Amazon EC2 Cloud Benchmarks vs. AMD Ryzen, Various AMD/Intel Systems

[gryning]

For fairness you should do some GCP comparisons. I find their single threaded computation far better than aws as well as the network throughout/latency.

I can't publish my data...

[aaronage]

Regarding whether or not AWS is cost effective... it's really missing the point.

I kinda surprised to see this debate - do none of you fine people work in a position that exposes you to "the cloud" in some capacity? It's pretty obvious why you'd want to deploy with AWS, Azure, GCE etc.

[starshipeleven]

If you are, e.g. small producer of mini aircrafts and spend up to €1M fo nice laser deposition 3D printer, and you need to do simulations ( liquid/gas flow, thermal, mechanic effects etc), would you really feel comfortable exposing your work to the friggin cloud ? A those prices, how much of a problem would be to spend €50-100k to a decent CPU/GPU cluster. ?

I can't comment on the $$$ part as I would need more info than that (did they pay the 3D printer with cash they had lying around or they asked money to a bank? What is their sale estimates, how high are their spikes of work, do they have decent staff to babysit the cluster?)

But I'd like to point out that what you are offloading to the VM is the number-crunching. An hypotetical evil guy looking into the VM may see that you are crunching stuff about some physics simulations, but there is no reason to have the VM contain enough info about what is that for or what you were trying to find with them.
Plus the fact that the VM exists only for a limited time, and is in a random physical server, makes any such hypotetical evil guy's job harder as he would need to pwn the whole Amazon's local cluster or something. Unless the evil guy is Amazon itself, of course. I'm sure that a tinfoil hat would help in that case.

Sure some can still be gathered with the help of some guesswork, but it's much less useful than you might think.
It's usually far easier to just send someone to steal far more useful data from the company's systems directly.

[starshipeleven]

Regarding whether or not AWS is cost effective... it's really missing the point.

I kinda surprised to see this debate - do none of you fine people work in a position that exposes you to "the cloud" in some capacity? It's pretty obvious why you'd want to deploy with AWS, Azure, GCE etc.

Explain plz.

[Brane215]

I can't comment on the $$$ part as I would need more info than that (did they pay the 3D printer with cash they had lying around or they asked money to a bank? What is their sale estimates, how high are their spikes of work, do they have decent staff to babysit the cluster?)

Plus the fact that the VM exists only for a limited time, and is in a random physical server, makes any such hypotetical evil guy's job harder as he would need to pwn the whole Amazon's local cluster or something. Unless the evil guy is Amazon itself, of course. I'm sure that a tinfoil hat would help in that case.

Why ? How many times has the coat of "national security" been used for private purposes ? You have this kind of misuse all over the place these days.

[starshipeleven]

Why ? How many times has the coat of "national security" been used for private purposes ? You have this kind of misuse all over the place these days.

Huh no I don't know. I guess that EU is better than USA on these things.

Anyway, the second part is still valid.

[aaronage]

Explain plz.

1) Running your own datacentre, or running hardware is a shared datacentre (colocation) is expensive (rent, power, maintenance, engineer time etc.).
2) Acquiring, maintaining and upgrading hardware is expensive - especially for small business/startups that don't have the funds to reach the scale they might need.
3) Having a presence in multiple regions is (you guessed it) expensive. If you're rolling your own, consider the challenges to deploying hardware in another country. Unless you're hiring an engineer that works remotely, you're probably going to need a support contract with some 3rd party that will handle things like outages (e.g. what happens if a disk fails in a datacentre 1000s of miles away? You're not going to buy a plane ticket for one of your local engineers).

The cloud solves all of the above. For very little cost, you could bring up instances in multiple regions that are fully provisioned with public IP addresses (or behind load balancers that expose ports on publicly accessible IPs), have fast and redundant storage and no engineer time is spent managing the physical infrastructure. If you're not managing physical infra, you probably don't need as many engineers.

That's just scraping the surface really, and only looking at some of the benefits to not managing the physical infra. When you start looking at the "infrastructure as code" benefits (automation, scaling, testing, monitoring etc.) it opens even more possibilities you (realistically) couldn't replicate by rolling your own.