How to correctly allocate server resources for virtual machines?

Question

linux

Hello.
Tell me how to properly allocate resources for virtual machines?
We have an HP Proliant DL380 G8, with 2 XEON E5-2630 2.6GHz processors, 32GB of RAM, and 4x600 SAS drives.
Already installed:
Windows Server 2012R2 with Hyper-V
Virtual machines:
DC=Windows Server 2012R2 (DNS.DHCP)
Windows Server 2012R2 = AD FS, DFS, Symantec Endpoint Protection Manager with SQL 2014 (Symantec serves about 200 clients). The file sphere is not very actively used, only the exchange of office files, no more.
Plans to raise the WDS role for distributing Windows images.
Ubuntu 14.10 - Squid with AD integration (4GB of RAM and 4 virtual processors are already hard-coded for it)
Centos 7 - Zabbix (about 20 printers, 10 servers, 5 firewalls), Mostly simple ICMP checks. For Widnows and Linux servers, standard templates from Zabbix. For firewalls traffic monitoring and VPN.
Here's how to properly distribute the resources of the Proliant server so that virtual machines do not suffer from their (resources) lack?
Under DC, it’s not worth highlighting much, there are only the main roles of DNS and DHCP.
But under combat Windows Server 2012R2, how much to allocate so that SQL has time to process operations and the file ball does not slow down?

Answer 1

4х600 SAS is not enough for that infrastructure which you listed. 4x600 is definitely 10RAID and we get only 1.2Tb of space, of which immediately at least 80 GB goes to each Windows server. Although Squid can be deployed on 10Gb, you will most likely need reports and a good supply of logs in time and will have to build on the load on the server itself, because. 4 processors with 4Gb RAM is not correct, it is better to increase the RAM, reduce the CPU, because. it is better for a squid to keep a cache in memory for productivity.
Zabbix, here you also need to count from the number of requests, and based on the described 2CPU / 2Gb is enough, if you rely on the official documentation.
Do not forget to duplicate the DC on another hardware server, as recommended by Microsoft.
The planned WDS must be calculated, because in idle mode, it will simply eat up disk space, and during operation it will load the network and disks.
SQL is a separate topic, because You need to virtualize SQL wisely. ideally, it requires a separate disk subsystem. But you do not specify what size the bases will be, how many there will be, the approximate number of transactions, and so on.
For each of the servers / services you described, there are recommendations from manufacturers that you should familiarize yourself with, regardless of what they advise here or elsewhere, because. recommendations of manufacturers in the context of loads will help you formulate the requirements and the necessary attributes for the calculation yourself.