Do any longterm Developers administering their own Apps on the major Virtual Machine hosting services notice this problem while running your app on a multi-tenant “scalable” cloud provider?
Noisy neighbors: If another tenant in a VM server runs a resource-heavy app, whatever you are hosting on the server will see a performance drop. The noisy neighbor factor can lead to a 20 to 30% performance loss for all VM users by causing resource bottlenecks.
I haven’t luckily noticed this yet, but if I understand you correctly you’re referring to using shared machines. Correct?
We’re using DigitalOcean and using a ‘shared cpu’ for our droplet.
However, if I start noticing performance drops/ instability because of that I’ll quickly upgrade to a dedicated CPU. Starts at $42/month I see.
Noisy neighbors will be a problem on pretty much any shared hosting server, as a lot of the $2/m VPS providers will cram as many machines as possible onto the host node since their margins are so low.
The best thing you can really do about it is going with a reputable company that monitors their servers for overloading and doesn’t oversell. Digital Ocean / Vultr are pretty good at that, and OVH is also quite good in my experience. If it’s a host that isn’t overselling, you can normally just drop them a ticket about the performance, and they’ll check the node for you, since it’s probably someone abusing (crypto mining usually). Pretty much all VPS provides that offer shared cores have some sort of TOS about not using 100% CPU for more than X amount of minutes.
If the provider has set the load balances correctly then each VM should be scaled individually so no single instance should be able to utilise power to the level where it effects other instances. If it can then the vm design of the provider is flawed.
As I see it, it is pretty easy for a hosting provider to push the “limit” of containerization. Unless you are paying for a bare metal VM setup the whole idea of VM is to SHARE resources in the first place. Depending on the Hypervisor setup and internal company standards “scaling up” may not execute across less-loaded containers and/or Virtual Machines the way Marketing promises.
I guess that’s why there is such a price variation between providers.
Certainly my provider in UK use leading edge technologies and on demand scaling and load balancing even on it’s shared platform. All sites are monitored in real time and any trying to hogg resources are flagged, the owner notified and where appropriate the site suspended until the developer addresses the issue. Having said that, for node servers I do use a VPS. Including a 30 site cPanel licence this works out about £5 (GBP) per site per month.
I have had a few customers who have tried a save a few pounds by going to who appear to be the real budget providers (godaddy, namecheap etc) who have had huge issues and minimal savings once the hidden costs kick in.
I’ve been using Linode.com for slightly over a year now with main host in Frankfurt.
3 sites hosted including a Metabase instance for data analysis & the other 2 are variants of Mastadon – one operating as the default Mastadon community at large interlinking cooperating Mastadon hosts and the other one as a modded private community.
At times I have found that a site just simply did not Spin back up from their virtual existence. Linode decided to back-burner the application in deference to other users, I’m guessing.
The Virtuality of the whole experience bothers me. That there are no files, nada, of my sites when I FTP in , or make a Terminal request, until it jogs the Linode host of my Volumes and Containers server to take my request seriously and spin up my darned files and folders!
Have you dropped them a ticket? Linode is well respected, and generally production worthy.
They have gotten quite a few tickets from me.
The most difficult time I have with them is how many issues they do not support.
It’ll depend on what you contact them for. No unmanaged provider (do/linode/vultr, etc) will offer support for anything on your VPS since they are all unmanaged services, they only things they are really responsible for is the uptime and stability of the host machine. Digital ocean will usually point you in the right direction if they have a KB article for it, though, since they have hundreds and hundreds of them. If you need general support, you’d probably have to go with a managed provider instead.
Whenever I get a new server, the first thing I’ll do is run yabs.sh (https://github.com/masonr/yet-another-bench-script) - This will give you some great benchmarks, and you can check if anything is lower than it should be. If something’s particularly slow on this, you can pinpoint the issue and send the output for that test to your VPS provider support.
Following your github.com benchmark script link I found these HOSTING forums
which have lots of valuable feedback on Hosting related issues in a bunch of different categories.