Ask HN: How much do you pay for Google Cloud?
3 comments
We are around 3-7k / month.
We do: - big nodes -> less k8s overhead https://cloud.google.com/kubernetes-engine/docs/concepts/clu... - https://github.com/kubecost -> grafana shows us what we spend - autoscaling -> default nodepool for stuff which has to run - keeping logs and especially metrics on gcloud low. Metrics are really expensive
Thats what we do professionally. For my startup, i use digital ocean as it is just so much cheaper and my trusty homeserver (core i7, 32gig ram, 4 hdds with zfs, 1tb nvm ssd) runs argocd + gitlab builds.
This way i have the much cheaper server at home for all load related things and as my server was build as low power machine, it only consumes 50watt if it is not doing anything.
With this setup i predict something like 100-250,- Euro per month infrastructure costs.
Of course buying that server would cost something like 500-700,- Euros but i had it running already and i'm using it for tons of other things. And the hardware inside is already 3 years old (core i7 2700k) but still much cheaper then using GKE nodes.
If i would not have the option to have a homeserver, i would probably configure my GKE cluster to connect with a very lightweight service mesh an additonal node from some other server provider like hetzner.de. They offer very current core i7 with 32gig/64gig for 50,- Euros per month.
It clearly depends what you have more: Money or Time. I personally have more time on my hand then Money so i don't mind really having a slightly more complex and more self managed setup. I also did not spend more then 1-2 weekends on getting it running. Nonetheless if you can bill your hours, thats probably not worth it.
We do: - big nodes -> less k8s overhead https://cloud.google.com/kubernetes-engine/docs/concepts/clu... - https://github.com/kubecost -> grafana shows us what we spend - autoscaling -> default nodepool for stuff which has to run - keeping logs and especially metrics on gcloud low. Metrics are really expensive
Thats what we do professionally. For my startup, i use digital ocean as it is just so much cheaper and my trusty homeserver (core i7, 32gig ram, 4 hdds with zfs, 1tb nvm ssd) runs argocd + gitlab builds.
This way i have the much cheaper server at home for all load related things and as my server was build as low power machine, it only consumes 50watt if it is not doing anything.
With this setup i predict something like 100-250,- Euro per month infrastructure costs.
Of course buying that server would cost something like 500-700,- Euros but i had it running already and i'm using it for tons of other things. And the hardware inside is already 3 years old (core i7 2700k) but still much cheaper then using GKE nodes.
If i would not have the option to have a homeserver, i would probably configure my GKE cluster to connect with a very lightweight service mesh an additonal node from some other server provider like hetzner.de. They offer very current core i7 with 32gig/64gig for 50,- Euros per month.
It clearly depends what you have more: Money or Time. I personally have more time on my hand then Money so i don't mind really having a slightly more complex and more self managed setup. I also did not spend more then 1-2 weekends on getting it running. Nonetheless if you can bill your hours, thats probably not worth it.
Great insights, thanks for sharing. Recently I found myself plotting how much extra time I spend on maintaining and configuring, which is basically the money vs time problem. No conclusion reached yet on my side.
I pay ~$300 a month for a small application. gcp & K8s is great services but it is not good for your bottom line or startups.
Kurian is focusing on enterprise customers
Kurian is focusing on enterprise customers
Yeh I agree.
You need to look at the billing tab to see what's costing you.
I recently reduced a company's spend by 60%
- increase VM size in k8s, force cluster pool size down - turn off stackdriver logs (want to run EFK here instead)
Why do you need a cluster per project? The point of k8s is to run lots of applications in the same cluster to reduce costs.
FYI, your k8s clusters will start costing ~$72 / month in June (iirc)
I recently reduced a company's spend by 60%
- increase VM size in k8s, force cluster pool size down - turn off stackdriver logs (want to run EFK here instead)
Why do you need a cluster per project? The point of k8s is to run lots of applications in the same cluster to reduce costs.
FYI, your k8s clusters will start costing ~$72 / month in June (iirc)
Hey thanks for the tips. I think stackdriver is really eating up a lot, will probably scratch that.
These are completely unrelated projects so I'd like to avoid keeping them on the same cluster. I also created separate projects for them from the console.
+$70 per cluster wow. That's the type of news that feels like it should be next to some good news.
Do you know any way I could get my hands on Google cloud credits?
These are completely unrelated projects so I'd like to avoid keeping them on the same cluster. I also created separate projects for them from the console.
+$70 per cluster wow. That's the type of news that feels like it should be next to some good news.
Do you know any way I could get my hands on Google cloud credits?
As far as i know, k8s stackdriver logs are included.
We create a few hundred gb k8s logs per month and we also run grafana/prometheus but i would highly doubt that running grafana/prometheus/loki etc. yourself is cheaper then using stackdriver.
We create a few hundred gb k8s logs per month and we also run grafana/prometheus but i would highly doubt that running grafana/prometheus/loki etc. yourself is cheaper then using stackdriver.
It is, we have proof by billing statements
1TB of logs is $500
1TB of logs is $500
But boy oh boy, every time I check that billing tab it gives me a mini heart attack. I am scared to look anymore, the damn thing gives me anxiety.
I got two Kubernetes clusters with small instances for two of my side projects, and a VM for automation stuff.
One month the forecasted cost was over $800 ... I panicked, I cut as many corners as possible. This month I got an invoice for $250+ ... which is manageable I guess, but the forecasted amount for April is around $450... Come. On. Google, you're killing me here.
So, I guess what I'm trying to find out is how much do you guys pay for Google Cloud and how are you keeping your cost under control?