4 ms·

A nice big Threadripper in the corner of your office is all you need, I can almost certainly guarantee it. There are very very few businesses with less than 100

A nice big Threadripper in the corner of your office is all you need, I can almost certainly guarantee it. There are very very few businesses with less than 1000 employees that can't be run on one modern high end server.

leesalminen 4y ago
What happens when the power goes out? Internet goes out? Someone spills a cup of coffee on the server? Landlord needs to do some work and drywall dust flies around everywhere? It's not that most CRUD apps could run on a single machine, it's that they probably shouldn't. Customers want redundancy, automated failover, etc. Datacenters are much better at offering these things than your office park.
- re-thc 4y ago
  Interestingly enough the chances of that happening are actually less than an AWS region outage :).
- sourcecodeplz 4y ago
  You invest in 1-2 UPSes and a 4/5G mobile modem data plan that kicks in when needed. If you really want to be sure, get a diesel generator also. Will take some time to setup but it is worth it. - No one can see your code - No one can shut you down
- CyberDildonics 4y ago
  Someone spills a cup of coffee on the server? Are people spilling coffee on your servers at work?
- steponlego 4y ago
  You can do automated failover to Raspberry Pi, it'll hold while the main server is cared for. If you go cloud, you deserve what you get. And what you get is what you get, because it's not your computer, and it's not your data.
- margorczynski 4y ago
  Using a colo where each instance has a backup in another colo would probably be sufficient for most.
_cenw 4y ago
This always assumes these applications are built by competent people who value efficiency. The job market is full of people that think they need high scalability but then underestimate the cost of an HTTP request or the memory footprint of keeping an entire dataset in memory. It has happened several times in my career building platforms that I have to double check that packet capture I just took from that single Kubernetes node is accurate (my favorite is 22k DNS queries, all A IN s3.amazonaws.com, in a 0.2s span). Similar things happen with memory and CPU.
cantSpellSober 4y ago
You lose the benefit of a CDN though (caching content on a server closer to where each individual user is accessing it)

leesalminen 4y ago

What happens when the power goes out? Internet goes out? Someone spills a cup of coffee on the server? Landlord needs to do some work and drywall dust flies around everywhere? It's not that most CRUD apps could run on a single machine, it's that they probably shouldn't. Customers want redundancy, automated failover, etc. Datacenters are much better at offering these things than your office park.

re-thc 4y ago
Interestingly enough the chances of that happening are actually less than an AWS region outage :).
sourcecodeplz 4y ago
You invest in 1-2 UPSes and a 4/5G mobile modem data plan that kicks in when needed. If you really want to be sure, get a diesel generator also. Will take some time to setup but it is worth it. - No one can see your code - No one can shut you down
CyberDildonics 4y ago
Someone spills a cup of coffee on the server? Are people spilling coffee on your servers at work?
steponlego 4y ago
You can do automated failover to Raspberry Pi, it'll hold while the main server is cared for. If you go cloud, you deserve what you get. And what you get is what you get, because it's not your computer, and it's not your data.
margorczynski 4y ago
Using a colo where each instance has a backup in another colo would probably be sufficient for most.

_cenw 4y ago

This always assumes these applications are built by competent people who value efficiency. The job market is full of people that think they need high scalability but then underestimate the cost of an HTTP request or the memory footprint of keeping an entire dataset in memory. It has happened several times in my career building platforms that I have to double check that packet capture I just took from that single Kubernetes node is accurate (my favorite is 22k DNS queries, all A IN s3.amazonaws.com, in a 0.2s span). Similar things happen with memory and CPU.

cantSpellSober 4y ago

You lose the benefit of a CDN though (caching content on a server closer to where each individual user is accessing it)