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
Our network usage varies a lot during the day.
Rarely there is a need to ask someone to slow down, below 500 mbps it is acceptable, but when someone goes at 800 mbps that is dangerously close to node's limit and others start to suffer and since we give out a lot of traffic with the plans, it can take a long time before that is exceeded.
However, this seems like a discussion about physical port and real commit.
My personal coloed server has 1 gbps port and the dc also offered 1 gbps port but i cannot max that out, even at offpeak times, it is obviously shared with other ppl.
So, in our case, lets say i have a node with 1 gbps port but my network from dc is 100 mbps dedicated flat, then my physical port speed is irelevant, internet link will be 100 mbps, no matter what.
Also, if the dc offers 1gbps shared, i will never be able to max it out, except in a theoretical situation when nobody else does any traffic. Otherwise nobody will be able to reach 1 gbps (presuming fair share, no special QoS rules).
So, while it can happen to reach close to 1 gbps, the actual speed depends on a lot of factors:
1. Network load on the node;
2. Network load on provider;
3. Network load in DC;
4. Number of hops;
5. Quality of carrier (some oversell and dump faster that others that guarantee more point to point);
From here we go in reverse and it depends on the other provider like this:
3a. Network load in DC;
2a. Network load on provider;
1a. Network load on the node.
5a and 4a were already considered.
There is no such thing as guaranteed network all around the world. Nobody can guarantee outside of their network, even same carrier has fluctuations between routers, some routes collapse and the others might need to take the extra load, at times a new virus starts packet storms across the internet and blow out the calculations done for a "normal" load.
The more stable the link is, the better, but nobody can guarantee anything unless in their own network, and even then...
1GBPS/Gbit is 100mb/s
100Mbit = 12.5 MB/s
1Gbit = 128 MB/s
That's not too bad
It depends, I got more than 10MB/s when scp files between some VPSes, but for some location, this can be as low as x00kB/s. Any way, enough for me.
I thought it was 125MB/sec
I thought it was 125MB/sec
Fix it......
Its 128 MB=>
1Gb= 1024Mb
i.e. 1024Mb/8 =>
128MB
Nope, it's 128 MB/s
Google agrees too
1 Gbps usually means 1024 Mbits, not 1000Mbits
1000 Mbps = 125 megabytes / second
(Google^)
1Gbps = 1 Giga (billion) bits per second
1000000000 bps = 1000000 Kbps
1000000 Kbps = 1000 Mbps
1000 Mbps = 1 Gbps
1 Byte = 8 Bits
1000 / 8 = 125, thus 1Gbps = 125MBps
I'm running my box with RamNode @Nick_A
I got a nice and steady download speed at 10.5MB/s (Megabyte)
Upload speed to a dedicated server is 1MB/s+ (Megabyte)
The dedicated server is downloading the same file as my vps with: a nice 10MB/s (Megabyte)
1MB/s is good enough for me because I use it mainly for openvpn (watching netflix) and my network is only 4Mbit/s
Edit, downloaded file is from here: http://109.73.174.10/1gbfile.tgz
So basically the upload speed in the end is always asymmetric to the download speed (compared to what they advertise) IMHO. My host says I can reach 100Mbit both in upload and download. Truth is upload can't never go higher than 10-11MByte/s instead download can easily go over 50Megabyte/s!
11MB/s is really good for something that's advertised as 100m, not sure what you're complaining about.
No issues reaching good uploads from any of my hosts.
it is not what I'm saying
No.
While we do have differences between upload and download and they even reverse a few times during the day, it is mostly the same.
I guess some providers allow fewer things and mostly this means hosting and for hosting outgoing is much higher that incoming, therefore the speed will be skewed.
I cant think of another reason, unless they have some unmettered in one way from the DC.
It may be the coder in me, but you are wrong. Computers don't work that way because its just 1's and 0's. There are 1024 Bytes in a KiloByte (KB) and 1024 KB in a MegaByte. It's not a factor of 1000 its factor of 1024.
1 Gigabit is 1024 Megabits, 8 bits to a byte = 128MB/s (minus overhead of the transport protocols)
The only people who don't do that are storage companies, but they legally state that their values are in 1000's instead of 1024 the way computer system's represent it.
If they are advertising 100 megabit port speed, then your max speed for the entire port is around 12MB/s upload and 12MB/s download. The fact is that most servers don't download as much as they upload, so the upload portion will be used much more than the download, hence "asymmetry". But in no way should you be getting 50MB/s download speeds on a 100 megabit port.
http://en.wikipedia.org/wiki/Template:Quantities_of_bytes
2^10 = 1024 i know that.
?
You will see that a kilobyte is actually 1000 bytes, whereas a kibibyte is 1024. Read the table on the right-hand side carefully.
Up and Down = 11.3 MB/s max.
Please realize that it more has to do with routing than with the actual amount of available capacity. Just because I can offer you a capacity of up to 100Mbit or Gigabit, does not mean the routes to your location will support this amount of transfer. One of the things I test most often when purchasing any type of hosting is the routes they take to get to me. I for example have Comcast as my ISP at home and I know that they will 'peer' or route better using certain routes. Such routes might be Level3, NTT, Telia, Cogent, He.net, etc.
I know for me I will get the best routing from areas that route to hubs in Ashburn, VA as I know that my primary network access point for the rest of the US is through Ashburn,VA from my cable modem. This will vary for each person based on their ISP, locations and the networks/capacities for which their ISP chooses to use when routing their data traffic. This is why, for example, you see data centers using things like Internap (PhoenixNap) type of technologies where they peer with the larger ISP networks more closely to be able to provide a greater network speed to the home users isp.
When purchasing any type of hosting weather it be VPS/Dedicated/etc. I always do the following:
A. Perform a traceroute from the machine you will intend to do most of your downloading from (In windows you can go to command prompt and type: tracrt (ip address) and in *nux/BSD: traceroute (ip address)
This will allow you to see how many hops and what networks your hops are going through, and the latency of the hop. The more hops and the higher the latency of the hops, the lower your throughput will be usually. (Not always)
B. Take advantage of test download files, download a 100gb file to your self or the machine you intend to use most of your traffic to and see how fast it download and if you can sustain these speeds on multiple downloads.
Also note, not all providers use the same routes incoming as they do outgoing, and vise versa. So a network that may provide great routes in one direction for you, may have poor routes in the opposite.
C. (For the more advanced users) Use a tool like bgp.he.net (looking glass) and check what type of bgp peers the network has and what percentage of their peering goes through a network you are close to.
Using tools like a looking glass will give you an easy way to see what bgp peers and routes a specific provider is using. Also, you can see how many networks they are directly peered with. This is why you sometimes see conversations like "It seems you only use Cogent or He.net", if you were to be someone purchasing bandwidth you would know that Cogent and He.net are some of the cheapest, and sometimes most oversold routes you can purchase. However, they are also sometime the most available routes in some areas (or only routes available). Someone who would be using Level3, NTT, Telia or other such routes are likely paying more for their bandwidth, and will likely have some better routes to offer to some areas.
I know this is a long post, but I hope it helps people to make better decisions when they are purchasing VPS/Dedicated/Network services. Also, I am not saying I know everything or that I can not be faulted for making a mistake from time and again, so if there are any corrections to the above the community would like to offer, please do.
My 2 cents.
Cheers!
Well, a lot of people has a good point here.
The advice that i can give is this: try to find hosters with gigabit line AND low bandwidth on plans (like 100-200 GB/month) because most of the leachers (torrenting, vpn) are looking for so called "unlimited" monthly transfers and big disk spaces.
Although that transfer doesn't seem much, it is just enough for a site on a vps. If you need anything more than that, you should buy a dedicated. 100-200 gb monthly bandwidth can handle a huge amount of visitors with optimized sites.
Read what @joepie91 said. You're thinking of a different unit entirely. A gigabit is 1000Mbps. A gigabit is 125MB/s.
There so much bullsh!t when it comes to speeds in this industry it's sickening.
Problems are rather simple, the upstream providers (think colo or dedicated server owner) often might be engaging in traffic limiting and shaping. To them, a dedicated server isn't much different than an end VPS customer. You poor hanging chads.
The upstream provider might be routing traffic over an already pretty clogged HE or Cogent link (see this all the darn time).
Then there is the VPS itself. How to balance all sorts of nutty usages, sockets perpetually open, etc. Plus the throughput contention issues when people actually their services.
I've gone through dozens of providers, colocated gear in a number of locations, etc. Problems are just part of the industry. Unsure why the hungry lawsuit lawyers haven't come to beat folks down over the speed claims then when there, total rubbish network.
Things like axel are a hack. Sometimes you see an improvement, but far from always. If end test server is configured highly to also regulate throughput, those additional connections might be ignored or total speed capped.