New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Please contact us to discuss further @ support[at]readydedis.com
I'm glad you actually read the reviews :-) - most don't.
Whatever you end up doing, don't rush it. There are a lot of things that needs deliberate consideration in your situation.
Price increases can happen with any provider. My advice is to leave production as it is and look at your architecture and see where you can make changes and work on V2. You are saying the instances are idle 90% of the time, that doesn't sound very good. You say each instance requires a public IPv4, check if that is really the case or you could use something else.
Agreed, this seems like a super funky architecture to have so much compute provisioned but barely used, it doesn't make sense. Maybe FaaS could be helpful here but it's difficult to say without actually knowing the use case.
Don't cut costs by going for lower end and potentially less reliable compute providers. Cut costs by optimising your usage of your already provisioned compute.
in Saas solutions, may be their clients use their panel/website only few times a day so most of the times, resources may be idle but when a client use panel...it should be live and running.
if i were you,
1. i do it on a small scale say 10-20 client panels first then move all other
2. i do it on a high resource available dedi and splitting it into multiple vps and i can over-sell internally as i am aware, most of the times resources don't get used much, Also monitor load and adjust hardware or move accounts accordingly if there are few noisy neighbors in a dedi who are using resources at same.
3. hire a good guy or take help of dedi providers to install proxmox, asking for suggestions on hardware, live transferring accounts betweens dedis etc.
but over the period, i am sure you can save much more money.
actually i am surprised why no one mentioned @Francisco buyvm, he has capacity to host vps at this scale and you can adjust plans/resources too.
Also, do you actually need IPv4, or is it nice to have?
Assuming your services are all HTTP/HTTPS on standard ports, you can use IPv6 with Cloudflare (who have a great API) and that would immediately save you 1,500 * 0.6 = €900.
You can then run your own internal network for management/backups, etc...
As far as I'm aware, there are no domain limits on CF free.
Just open a ticket and let them know you're about to drop 1,500 domains on them 😅
I'd elso echo what the others are saying, look more into optimising your stack - if you can.
I'm not a VPS reseller or anything. I'm running a SaaS business and there are specific applications that are running on those servers. Each have their own ports and they're not web apps or anything so no HTTP/HTTPS. We're not giving our customers access to these servers. It's a SaaS business so everything's managed for them. Think about Game server hosting. Not exactly the same but a similar concept with a few differences. Although in Game server hosting you still give your customers some sort of access through a game control panel. We don't give that either. Our customers order our services and they never even log into their accounts or interact with their services in any way. They don't even think about it. We run these services for them and they just pay their fees that we're charging them for our services.
IPv4 is a must have because otherwise these apps won't work. They only support IPv4. Some of them actually support IPv6 as well but it's not very stable so we don't use it.
If we can find a solution here, either building our own or partnering up with a provider here, we will build our own migration tool to migrate all of our applications to the new place. Everything will be fully automated just as they're now with Hetzner's Cloud API.
Ah i gotcha!
Well, I hope you find a long-term solution
As we don't have more information, it is hard to judge, but the fact, that each VM uses ~1.5GB RAM suggests that a lot of information is the same on all the VMs. Can you deduplicate the information? What makes the CPU usage spikes? Do they all happen on all VMs at the same time, or are your clients triggering that? If it wluld be possible to merge multiple clients on single VM, lets say deduplicating 60% of heap, you could have at least 10-15 clients on one 8GB VM. Then as mentioned by others, there must be a way to solve the IPv4 problem - use IPv6, make use of different ports, if its http(s) based, proxying by hostname (this could be subdomains, or FQDNs, doesn't matter). This way you could easily save at least 50% on your wallet. If that works, nothing stops you from aggregating even more and going to 32-64GB VM, or even a dedi...
Basically, it feels like the problem of software architecture which you are trying to solve with infrastructure.
As for IPv6 have you tried integrating your service into Cloudflare? Then you may not need IPv4.
Also sounds to me like LXC/OpenVZ would be a better option if you go down the route of using your own hardware.
Each IP address has 65535 ports.
You can put everyone on different ports of the same IP address.
Both IP address and port number are to be published in DNS as SRV record.
I have all sorts of questions about your architecture with a footprint several small resourced VMs like that.
Is this a clustered platform? Is this a multitenant platform where segmentation is VM?
Is the storage you are using in each host for ephemeral tasks or is it intended to be persistent.. and if so how have you been backing the storage up?
Is your scaling because of traffic spikes or resource contention / noisy neighbor problem from shared hosting.
If your platform is container-ready this might be a good candidate for baremetal k8s. Your base workload could then be augmented with some k8s VM worker nodes as needed for additional scaling.
Thanks everyone for all the replies. I've made a deal with a local provider for $2 per 2GB RAM and 30GB SSD including IPv4. Using LXC. Also hourly billing which is the perfect solution for our use case.