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
What is the Bandwidth limit and also on average what BW can be expected ?
The people who actually need 100gbps can afford that
I mean, I dont have 100Gbit but I can do 25Gbit on a VPS with 100TB per mo if the price is right
This has absolutely nothing to do with hypervisor overhead but rather with the way virtio/tap networking works.
The rest of your message is also wrong. Yes, it’s unlikely virtio/tap will ever exceed 20-30 Gbit/s. That’s not a limitation of “the technology of our time” though and really just tied to the way things are currently implemented.
Regardless there’s countless other ways to go beyond the limitations:
SR/IOV (PCI passthrough of a single NIC port or the whole NIC, depends on IOMMU configuration)
SR/IOV with Virtual Function (VF), something supported by more expensive NICs like Mellanox ConnectX
openVswitch with DPDK and HW offload, using vhost-user instead of tap, or alternatively an XDP backend using vhost-user as well
Sure, none of the solutions above are likely implemented by your average LET provider. Nevertheless it’s pretty easy to push 100 Gbit/s through a VM with the above.
Great. Show us a VPS that costs up to $100/mo (traffic fees extra, I'll grant you that) that can and does push 100 Gb/s through a 100Gb/s line with a hypervisor of your choice!
You see I didn't say that it cannot be done in theory, but if you say that it can be done, then show it (actually working). Maybe the provider who thanked - and hence agrees with - you can provide such a VPS to you (I won't hold my breath though).
You can't utilize 100 GBits in a meaningful way. You are better of having multiple servers so you get some redundancy
For me this was a hypothetical question. I was curious to your responses. I have no direct use case in mind. But I get it, 100Gbit connection isn't realistic anytime soon 😁
Of course depends on what target you are trying to reach, how much compute, how is the routing and multiples of other factors.
But generally physically the usage of the nodes never exceeds 20gbit concurrent, which means you could theoretically easily use the full 40Gbit limit the vps is set on, but just to be safe i would only guarantee 20gbit always, which is more realistic to be able to be used on a vps.
Traffic Limit is a fair use policy based on the tier of vps you choose. Can be looked up on our Site.
only found 50 Gbps Unmetered
https://fdcservers.net/product/vps $699.99/month
Must be be port speed.
no way you get 100gbps unmetered bandwidth at that price.
There are entire datacenters who barely have 100gbps capacity
ZETServers is cheap but I'm pretty sure you can get 100G dedicated ports with T1s like GTT for ~$4k MRR in a lot of big DCs. It's not entirely unreasonable.
As for DCs with barely 100 Gbps of capacity, they must either be small DCs or primarily consist of a certain type of customer that doesn't require good connectivity.
i dont know how is it in usa but in eu some cities are interconnected with at best 10gbps lines. so its impossible to guarantee 100gbps unless you do your own fibers everywhere. no isps will reserve this capacity at this price range.
I want to know who is going to rent the first 100 Gbit VPS for $7 / month. God help us.
FDC
https://fdcservers.net/product/vps
who has 100gbps to test it?
Can you name some providers that have 100gbps on external, guaranteed, on a single fiber, not multiple (combined) From let. No ovh or hetzner.
I'd assume many providers do? It's not something extraordinary these days. With that being said, we just hooked up additional 100G to GNM-IX and 100G to CloudFlare in the past 10 days.
thats just a port.
does not guarantee you 100gbps inside another client from the ix peers.
is at best to isp router.
Most of the DCs that offer services on LET (such as FiberState, WebNX, ColoCrossing) have 100G ports with T1s. For VPS/dedicated server providers, it's generally more rare because LET consists of a lot of smaller providers and these types of products just don't take up enough bandwidth to justify 100G for many.
--
If you're asking for providers that offer 100G dedicated ports, yes, it is uncommon but I'm saying that it's not impossible. NOCIX (AaronW on LET) can give you up to 400G on a single port
If you are in basically any big or well-connected DC, you should be able to get 100G uplinks. I'm sure that 100G in the middle of Liechtenstein will cost significantly more than getting a 100G uplink in major network hubs such as Equinix Frankfurt or Amsterdam.
As I mentioned before, T1s like HE.net, GTT, Cogent, etc. will happily give you a 100G uplink in large DCs for around $4k/month.
would be tough to find a VPS provider that does it, lots of challenges to do that. but I'm sure someone's out there who can offer it.
Yeah, this is true. For example: From some AWS locations in the EU, max speeds are 1 Gbps at best, although being on 10 Gbps hardware.. Same reason probably.
You will want a pretty decent CPU. Even with virtio the overheads for PPS is still quite high. And thats before processing / application layer (e.g web server). I use the rule of thumb of 10Gbps per core generally.
Ive got a specialised load (local http server serving largely cached requests) at on a 4 AMD virtualised cores (workload limited to 2 cores) on a 40Gbps (limit) nic. It's heavily tuned (sendfile etc) and can hit about 20Gbps. Look at perf top for it and you will see its syscalls, nic drivers, gro etc. This is not a heavy application load, just serving static files (no compression, etc) to internal consumers.
I'd be surprised if a decent VPS could do more than 60Gbit without exceeding the cost of a good bare metal system that for the same cost (and same connection) could do 100Gbps. And it certainly wont be at LE pricing.
Its rare to see this sort of workload on VPSs too, generally dedicated clusters are used for heavy processing and serving for the performance that bare metal provides. The only reason my workload exists in a virtualised space is that its using excess resources on an otherwise paid for cluster, and that its not always at peak load. Its a gap filler.
100gbit of soccer streams.
Francisco