 pm
Hello, while moving big VMdisks between a Petasan cluster and an Equallogic SAN, I noticed that the Petasan read throughput was around 150-200 MB/s, while in writing it hardly reached the value of 40 MB/s (see pic), in fact copying in one direction took one hour, while in the way back it took more than 4 hours !
Is it normal this big difference ? All components (Petasan, Hosts, Equallogic) have 10 Gbit/s connectivity, on the same optical switch. Or maybe is there something that might be better tuned ?
Thanks and bye, Ste.
Hello, while moving big VMdisks between a Petasan cluster and an Equallogic SAN, I noticed that the Petasan read throughput was around 150-200 MB/s, while in writing it hardly reached the value of 40 MB/s (see pic), in fact copying in one direction took one hour, while in the way back it took more than 4 hours !
Is it normal this big difference ? All components (Petasan, Hosts, Equallogic) have 10 Gbit/s connectivity, on the same optical switch. Or maybe is there something that might be better tuned ?
Thanks and bye, Ste.
admin
2,930 Posts
Quote from admin on July 1, 2022, 5:00 pm
-Can you describe the hardware, the disks and their types
-What is the client OS: Windows/Linux ? is virtual/baremetal ?
-How are you copying : using cp command, windows copy, is it 1 file at a time ? Or is this vm migration done from vmware ?
-Are you using a replicated 3x pool ?
-Can you describe the hardware, the disks and their types
-What is the client OS: Windows/Linux ? is virtual/baremetal ?
-How are you copying : using cp command, windows copy, is it 1 file at a time ? Or is this vm migration done from vmware ?
-Are you using a replicated 3x pool ?
Ste
125 Posts
Quote from Ste on July 6, 2022, 9:44 am
Petasan cluster HW:
- HD are HGST (10 TB and 14 TB),
- the journals are on NVMe SSD Samsung 980.
- Intel i7 cpu
- 64 GB ram per node.
- Two 10 Gbps SFP+ nics (one for backed and one for iSCSI).
- pools with replication 3x
- both datastores (petasan end Equallogic) are created on iSCSI volumes
The other SAN is a Dell Equallogic PS6210X, equipped with 24x 2.5" SAS in RAID6.
The transfer is done with "storage vmotion" inside vCenter 6.7, in both cases the same VM with a 480 GB vmdk file, from petasan -> Equallogic in the first move, and on the way back (Equallogic -> Petqasan) in the second move, same VM (linux CentOS 8) same files.
Bye, Ste.
Petasan cluster HW:
- HD are HGST (10 TB and 14 TB),
- the journals are on NVMe SSD Samsung 980.
- Intel i7 cpu
- 64 GB ram per node.
- Two 10 Gbps SFP+ nics (one for backed and one for iSCSI).
- pools with replication 3x
- both datastores (petasan end Equallogic) are created on iSCSI volumes
The other SAN is a Dell Equallogic PS6210X, equipped with 24x 2.5" SAS in RAID6.
The transfer is done with "storage vmotion" inside vCenter 6.7, in both cases the same VM with a 480 GB vmdk file, from petasan -> Equallogic in the first move, and on the way back (Equallogic -> Petqasan) in the second move, same VM (linux CentOS 8) same files.
Bye, Ste.
admin
2,930 Posts
Quote from admin on July 6, 2022, 3:49 pm
Yes reads are faster, in writes you have to write to several replicas across the network before returning.
vmotion across different clusters are known to give low performance, it is done using a non ideal block size of 64k. PetaSAN does support VAAI to accelerate vmotion when moving vms within the same cluster. We are able to reach 400 MB/s writes per vm using SSD cluster. Refer to our vmware guide as to recommendation to get good vmotion setup.
Yes reads are faster, in writes you have to write to several replicas across the network before returning.
vmotion across different clusters are known to give low performance, it is done using a non ideal block size of 64k. PetaSAN does support VAAI to accelerate vmotion when moving vms within the same cluster. We are able to reach 400 MB/s writes per vm using SSD cluster. Refer to our vmware guide as to recommendation to get good vmotion setup.
