Right VM/VPS sizing for Shared Hosting

[hostyler]

Hello guys
We are evalutating CloudLinux, with license CloudLinuxOS Shared or CloudLinuxOS Shared Pro

But we are not sure that we have correctly understood the Cloudlinux method of managing CPU and memory limits.

Example
Hosting Plan with 1GB vRAM 1vCpu
For host 100 customer the sizing of the VM/VPS, excluding overselling and concurrent uses, should require 100GB of RAM?
And how many cores?

We understand that, as in virtualization, resource usage is rarely concurrent, but is there a guideline for right-sizing?

We are analyzing the hosting package offers of our main competitors, which offer with defined resources (cpu, memory).
We see hosting package from 1 to 4 GB vRam and from 1 to 3 vCpu, and certainly to maintain aggressive prices the number of customers per server/vps is not low.


Thanks for any suggest / information.

[bogdan.sh]

Hi,

Well.. it's always about overselling and yes, due to non-concurrent usage they are shared between customers.

The limits does not mean resources are allocated for customers, instead - they are limiting users of resources overuse. Basically it's the same as you current server can handle 100 customers without CloudLinux on, the only difference is that without CL one customer can allocate 80% of RAM. While with CloudLinux limits this user can allocate only selected amount of it, e.g. 1Gb and no more.

There is no silver bullet of calculation. Better to start with the default limits for all the customers AND some increased limits for high-end customers,e.g.:

Typical Hosting Account

• SPEED=100%
• PMEM=1GB
• VMEM=0
• IO=1024KB/s
• IOPS=1024
• NPROC=100
• EP=20

High End Hosting Account

• SPEED=200%
• PMEM=2GB
• VMEM=0
• IO=4096KB/s
• IOPS=1024
• NPROC=150
• EP=40

Then monitor usage with CloudLinux Manager and adjust if needed.

[hostyler]

Hi,

Well.. it's always about overselling and yes, due to non-concurrent usage they are shared between customers.

The limits does not mean resources are allocated for customers, instead - they are limiting users of resources overuse. Basically it's the same as you current server can handle 100 customers without CloudLinux on, the only difference is that without CL one customer can allocate 80% of RAM. While with CloudLinux limits this user can allocate only selected amount of it, e.g. 1Gb and no more.

There is no silver bullet of calculation. Better to start with the default limits for all the customers AND some increased limits for high-end customers,e.g.:


Then monitor usage with CloudLinux Manager and adjust if needed.

Thanks for informations

SPEED 200% = 2 core .. right?


Comparing various online offers I understand the overselling of memory... but making a relationship between the price of the plans and the number of estimated customers I don't understand how 2 or 3 dedicated vCPUs can be offered, without sending the server / VPS into critical situations.

[bogdan.sh]

Yes, 200% means 2 cores.

It's a marketing trick as vCPUs are not dedicated for a user. And even more, there are cases when VPS CPU are also shared between virtual machines (read about the steal time in top).

However, with CloudLinux onboard you will have more free CPU times since no users can jump over their SPEED limits.

[hostyler]

Yes, 200% means 2 cores.

It's a marketing trick as vCPUs are not dedicated for a user. And even more, there are cases when VPS CPU are also shared between virtual machines (read about the steal time in top).

However, with CloudLinux onboard you will have more free CPU times since no users can jump over their SPEED limits.

Thanks for help and info