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.
Speed testing
@VirtixTechnologies said: Download speed from
Why does everyone waste their time with "speed tests"? Such a small minority of this community actually understands it, and you will rarely hit full port speeds.
The only reason I even pay for gbit access for colo is to stop fighting with those that think they need it. Well and I need over 100mbit now.
Thanked by 1Asim
Comments
And on top of that some of these sites where this shell script does speed test from just suck. At least find some decent places to test from.
I agree. It's only too common for people to want more resources than they're realistically actually going to use. I'm a great believer of paying for what you need rather than the whole "unlimited" scene or what have you.
How much is the 1 gig port costing you?
I pay for traffic I consume per TB, not on the 95th
I agree, for normal sites it's not needed. However certain situations require some good boosting speed and for them it's useful.
Example?
@miTgiB backup/migration, streaming, file server and other things that no average user does.
Streaming is my only reason/excuse to care about port speed & speed tests.
Torrents work on any line, just leave a router on with a hdd attached over night and will finish dl till in the morning.
I did many low power torrent boxes here you don't need ultra-fat pipe for that.
Taz is right, also that not everyone needs that, but the ppl that do, are looking for it and the others say, well, why not if the price is the same...
M
It's useful for a number of reasons:
A few people have pointed out that we should support IPv6 in ours plus a few other things.
http://notwerk.com/ looks interesting, a massive visual ping at a network level (rather than node level).
@serverbear
Telecom Italia in Singapore to TPG, around the world in 500ms. Just uncovered this while doing research on AU's networks. I wonder if 1gbps is achievable.
Tracing the route to syd-nxg-men-crt1-ge-6-1-0.tpgi.com.au (203.29.135.41) 1 xe-11-1-0.franco31.fra.seabone.net (89.221.34.183) 196 msec 196 msec 192 msec 2 ix-3-3-0-0.tcore1.FR0-Frankfurt.as6453.net (195.219.180.37) [AS 6453] 248 msec 188 msec 192 msec 3 if-7-2.tcore1.FNM-Frankfurt.as6453.net (195.219.50.2) [AS 6453] [MPLS: Label 775714 Exp 0] 348 msec 348 msec if-9-2.tcore1.FNM-Frankfurt.as6453.net (195.219.50.42) [AS 6453] [MPLS: Label 775714 Exp 0] 364 msec 4 if-5-2.tcore1.AV2-Amsterdam.as6453.net (195.219.194.13) [AS 6453] [MPLS: Label 415731 Exp 0] 352 msec 352 msec 356 msec 5 if-2-2.tcore2.AV2-Amsterdam.as6453.net (195.219.194.6) [AS 6453] [MPLS: Label 316019 Exp 0] 360 msec 372 msec 376 msec 6 if-5-2.tcore2.L78-London.as6453.net (80.231.131.13) [AS 6453] [MPLS: Label 568389 Exp 0] 352 msec if-8-2.tcore2.L78-London.as6453.net (80.231.131.5) [AS 6453] [MPLS: Label 568389 Exp 0] 344 msec if-5-2.tcore2.L78-London.as6453.net (80.231.131.13) [AS 6453] [MPLS: Label 568389 Exp 0] 352 msec 7 if-20-2.tcore2.NYY-NewYork.as6453.net (216.6.99.13) [AS 6453] [MPLS: Label 617585 Exp 0] 352 msec 348 msec 380 msec 8 if-11-2.tcore1.NYY-NewYork.as6453.net (216.6.99.2) [AS 6453] [MPLS: Label 538641 Exp 0] 360 msec 356 msec 356 msec 9 if-1-2.tcore1.PDI-PaloAlto.as6453.net (66.198.127.5) [AS 6453] [MPLS: Label 307600 Exp 0] 352 msec 348 msec 344 msec 10 if-2-2.tcore2.PDI-PaloAlto.as6453.net (66.198.127.2) [AS 6453] 360 msec 348 msec 356 msec 11 Vlan3254.icore1.SQN-SanJose.as6453.net (66.198.144.6) [AS 6453] 348 msec 352 msec * 12 Vlan505.icore1.SQN-SanJose.as6453.net (209.58.116.54) [AS 6453] 348 msec 352 msec 348 msec 13 pdx1-ar4-xe-0-1-0-0.us.twtelecom.net (66.192.244.74) [AS 4323] 424 msec 368 msec 368 msec 14 syd-nxg-men-crt1-ge-5-0-0.tpgi.com.au (202.7.171.165) [AS 7545] 512 msec 512 msec 508 msecNone of your listed reasons are dependent on port speed or relevant. What is relevant is can I serve the data as fast as needed. Video streaming is about the only possible excuse I have seen given, but rarely is over 10mbit needed for quality video streaming.
Now I am not singling you out, but this is why I say a minority understand speed, here is the past 24h combined use of over 1000 VPS
Thats quite good for 1k VPS.
Here is export for the last 30 days from 1 of the switches on my side. Around 70 clients: http://i.imgur.com/fkdTB.png
@miTgiB If you're doing 10 MB/s from Cachefly & the next host is doing 50MB/s then that will impact a users decision (regardless of whether they need it or not).
Quick example, we have a 1Gbps connection at the moment, setting up a new server from scratch (with all the upgrades & packages + backups, to test how well we can recover from a backup) is normally complete in ~10 mins. We set up a new server which was limited to 100Mbps & the decrease in download speeds meant the packages + backups took 100 mins to download. Not ideal if you're trying to restore a site quickly.
And whilst we don't sustain those sorts of speeds for normal everyday use, for an extra $5/month it's nice to know we can if needed.
What I said in many occasions, especially about buyvm and the old guard of the anti-highbwapps coalition is that, after a careful selection and long time in business, you can select the customers. You boot the ones that use resources while also banning various apps and then can sell a huge amount to thousands, 10 times your hw and 1000 times your bw.
In reality, the average resource usage is higher than that, but the ppl that get the boot for using what was advertised under various pretexts are going to the ones that just start in the business, and make the life hard for Randy and the like.
I dont say all the booted ones were innocent, far from that, maybe half are real abusers, but this is the natural advantage of established hosts, their customers use very little compared to new hosts.
The ability to select your customers is a natural advantage that comes after natural selection worked. The winners share the spoils, nothing wrong with that.
My point is that you cant compare resource usage of new and old hosts.
M
Now why did you have to put your tin foil hat back on, you were doing so well too. It is a conspiracy that has no valid proof. I am showing my actual use for the past 24h, I do not kick any customers for valid use of their service. I won't lie I have never kicked a customer, but for clear TOS violations, and no I do not allow TOR, but not for bandwidth reasons, I support the goals of the TOR project, but the reality of the TOR project is rife with abuse. In the past year I have received 3 search warrants from the Department of Justice, all three were TOR related.
True, but I have only consumed more resources as time passes, but you are right, I cannot speak for the new host.
You invented the wheel too? I use cacti, which has been around a long long time, 2005 is as early as I can find.
I've hit the limit of 100mbit from really high traffic and mildly heavy sites. Depending on the intention of the site you can only optimize so far. I like to know how much I can burst total because a few thousand unique hits in less than hour on an image heavy page (where quality is important for images over thumbnail size, on the front page and not buried further in) and things start to cap.
Keep in mind I'm referring to my client getting tweeted by a major author, and this is only one example, but it is my one example
No, but I'll guess now that you say it, I helped or suggested using cacti?
I find it useful when installing/upgrading - being able to download from multiple sources all at once. Also useful for adding/rebuilding nodes in the database cluster that could be a few gb off from a backup.
It's also very useful as a Java Developer. Not sure if you've heard of maven, but a lot of software are built on it. Sync-ing a new maven repositories takes years here on my local PC as we have slow Internet connection. We run a continuous integration build server on a VPS and that needs to update the maven packages from maven repositories and get new ones as we add new components.
I can think of a lot more legit examples.
Still toying with your centreon template @miTgiB, I haven't got my switch setup using snmp yet.
I have been playing with observium.
It is normal to use more resources as time goes by, sites are getting heavier and ppl watch hd over vpn, I just wanted to say that ppl which really use a lot of resources are either abusers or some in the gray area, few really need those for legit purpose but are not the regular leb ppl.
Once a host is known to terminate abusers fast and afford to cherrypick customers, they will not apply, even for leb prices, as such, you will have much lower usage than a new host.
I would say a vps business can only be profitable for real from the second year. Some get better, some get worse, depending on local conditions.
From the second year, tho, the loyal customers are alredy recommending to legit friends for stability, mostly, and this is when you can start really to oversell. If you do it smartly, everything fine, but you cant do that really in the first year, abusers will impact all the nodes and legit ppl will move instead of recommending, so the only real choice is to have a reserve over what you sold. No way to oversell if you want to make sure you survive the first year.
M
Seven years ago when Hostigation was starting with VPS I used hardware that was just crap, most desktops of the day creamed the nodes, for OpenVZ. I follow your logic, and I don't disagree with it, but I know my nodes so well now and what I can put on them before people complain I don't think I can really relate to it.