Advice for a SmallBusiness Domain and VSphere

[Radames2019]

Hi,

I'm newbie here, and I ask your help because...

I'm planning to renew my old SBS2011 domain:

20 users, 20 PCs, 20 mailboxes

about 300Gb of network data

about 30Gb of exchange data

In my mind Vsphere 6.7 on a new server (1x Xeon 6230 - 192Gb RAM - 4x1Tb SSD-SAS in RAID10),

The new server has 4port 1Gb NIC, I'd like to team them 2+2 (team+failover).

I'd like to create from scratch 3 VMs with Server2019std (S1=DC+FileServer, S2=Exchange2019, S3=Applications)

I think this HW is good enough, do you agree?

I have more doubts about internal storage, i'm thinking to

64Gb for host partitioned in the RAID10 array (shared with VM)

and I'ld like to assign 600Gb for S1, 200Gb for S2 and 100Gb for S3

but i'm not sure if it is better to keep (for ex) :

S1: Windows and network data in the same "volume C:"

or to create two "volumes" of S1: 100Gb C: (WIN) + 500Gb 😧 (DATA)

or another solution that is more easy to manage and to backup.

Same doubt for S2 and the mailbox database, of course.

What do you suggest?

Thank you very much

and sorry for my english!

[Biemer]

The hardware is good, I don't know which brand it is but we have Dell server with mirror SD cards of 32GB for ESXi and this works fine.

For the VMs part, I always use at one disk for OS and one for data.

[Radames2019]

Thanks for answer,

I don't understand the host running on raided SDs,

isn't better (and safe) to run from the SSD-RAID10?

Some disadvantages I don't see?

Thanx again!

[scott28tt]

Mirrored SDs is a pretty common deployment choice. ESXi doesn’t need large storage for itself, and you want best possible performance (and room to grow) for your VMs.

ESXi Hardware Requirements

[IRIX201110141]

We are a Dell shop....

Your HW looks way over powered to me.

- CPU has 20 Cores so you need additional Windows Server LICs (2x 2 add. Cores)

- You choose a XEON Gold 62xx which comes with 6 memory channels and you choose something like 6x32GB which is a balanced config

S1 = 2vCPU, 8GB

S2 = 2vCPU, 16GB

S3 = 4vCPU, 8GB

iam just guessting. But also with a Tiny VCSA, a OMIVV VA and Veeam Server and Proxy the 192 are way to much. But it depends on the server modell if you can use all 6 memory channels. Sometimes you cant archive the use of 6 channels and also not the 2999Mhz memory speed. Than you back to 4 Channels and 2 Bank which means 64, 128, 256GB of RAM.

The slowest SSD will normaly outperform any 15K disk. So using a SSD RAID 10 for 3 VMs is a pricy option. The standard vendors have 480, 960, 1.92TB.... a 1TB sounds like a customer driver and rather a server "grade" ssd.

If you have the budged for for it. If you need some money for Veeam Backup, iDRAC Enterprise or 7Y 4H MC Support or other stuff take a look to smaller config.

Regards,
Joerg

[Radames2019]

Thanks a lot for the tips!

We have a special financing for "IT innovation in SmallBusiness", so I'm planning a (little) oversize server (HP DL380), able to grow with us in the next, let's say, 4 or 5 years without any HW upgrade.

...And the actual HP ML350G6 (16cores and 48Gb RAM, but only 32Gb usable with SBS2011) is veeery slow and intricate.

Because I'm newbie to VMware, in the beginning I'm thinking to assign

S1=     DC                      8vcpu 16Gb

S2=     ECH                    8vcpu, 128Gb (according to  MS system requirements, just to be absolutely sure)

S3=     FILE and APP     8vcpu, 16Gb

and see what's going on, then calibrate the resources in real life.

I know that the RAID10 SAS-SSD (960Gb - correct) is pricy but the proposed price for SATA SSDs was similar, and I'd like to have both:

• performance (for ex. a APP we use has a terrible indexing system, so every search in the database takes now 10/15 secs - when it's a good day!)
• safety (a couple of 15k died with no notice in the years, this mades the Boss nervous)

Last to say, at the moment I don't have a clear "disaster recovery" solution:

actually I backup the data (network, databases, etc) of the 3 small and old servers we have, but if one dies we lost all his services.

For the future I'm planning to backup the network data on NAS1, and the full VMs on NAS2,

but If the new server dies I don't have any HW able to recover the VMs from backups, i think this is a said point of failure

do you agree?

Please, can you suggest me, in your experience, if and how I can mitigate this PoF?

Thank you again for your patience!

[Biemer]

You can always add more CPU and memory if needed.

For the DC I suggest that you start with 2vCPU and 8GB, if needed you can add more. This can be "hot add" for CPU and memory if you enable this.

This can being used for the File & App server, but here it depents on what kind of app you are talking about.

If that server has a hardware issue then indeed all VMs will be down, but this will also be the case if you update the ESXi.

For the Windows License it is better to scale down the number of cores to 16.

Because Windows 2019 std license is for 16 cores and you are allowed to run 2 VMs with this. So for this CPU this will be 2x Windows 2019 license for 2 VMs, so you need 4x Windows 2019 Std license to run the 3 VMs.

[Radames2019]

Hi, thanx for answer,

but in the price-quote I asked for my plan, the proposed licensing for my server (20cores for 4x VMs max) was always:

2x     Windows 2019 std (allows 1x phisical + 2x virtual each = 4x VMs max)

4x     Addon Windows 2019 2x cores    (add 4cores for each 2019std)

I was quite about that, ... please don't make me worry! :smileyconfused:

For the disaster recovery,

do you think that buy a decent used server with minimun CPU/RAM requirement for my domain and ESXi 6.7U3

is a possible solution? If so is it advisable?

Thank you!

[Biemer]

Windows 2019 Std allow indeed 1x physical and 2x VM, but the physical may only have the hyper-V role.

That covers indeed the license for max 4 VMs.

Yes, I that is a possible solution.

You can also add a extra dual network card to the server and devide the network across these cards, 2 on the embedded card and 2 on the extra card.

Also make sure that the server have dual powersupply with one connected to a UPS.

Then only thing that can cause a disaster is a motherbord problem.

[Radames2019]

Yes, good idea, thank you,

redundancy of the NIC is cheap....

In your experience is a good thing to use the HPE customized ESXi image?

or is better to use the original VMware packge and customize by my own?

Thanks

[Biemer]

Use the customized HPE ESXi image.

[Radames2019]

thanks!