Hey thanks for your time. I am having issues finding metrics for cpu per concurrent users type of deal.
Http requests per second per cpu benchmarks and that type of thing. Nginx and apache are the most likely candidates.
I struggle to see the advantages of stuff like ecc, xeons and server racks. Since the websites I'll deploy are not meant to be a real business, rather they are fun python apps. It'd be cool to know the what's the capability of a ryzen 1600 vs a it's Intel equivalent.
I have an interest to build a ddr3 system since i could use its parts to debug other pcs. And keeping costs as low as possible. I'd only want to buy old corporate hardware if there is a clear objective advantage for it rather than using consumer grade parts.
I wouldn't want to spend more than 200 bucks for a cpu.
I have a project similar to yours for a gateway server with some web services on it and some target servers behind it. I am waiting for Ryzen 3 Pro. The 3.7ghz one would be a very snappy gateway with quad core that will overclock to close to 4ghz and have ECC support and SVM features. Then I can securely run my gateway in a VM and pass through some PCIE NICs. Even though it quad core it still supports the 64GB of ram I believe. For that matter though 32 GB is a Huge buffer for any gateway server.
One thing you should note Is what your upload bandwidth is. Just simple straight HTTP should not consume much probably 200kbps per page request. say you have 5mb upstream the max would be 25 users loading a page at one time.
I'd think a rasbery pi 2 or 3 would service maybe 10 page loads at the same time. I like low power systems like ASRock J4205 I'd estimate that can do maybe 50 page loads at a time.
Whatever use case you have you should think of how it would be used. Say if you have a class of 30 and you tell them all to load your site at the same time things might be sluggish for a sec. It is good to think of a user not using your website when they are looking at a page but only when they are clicking a link to load a page.
Thanks for pointing that out. Ideally, I'll have a google fiber connection. It would be cool if I host a couple hundred users at the same time.
What exactly are you using the passthrough for?
With the Fortinet Gateway OS in a VM it needs to have full control of the NICs. If I were to let the host system share the NICs with the VM then the FortiOS system looses track of the physical hardware and makes rebooting the VM a guaranteed loss of service. If you pass through say a quad port NIC then the FortiOS VM will have full control of the network devices and the host system will be unaware of their presence in the system. I'll leave the onboard Ethernet from the ryzen board as a remote management interface for the host system but all other networking will run through the VM with the NICs passed through to it. As for how many users could it support I'd say it depends on the services that are being passed through it. For example a VPN tunnel is a much more costly resource than a web redirect. But I'd be comfortable with 150 or more simultaneous VPN tunnels on that Ryzen 3 quad core. Even the smaller devices that Fortinet sells with RISC processors and flash storage can support 50 tunnels. But that Ryzen 3 Pro with NVME storage and 64GB Ram would be a beast router. FortiOS could actually run just fine from the Intel Optane NVME drive passed through as the boot drive for the VM that'd be about as low latency as you'd need. With the Optane drive you could get by with less ram as well.
Use a Raspberry Pi 3 Model B. They were designed with your use case in mind.