One powerful server and VS virtualization several "weak" servers of the same type?

Question

System administration

Reshaping the infrastructure of the enterprise. Current situation: 3 PCs running win2k3.
1st: SQL server (base 1.5 GB, rarely used), terminal server (30 users, work with MS Access, Google Chrome, MS Office),
2nd: Terminal server (40 users, work with Google Chrome and MS Office)
3rd: File cleaner, print server, TFTP server for downloading thin clients.
In the future we are going to raise a domain controller, WSUS, WDS.
Possible options:
1) Purchase of three HP DL320e v2 G8 (two with xeon 1220 v3 / 16GB of RAM, one with Pentium G3220 / 4GB of RAM) and distribution of roles according to the previous principle: for terminal servers on a good physical server, everything else on a physical machine on -weaker.
2) Purchase of one HP DL380 G6 (inside two Xeon L5639 / 72GB of RAM) and one self-assembly server on a platform from Intel (Pentium G3220, 8 SFF HDD bays, Constellation ES hard drives) with virtualization of all work tasks on the first one and file washer/storage backups / backup CD in the virtual machine on the second.
Questions:
1) Will the DL380 G6 have enough resources in the above configuration to perform all the above tasks? Any server capacity calculators say that there is still a reserve for the future, but I do not know if they take into account virtualization.
2) How to minimize the risk of failure of the entire system at once?
3) What are the risks when buying EOL servers apart from the expired warranty?
4) The choice of DL380 G6 is due to almost three times lower price than the eighth generation server in a similar configuration. How justified is this in terms of server capabilities? Are there any more significant than the 10% performance increase between L56XX and E5-24XX according to synthetic tests?

Logs:

Answer 1

Both presented options do not like, sorry.
I offer 2 servers for virtualization (xeon e5x2 64GB) and 1 physical (xeon e3 16 GB), the configuration is approximate, you do not specify performance data anywhere.
1st server host VM:
a. SQL server
b. terminal server
c. Print server, TFTP
d. WSUS
2nd server host VM:
a. terminal server
b. file washer
in WDS
, domain controller
3rd server - domain controller, backup.
Virtual machines replicate from server to server.
Answers to questions:
1. It is not known, there is no data from you, for virtualization (VMWARE HYPERV), lay 5%.
2. I wrote above.
3. I don’t want to answer, I can’t formulate my idea beautifully.
4. l5639 - 4858, e5-2430 - 7288, e3-1220 - 7022, where is 10%?
http://www.cpubenchmark.net/compare.php?cmp[]=1983... There are many arguments, I gave only one.
NB! do not forget about the cost of software.
Maybe contact an integrator?

Answer 2

I didn’t read the question itself (it’s very large, and I don’t understand the brands of equipment). I answer according to the title.
1. If you need fault tolerance, and you usually need it, then there should be at least 2 servers, and all your performance requirements should fit in 1 server, the second will be a hot standby.
2. If you don't care about electricity, then a bunch of medium-sized servers will do.
3. And if you want to get flexibility, then it is better to choose a lot of medium-sized servers combined into a cloud platform, i.e. you will kill all the rabbits, such as flexibility (how many machines I want, I have as many), fault tolerance (it is better to let 1 out of 10 servers fail than 1 out of 2), and of course you will get your own problems.

Answer 3

G6 is still generally supported by the vendor then?
in any case, there should be at least 2 hosts in case of a complete server failure.
and so - you can configure for example:
2 hyper-v / vmware hosts with:
1. ADDS
2. SQL
3. WSUS
4. FileServer + printing
The second virtual
RDP server.
in fact, all roles are already either clustered (SQL) or replicated - ADDS, WSUS, FIle (DFS) in one way or another,
and when deployed correctly, this all eats quite a few resources due to server core s 2012R2