Design Advice (SoHo)

Rand__

Well-Known Member
Mar 6, 2014
4,572
910
113
Hi,
looking for some design advice.
Currently run a vSphere/vSan environment which I am not entirely happy with as indicated in other threads, primarily relatively slow performance for the invested hardware as well as instability while tinkering (might be user fault).

Before I start adding/repurposing more hardware I thought I should discuss with someone;)

Basically I have these VMs for Home use - there is a bunch more for work topics but those are not critical (ie backup on demand, and mirrored datastore are sufficient):


Critical Infrastructure - HA
PDC (1xVM, can't mirror all components afaik [DHCP], [WSUS?]), currently using FT
Sophos (1x VM, issues mirroring), currently using FT
Horizon View, 1 VM, (not sure can mirror) currently using
vCenter (3VMs) (vCenter HA setup)

Critical Applications - HA
Client VMs (2)

Noncritical Infrastructure (No HA)

FreeNas (x2)
Backup Tape (Win)
Backup Disk/VM (Nakivo Appliance)
Powerchute Appliance
Monitoring/Logging (3)

Noncritical Applications (No HA)
Multimedia (2)
Client VMs (2+)


I do see several options:
1. Stop complaining and just use it the way it is - 3 (or4) nodes are enough for my use case
2. Add another node to vsan or some more disk (scale out/up)
3. use another design (2 * Robo+Witness) for improved availability (ie two hard failure domains). Would fix my Nakivo license issue too;)
4. Use another platform - not sure which one would be most applicable, but I guess I can run vmware and just replace vsan as shared datastore with something else

I currently have a bunch of E5 servers (4) I can use for this and also some E3/i7 boxes than I can beef up to use. I think I overextended a bit though (STH virus) so looking for some back to earth or best practices input:)

Thanks
 

Dean

Member
Jun 18, 2015
97
8
8
45
I'm about to setup Robo, with witness.

Sent from my Moto Z (2) using Tapatalk
 

Evan

Well-Known Member
Jan 6, 2016
3,123
522
113
Maybe time i get a demo unit for fun if they can ship one with a good amount of NVMe drives
 

Rand__

Well-Known Member
Mar 6, 2014
4,572
910
113
But since you're interested - since then i went from single diskgroup design to double dg's with 2 stripes per objects. That caused issues with data evacuation (couldnt completely evac a single node, only 'ensure availability') so I am currently implementing 6 nodes. Not sure whether I stay there yet;)
Also tried SSD based setup in the meantime (more drives) but that was not really good.
Want to look at ScaleIO as well... intriguing idea to use all drives all the time to pull/push data, especially without dedicated hba. That means slower performance per drive (ie due to sata attachment) but enables larger drive numbers. Not sure how it works out, just read documentation yet; was hoping for 3.0 to be out soon but it's still pending...
 

Evan

Well-Known Member
Jan 6, 2016
3,123
522
113
You know that thread was from May?;)
Oops , your right and don’t know what I was thinking I guess I just wasn’t thinking when somebody bumped it. So has anybody used one anyway ? Since it’s so old surprised nobody has reported how good (or bad) it is yet.