However in the same article if we get 25 GBE optical 4 port then we are at 25 GBps apparently
right now our cluster network is at 25 Gbps but IO module is 10 GBps.
My understanding is that cluster network is just for management purposes so we still working with 10 Gbps as I understand that IO module is communication between esxi host and storage disks
Ya i was under impression that cluster network is just for managing the cluster but it has nothing to do IO outputs between hosts and storage
Then i went to datacenter and got this. Here it clearly says that it is 25 GBps but for some reason other guy claims that this doesnt matter but cluster network speed matters. This was so confusing to me. Cluster network doesnt carry IO traffic but these modules do
So this module can be 25 GBps it is just wrong SFP used this is why PowerStore shows 10 GBps
As far as I understand this diagram ports are dedicated for iSCSI traffic NVME traffic they cannot be mix and match for example for iSCSI
My understanding is that if we implement iSCSi then we use port 2 on module and if we are going with NVME TCP/IP then we have to use port 0 on each controller.
Not also clear about LACP bonding can ports for NVME be bonded to form LACP for example from IO module 1 and 2
Example NVME TCP network port 0 on node 1 to be bonded with bonded with port 0 on node 2
based on what you said then we dont even need IO modules.4 port cards port 0 and 1 are connected. I was under impression that these ports are used just for Cluster networking but have no bearing in carrying traffic from host.
IN every other storage manufacturer cluster network is one thing and IO input output from hosts to actual disk is carried over IO modules. Cluster network is just management network but apparently it also carries traffic from hosts to disks which is first time I hear about after almost 10 years working with storage
Direct attached iSCSI or NVMe TCP hosts are supported on other ports, but not for ports 0 and 1 (bond0) on the 4-port card.
from hosts to actual disk is carried over IO modules
Then over here I have 4 ports IO module which is for some reason 10GBps
These ones should carry traffic and Ports 0 and 1 on 4 port card should be only for Cluster network but apparently that is not the case
THe reason this is confusing is that I have extremely bad performance on NVME TCP IP hosts and based on internet articles NVME is supposed to outperform iSCSI however in my case NVME is extremely slow and basically it comes to the point that NVME cannot be used otherwise performance is way worse than iSCSI so I would basically go with iSCSi where I get decent performance.
Based on this I dont even need IO module but what I got from this document networks mapped for 4 port card are just for management and high availability but not IO traffic cause right now IO modules are configured for 10GBps
Apparently it is high misunderstanding on part but I cannot get any decent performance out of NVME TCP
DELL-Josh Cr
Community Manager
•
9693 Posts
•
43676 Points
0
0
Posted December 15th, 2023 14:50
You should call phone support and open a support case so they can take a look.