In this series of articles, we want to address the questions that people have when dealing with hosting providers and dedicated servers in particular. We conducted most of the discussions on English-language forums, trying to help users first of all with advice, and not self-promotion, giving the most detailed and impartial answer, because we have over 14 years of experience in the field, hundreds of successful implemented solutions and thousands of satisfied customers. Nevertheless, our answers should not be taken as the only correct answers of the first instance, they may well contain inaccuracies and even errors, no one is perfect. We will be grateful if you add or correct them in the comments.
Why is the cost of a server with a traffic limit of 100 TB and a 1 Gb / s channel much less than the cost of a server with a 1 Gb / s channel without traffic? After all, if you rent 2-3 servers with a 1 Gbps channel and a 100 TB limit, you can consume exactly the same amount as would be consumed by a server with 1 Gbps Unmetered, or even more channel in peaks, while the provider provides essentially more iron, more connections and a lower price?
The fact is that providers, when offering servers with a sufficiently large traffic limit or even "unlimited" for little money, take into account the average consumption profiles of their customers. It turned out that the majority of customers who buy such channels do not fully use the connectivity provided to them. It is because of this that it becomes possible to make such an offer.
100 TB of traffic is a pretty big limit. This is more than 100 Mbps Unmetered. After all, having a channel of 100 Mbit / s, without taking into account, you can pump a maximum of 100 (speed in megabits) * 86400 (number of seconds in a day) * 30 (days) / 8 (bits in bytes) / 1000 (megabytes in gigabytes, if we count by 1000, instead of 1024, 1024 is a bit in a kibibit) = 32 GB per month in each direction with a constant channel load of 400%. However, as we know, servers do not consume traffic constantly, and very often the daily consumption curves can look like this:
Some people may have peaks of bandwidth and require an honest 1 Gb / s at these moments. At the same time, the total traffic limit per month can hardly be exceeded:
Such clients, of course, are not very profitable for providers, and therefore the provider seeks to transfer them to Unmetered, since if it provides services to customers from the same region, it is likely that consumption peaks will coincide and this "honest" gigabit provider will be able to sell only 1,2 client. If the provider has clients from different regions, then it is likely that the channel can be sold to two or more subscribers at once, since audience consumption peaks will occur at different times. In reality, not every client consumes his own limit of 100 TB, therefore it is extremely profitable to provide servers with a limit of 100 TB of traffic.
Moreover, by connecting 10 gigabit channels to the racks, it is possible to very effectively divide the traffic between everyone. We manage to split a 10Gb/s channel into an average of 5 racks filled with servers with a 100TB limit. This is approximately 150 servers. Since one rack with a height of 47 units can accommodate either 41 one-unit servers or 21 two-unit servers.
As a result, the total consumption of the channel is as follows:
If you deny service to subscribers that generate a lot of traffic (the main contribution to the channel load is made by less than 10 servers out of 150 present on this port), then you can increase the number of servers to 300 or more. And everyone will be happy and everyone will have enough traffic.
However, there are other ways to save money and not upset subscribers - connect a cheaper transit uplink or send traffic to an exchange or peering point for free if you are a large traffic generator.
It is all this that allows us to provide low prices, not to refuse service to subscribers, paying transit providers 1500-6000 euros for every 10G, depending on how good the transit provider is and selling connectivity cheaper than prime cost with a certain oversell coefficient, when each of the subscribers has their own ordered an honest channel, while not interfering with each other.
It immediately becomes clear why the price of 1Gbps Unmetered is much higher, since if with 100 terabyte servers, not everyone consumes their limit, then the client who orders 1Gbps Unmetered will obviously consume most of the channel. Although we have seen the exception above and an example of how you can generate almost 1 Gbps of traffic in bursts and still be within the 100 terabyte limit, this is an exception, not a typical pattern.
My administrator installed the vnstatd program on the server, the traffic is taken from the interface, it is taken off every 5 minutes. Does he take everything into account? So, it shows me that 87 TB is used, while the provider says that 96 TB is used and the traffic is almost over. I am confident in my system administrator, he is an excellent specialist. And if he says that the provider overestimates the expense - it is. Moreover, this is evidenced by the fact that they began to play with might and main with values, giving in the process of discussion different values for traffic for the same period. To the question "how so?" we are still waiting for an answer.
The fact is that some traffic accounting programs keep records in TiB, not TB. Tebibytes, not terabytes. That is, accounting is carried out according to the binary system, and not decimal, based on the fact that in a kilobyte, or rather in a kibibyte, there are 1024 bytes, not 1000.
It is worth noting that in order to prevent this distinction from being used for marketing purposes, ISO (International Standardization Organization) has long introduced the prefix "bi" for binary bytes, that is, kibibytes, mebibytes, gibibytes, tebibytes. However, marketing still took place, and if drive manufacturers manage to indicate smaller amounts of drive capacity due to decimal bytes, then when measuring and accounting for traffic, the situation is reversed. The hosting provider, providing 100 TB of traffic, provides it less than it can actually be, if you take it into account in the binary system.
It would seem that the difference is small, only 24 bytes per 1000, the error from this is only 2,4%, but why is there such a big difference, at the level of 10%? Maybe they really did not take into account some kind of traffic?
The fact is that we must not forget that the “error” is growing, namely:
1024 bytes in a kibibyte (according to ISO standards), in a mebibyte it is already 1024 * 1024 = 1 bytes, in a gibibyte - 048 * 576 * 1024 = 1024, and in a tebibyte - 1024 * 1 * 073 * 741 = 824.
Unexpected turn? Yes?
When measuring traffic in terabytes, the difference between accounting units is exactly 10%!
Moreover, the difference in data taken from the port of the switch and from the port of the server can be caused by a DDOS attack, which does not reach the client, and can be eliminated at the “router” level, while still consuming traffic.
We should also not forget that sometimes the program does not take into account traffic on all ports, and some traffic may “escape” from monitoring.
And it also follows that when limited traffic is provided, the total incoming + outgoing is often taken into account, and if you have, say, a VPN service, the ratio is 1 to 1 and your customers can download a total of no more than 50 TB of traffic with a limit of 100.
To be continued ...
Some ads 🙂
Thank you for staying with us. Do you like our articles? Want to see more interesting content? Support us by placing an order or recommending to friends, , a unique analogue of entry-level servers, which was invented by us for you: (available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).
Dell R730xd 2 times cheaper in Equinix Tier IV data center in Amsterdam? Only here in the Netherlands! Dell R420 - 2x E5-2430 2.2Ghz 6C 128GB DDR3 2x960GB SSD 1Gbps 100TB - from $99! Read about
Source: habr.com