Traffic Limits and How They Work
We have traffic limits set for most of our seedboxes, and this is rolling 30 days. This rolling bit is sometimes a bit hard to understand. This rolling means we look at the past 30 days for the traffic, not the current or last calendar month, but just the last 30 days, and actually pretty much to the minute.
This means your traffic counted today, 26th of June, will include traffic done from 27th of May. Not from 1st of June. 2 Weeks ago on 12th of June, it would have included traffic starting from 13th of May. Sidenote: We currently look at the past 30days, however average month has 30.46days in a non-leap year. This gives you ~1.5% extra traffic per actual month.
Why do we look at the past 30days, and not just completely reset every 1st?
Good Question. Simple and short answer: Service quality and money. If all traffic limits are reset on the 1st of month, every user will start racing towards their limit on the 1st, causing the bandwidth consumption to balloon, and along with it resource demand. Both bandwidth and resource demand would be peaking super high on the first week of the month, to couple weeks of the month. This would mean we would need that much extra hardware, and bandwidth costs would be that much higher.
It is about being fair and just for all users
With rolling 30 days it is more fair and just. Bandwidth usage is more stable, there is no huge spikes in demand for resources etc. You “pay” for the bandwidth when you use it. Billing dates are not by calendar month, someone signing up say the 15th, but not renewing the next month (or purchasing new slot), could get double the traffic than those who keep renewing their accounts month to month. This double traffic will cause double the load on the server, and would be unjust and unfair for the other users of that particular server. Further it would be unfair to all the other users who would need to pay for these people’s double usage on top of their own usage as well.
Why not unmetered/unlimited?
Reason we avoid offering unmetered/unlimited traffic plans: It always winds up so that 95% of users are paying extra for the usage of 5% of top heaviest users. It skews the pricing, and would make managing service quality that much harder, as some people might be using 20 times as much as the average, and there is no traffic limit stopping you. So 95% of users would pay a premium to subsidize the usage of the top 5% most consuming users. Not very fair for the 95% of users.
Further, service quality assurance is much harder. Think about restaurants, all you can eat buffet compared to restaurants where you pay by the serving. In all you can eat buffet people tend to generally try to eat as much as they can, but price still needs to be low. This means the buffet needs to adjust their menu accordingly. Same goes for metered / unmetered hosting services as well.
For unmetered it is very hard to get high standards of quality and you will need to sacrifice a lot to accomodate those top 5% users, and because you offer unmetered/unlimited that 5% suddenly becomes 10-20%. Often this comes at the expense of service quality – prices just can’t be increased accordingly.
Where as metered services have more predictable usage levels per server, and there cannot be users which alone consume 100% of the bandwidth of the server on the lowest plan. Yes, that can happen too, and when that happens, you either need to eat that cost and leave that user as the sole user of the server OR you will just need to provision other users on the same server, knowing full well that their experience is going to be …. Not so good. In our opinion, neither of these cases should ever happen and we rather take the metered service approach. In our opinion, an user should always get what they paid for, or more.
Bandwidth is actually super expensive
Bandwidth is super expensive, the market norm per Mbps is right now probably around 1€ to 1.5€, and on the bulk side a 10Gbps from Hurricane Electric is somewhere around 3500 to 5000€ a month on multi year contract. I think 1 year contract is close to 5000€ a month, meaning 60 000€ total cost.
Even if your bandwidth is at the low side of things, say 3000€ a month for 10Gbps, that is 300€/1Gbps or 30€/100Mbps. Or roughly 1€ per each terabyte of data traffic done … If it were all flat! Unfortunately, bandwidth needs to be purchased by peaks. In seedboxes your bandwidth can fluctuate by as much as 40% commonly, so if you have 10Gbps peaks, lows are 6Gbps and average probably somewhere around 7Gbps. Peaks are usually short bursts, but you need bandwidth to accomodate even those short bursts, or otherwise latency jitter and packet loss creeps in! Hence that 10Gbps worth of datatraffic suddenly costs not 3000€ but 5000€ -> 10Gbps average, but 16.67Gbps peaks! But since bandwidth is usually bought in full link capacities at once, it is actually 6000€, but you have plenty of head room remaining. Now imagine if traffic limits were all at once reset on the 1st of the month … We could easily be talking then 60% or more in variance of bandwidth utilization!
This is not even accounting the network equipment you need to push those speeds … Which needs to be scaled also by the peak rate. Further, at the end of the day, customers need to pay for all of this, there is no free lunch. So if it were reset on the 1st of the month, your service price would need to increase accordingly. Not very fair, and not sensible to leave all that bandwidth unused for the end of the month. Hence, rolling past 30 days.
What is being metered, and what is being limited
We meter all the bandwidth an user does from his account out to the net. Whether it is torrent traffic, SFTP, BTSync etc. All of it gets counted in, at the kernel level. Nothing gets past this. What we don’t monitor is the ingress (download) traffic however. We only measure what goes out from your service to the internet.
Further, we only apply limits to torrent traffic. SFTP, BTSync, FTP etc. remains non-limited at all times. Even on services such as Super50 which is 100Mbps/50Mbps, you are on 1Gbps server and thus your BTSync and SFTP can actually go much faster than the 50Mbps advertised as the plan maximum upload capacity. EDIT: Upload limits are now applied to all upstream except datacenter internal traffic which is always without limits.