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
Everyone gets 1 Gbps unmetered port, in country in home only.
International Interneighborhood traffic is limited to 1 Mbps for IPv4 and 5 Mbps for IPv6.
I'll totally buy this one if it's priced proportionally.
Before the migration I had about 275 active VPS servers (different packages but all with 10TB/mo at 1Gbps then 100Mbps unlimited), without IPv6, and about 150Mbps IN/OUT constantly but that 1Gbps was almost never reached except in the happiest cases I had 800Mbps upload and 350Mbps download when I was running one YABS.
I currently have about 180 active servers.
At a guaranteed speed of 650Mbps, we can have a monthly traffic of 1,684PB only through IPv4 + 1,684PB only through IPv6 + ~5PB best-effort IPv6.
I hope I didn't make a mistake in the calculation, but if we ran 24/24 at the guaranteed speed, we would have 1600 packages with 10TB each. I intend to keep the limit of 10TB/month at maximum speed then limited because this protects me from illegal content hosted in our infrastructure. (especially streaming sites)
I intend to sell less than 500 KVM servers until I upgrade the network, but at any moment when I feel that things are going well, I would upgrade the network.
IPv4 speed is 650Mbps guaranteed at the rack level, but both providers have a best-effort of 1Gbps (even if one of them only in their network, but they have the most clients in Romania - including the old DC).
Yes, the fiber optic lines once installed in the location can be upgraded in 24 hours (today you sign the contract, tomorrow it comes into force) - tested when I upgraded from 150Mbps to 900Mbps.
The contract with the 2nd ISP is confidential, but for the first guaranteed 150Mbps I pay 66 euros/month in obligation for 2 years.
@FlorinMarian 330TB IN + 330TB OUT @ 1 Gbps.
Nah, pumping petabytes of data via GigE line. Compression algos and alike...
Are you going to blame ChatGPT for the miscalculations this time as well?
I suggest you delve deeper into this field of work instead of depending solely on ChatGPT.
Stats for a 4PB traffic:
I don't think you can reach 4PB of monthly traffic on a less than 1GbE residential line, sorry to point that out.
My fault this time, I assumed that I could be wrong.
Thanks!
The max one can get from the 1000Mbits line is 330TB. If there are 180 active servers, please explain how one will get guaranteed 650Mbits when the guaranteed bandwidth is 4Mbits if everyone started using the server up to his limit.
You and other’s have provided numbers
But no VPS client should expect a dedicated line
The one most interesting thing is: Exactly how many Mbps does his users use at once, let’s say with 95th percentile (not being able to max out your line once in a while in a shared envirement is standard) - even if 3 minutes a month 800 Mbps isn’t enough, that’s not extremely important for this budget level
So I think Florin should do a 95th percentile (or another more fitting number depending on how much he can allow users to get affected by congestion) calculation of speed requirements averaged out over a set number of clients historically for as long as possible, only then can he truly know - it’s very induvidual
Adding TBs & PBs into the discussion makes no sense, because usage isn’t linear, and @FlorinMarian don’t have a data cap, but, the linear part trumps that, because:
Not being able to utilize what you otherwise would want to utilise is perfectly fine, much better compared to Florin paying a lot for bandwidth that gets used a few minutes per month
And I don’t think TBs and PBs are a good measurement for the same reasons, as well as:
If users consinstantly use a lot more during the day for example, the capacity being avaliable at night dosen’t help if the speed sucks during the day; a total usage calculation dosen’t take this into account either
Tl;dr: @FlorinMarian has to over time look at averaged 95th usage and decide how often he can allow for users getting below xyz in speed. Well there, calculating users per Mbps is fairly simple
IMO the above has to be very carefully calculated, because doing it wrong would suck
Underestimating it means users getting shit speed and leaving
Overestimating it (or like @imgmoney suggested, calculating based on all users using their entire allocation at once) means spending huge sums of money for something that the vast majority of users wouldn’t ever notice apart from a YABS run at the worst possible timing, and thus, @FlorinMarian having to charge a lot more compared to the provided value
My point is just: What LET user would want their homehosted service to be $2 extra per GB RAM with the benefit that they’ll get their full network speed 24 hours a day instead of 23 hours and 30 minutes
Although what sucks for @FlorinMarian is that his usage (if I were to guess) is low enough to fluctuate a lot. Big companies (OVH for example) can probably estimate their future usage needs to a much higher certainty; avoiding overspending as well as too much congestion
average provider has at least 1G dedicated line per VPS node and 2x 10G per rack.
He has 650mbpa for whole rack… which is like nothing
People have more for home use
He has more than he claims his users have EVER used at once
Again, this is way too complex to just be guessed, he has data on his clients usage, he should carefully analyze it
Referencing other’s usage mean very little, he shouldn’t pay more because other people who aren’t his clients use more
OVH offers dedicated machines, with basically new hardware, and openly advertise 250 Mbps - and they seem to be widely used
The above obviously dosen’t mean that @FlorinMarian will be fine with 600 Mbps, but that’s my point
OVH offers 250mbps per server
He has 650mbps for WHOLE rack. That’s 6.5mbps per vps on 100 vps
Ok if you say so
But that’s speed of home internet in developed countries
why do you say that? I have lots of VPS where I do 100Mbits most of the time and run out of traffic quickly.
Also, he is allocating 10TB to every client which is 170 x 10 = 1700TB and 1.7PB of bandwidth which he never had and cannot provide.
Overall he can provide only 170TB via his network which means, 1TB of BW on VPS.
I didn’t say so, @FlorinMarian did when he replied to my question
If you never managed to eat more than 2 steaks, you don’t have to order 5 steaks just because the neighbouring table does
Your usage isn’t the average, what you do isn’t remotely important when calculating an average, and I’m 100 % certain that the average person here on LET that buys a dedicated, again, on average, are using more resources than someone buying a VPS.
Correct
Wrong, he should pay for bandwidth according to his actual highest possible (but realistic) usage, not according to his end-user allocated bandwidth.
Some random stats I made up:
According to you, should he do what basically no other LET host does and make sure to have the ability to push 10 000 TB? He wouldn’t have LET pricing, that’s for sure.
We’re talking probably at least 5-10x the pricing of the typical LET host.
When allocating a vCPU, even when a host say that the user can use the whole core if they’d like, they’re generally not dedicated unless advertised (@labze can tell you all about this, never once do all users use 100 % of their allocation).
Bandwidth is no different.
Finally, with your logic, a host offering unmethered/unlimited bandwidth (many hosts here does), they’d have to have unlimited transfer speed with their upstreams. No upstream have unlimited speeds, so they must all be fraudulent.
The entire game of a hosting company is finding out how heavily you can oversell without customers being impacted. I guess Florin just wants to find out what that number is.
I just discovered a new risk owning a DC at home.
https://hazi.ro/preview/HDD.mp4
what a lovely sound
Hello!
After a series of problems solved, I come with a small update.
I discovered a number of problems including:
- 30 out of 39 SSDs seemed to be non-functional (resolved - for some unknown reason they were encrypted at the disk level and any data changes were lost when you cut the power to the SSDs)
What happened in the meantime:
It follows that in the next period we will migrate in real time all the VMs from the current SSD server (over 100 VMs) in order to make some optimizations to it as well, after which we will include it in the cluster.
We are still waiting for the operator to come with the 3rd internet line to be able to create some special packages to really celebrate the migration.
I wish you all a wonderful day!
A few moments later...
Regarding the RAM, you could be right (I have no idea if all 5 are from the lot of 12 DIMMs without electrostatic protection) but in the case of the network PCI-E it was already mounted in the server, there is no problem that it is the same cause.
Cool. So what are the plans and pricing when you go live ?
You do know that your body can contain enough static electricity to damage computer parts, right?
If you open a server without being grounded and in an antistatic environment, you can damage the server parts by touching them.
€8/month for 4GB plan, same price as Starbucks cold brew.
That’s not BaseEnd range
You get dedicated resources.
Bro won't sell 72 VPSes with "1 dedicated core" on a physical server with only 40 logical cores.
On the contrary, I want bro too to sell such things so that we can get those for cheap. Price is a crucial factor that would make me buy such servers because there are minimal expectations regarding performance/uptime/reliability etc.
When you buy BroVPS, you can have maximum expectations regarding performance and uptime and reliability.