Question Using Plesk for a 150k+ hosting provider on AWS

[MEXXUS]

Hi All,

As a hosting provider, currently hosting around 150k+ websites on a self-built platform and hosted on AWS, we're looking into the possibility to use Plesk as a future platform. However, because of the volume, this seems to be a challenge. Plesk documentation says that i.e. a maximum of 900 sites per server are allowed, so that means 150+ servers. Of course, we would like to have minimal risk for SPOF and use a separate DB cluster to connect to Plesk. Also, we would like to host e-mail via Plesk as well, where we currently have over 400k+ e-mail boxes hosted on an external e-mail provider.

So far, I have found the preferred requirements for Plesk as follows:

• 1GB of RAM per 40 – 50 websites
• 1/2 amount of RAM recommended free disk space for swapping
• Between 2 and 2.5 GB per website
• Plesk doesn't say anything about CPU cores
• Without advanced configuration: 900 sites per serve
• With advanced configuration: 1000+ sites per server (altering kernel needed?)

Anyone here with knowledge of really large hosting environments to host on Plesk?

 

[pleskpanel]

The underlying disk subsystem is going to very much play into this. Running databases on the same instances (making assumptions about IO) is certainly achievable and can simplify the setup but mail at scale can create headaches. If anything, consider keeping your mail on its own clusters as those mailboxes will not only grow but also add to IO contention with those sites. It can be easier to address mail reputation issues when isolation it to its own stack. SmarterMail or other services might be worth looking at.

 

[weltonw]

Your "preferred" requirements will depend on your workload. If there are WordPress sites, I'd recommend against 1 GB / 40-50 for optimal performance.

While yes, you can host a ton of sites on Plesk, (Increasing the Number of Domains that Plesk Can Serve) 900 is a lot, and unless they're all very very light and aren't saturating any resources, that's already a decent amount / server. You have diminishing returns for things like QC/s as you add more sites, not to mention failure, etc.

I'd question if a cluster is necessary. The actual Plesk server remains a SPOF unless you have replication (which is quite cumbersome). Maybe you get improved resiliency/easier failover, but maintaining multiple clusters of databases has a greater overhead than one uniform local install/server

 

[FutureX]

@MEXXUS Just curious if you found a solution to this?

@weltonw For WordPress sites what would you recommend in terms of RAM/site? What do you consider a "decent amount / server" in terms of hardware specs?

 

[weltonw]

@weltonw For WordPress sites what would you recommend in terms of RAM/site? What do you consider a "decent amount / server" in terms of hardware specs?

How many/what kind of sites? It's highly dependent.

 

[FutureX]

Let's say 900 average WordPress sites. I know there's no "average"