Leigh Kalbli
asked on
iscsi nested esxi.
setting up a nested esxi lab to do some testing and learning. I have a dell R720 xd with dual 8 core CPU and 128 GB RAM. 6 x 300gb 10k SaS drives in RAID 10. I created four esxi virtual machines and have them stored on the local datastore. They will all be in a cluster with one as dedicated spare I then was planning on configuring the nested esxi hosts to connect to an iscsi san where the test virtual machines would be stored. I however was thinking about if it was necessary to have the local storage for the nested hyperviors or i could just put them on a LUN on the san. So the host would have an iscsi connection to the LUN with the hypervisors, then the hypervisors would be set up with iscsi back to the same SAN where the vm’s are. Would that be overkill and have a heavy impact on iscsi performance? The only reasoning i had behind it was to a. not have local storage, b. have the hypervisors on the SAN that is snapshot and replicated periodically.
ASKER
Hayes. i should have mentioned that the r720 has daul 1 gb for vm management and dual 10gb for iscsi and vm traffic.
ASKER CERTIFIED SOLUTION
membership
Create a free account to see this answer
Signing up is free and takes 30 seconds. No credit card required.
Any shared storage will give you options to Lab Shared Storage Functions in your Lab.
Install FreeNAS and create iSCSI LUNs.
Install FreeNAS and create iSCSI LUNs.
you may well use iscsi for the nested ESXs. the system-related traffic will drop next to zero shortly after boot.
the inner guest vms can either directly use iscsi or the local store. either way, that's one iscsi xfer. the former is a bit faster because there is one less layer of virtualisation, but barely : the local virtual io layer has neglectible performance impact compared to iscsi anyway.
depending on your goals, both are meaningful. separate targets are good to experiment with vmotion, local is better if you plan good I/O perfs on barebone and want to check what you need still works decently with local storage. other setups might be meaningful depending on the usage of the lab. in doubt, use separate iscsi targets for each guest or nested guest.
the inner guest vms can either directly use iscsi or the local store. either way, that's one iscsi xfer. the former is a bit faster because there is one less layer of virtualisation, but barely : the local virtual io layer has neglectible performance impact compared to iscsi anyway.
depending on your goals, both are meaningful. separate targets are good to experiment with vmotion, local is better if you plan good I/O perfs on barebone and want to check what you need still works decently with local storage. other setups might be meaningful depending on the usage of the lab. in doubt, use separate iscsi targets for each guest or nested guest.
Since its a test environment - if your not doing "heavy load" type operations, you might get away with 1GB (but i doubt it)