Virtual appliance transport

[Giacomo_N]

Hi guys, tomorrow we'll begin the migrate from Nutanix to VXrail (Vsan).
Actually we used Direct access by NFS, but on Vsan enviroment we need to switch on Virtual appliance transport mode.
For the best performance, the esxi IP must be on the same network of Veeam server (without routing between) or nothing matters?

Thanks in advance!

[Andreas Neufert]

There is nothing really special to plan for other than installing the proxies on VMs on the cluster and use the same SCSI adapters in the VM as your workload VMs have. (Usually the LSI)

If you want to optimize for performance and optimal dataflow, then add more small Proxies and spread them accross the vSphere cluster as we have a logic for optimal dataflow within vSAN. For example if you install a Proxy (potentially Linux Proxy) on each host and we will choose the optimal patch automatically to reduce east/west traffic within vSAN.

https://helpcenter.veeam.com/docs/backu ... ml?ver=100

[Giacomo_N]

Thank you Andreas for the reply,

ok I'm ready, I can reuse the actual proxy (one per esx host), I've the same Paravirtual controller on all VM's.

[Gostev]

Hi guys, tomorrow we'll begin the migrate from Nutanix to VXrail (Vsan)

I'm just curious what prompted you to migrate?

[Giacomo_N]

Hi "legendary" Gostev!
The main reason is the performance, after Vsan POC I've verified the claimed "low footprint" storage Vsan (because are integrated on esx kernerl..bla,bla) is true.
Less CPU/host overhead (average 20%) compared to Nutanix CVM and low latency disk access.
Second, Nutanix now are very "Acropolis" oriented compared to five years ago (we bought Nutanix on 2016), we're Vmware oriented.
Third, on Nutanx stretched cluster only 50 Vms are supported on Vmware, actually we've 60 Vms.
Fourth, less hardware customization compared to Vxrail/Dell, we choice 6 hosts with single socket 20 core CPU (Gold 6242R). The same configuration are not possible with Nutanix Hardware, only by partner and support "isn't the same".

[Gostev]

Got it, thanks for sharing!

[NightBird]

Whouah, Nutanix is limited to 50VM in vSphere Stretched Cluster ???

[msteingass]

In short - no

Theres a limit for the amount of files in a protected datastore and a limit for Protection Domains which are a logical construct for configuring Metro Availability on a container.

I can‘t say exactly for how much VMs that makes but definitely more than 50. Maybe there was a misunderstanding with the Protection Domains.

https://portal.nutanix.com/page/documen ... ity-c.html

I‘m familiar with Nutanix and also curious about the other points. But I don‘t intend to start a discussion. Maybe VxRail just was more fitting in some way.

[Matts N]

Yes, is that correct? Only 50 VM's on a stretched Nutanix VMware cluster??? Would explain a lot of problems we had and to some extent still have at a customer (300+ VM's). The problems were dire enough for us to decide not to sell Nutanix for stretched clusters anymore, only suggest VxRail now.

[Matts N]

Had a quick chat with my colleague who works with Nutanix. The 50 VM limitation is per datastore, not for the total cluster.

[Giacomo_N]

Hi guys, sorry for my "disinformation", for sure was a limit some years ago, I remember the support confirm me the limit in 2017.

Anyway, Vxrail implemented and I can confirm, using the Moritz words "VxRail just was more fitting in some way".
Very fast and very low host resource footprint. DR procedure, failover and failback, are fast like a single site farm; With Nutanix I've build a procedure during the last 5 years of about 10 pages and required 2 hours.

By Veeam side, Virtual appliance mode, one proxy per host (affinity rule "should") and registry setting to increase the maximum snapshot per datastore, give me a 5X performance compared to old Nutanix (old Nutanix=old Hardware Hybrid HDD/SSD compared to SSD only in Vxrail).