SinEmpire, I have to tell you the truth, your post was very well written. What you say is very true and accurate. This was the best written post on this topic i have seen. :thumbsup

Good posts,

Some very good points mentioned here. The well established hosts will always thrive to achieve the best quality network and support for their customers. By offering real redundancy, real multihomed network, real 24 hour support. I really don?t feel there is a big problem with using cogent bandwidth. If price is important factor for you then cogent will be your best bet. That?s why some of us drive saturn?s and others roll around in brand new cl55?s. Unfortunately some hosts false advertise their services, promising things they don?t even have. Some even resort to comparing their host to others by using one factor, which is price. When sopping for a real host remember you get what you pay for.

Very good points indeed!

Thats why I colo all of my boxes at GHETTO-COLO. They use Sprint :thumbsup

I'm really glad I did this post. It's nice to finally see some positive feedback on quality hosting!! It's reassuring after seeing all the bashing that goes on here on GFY with hosts that charge reasonable rates :)

What I find funny is that there are a LOT of people that haven't posted in here!! lol lol They seem to have a lot to say all the time... but not in this thread! hehe

Brad

Technically your pretty close when presenting those numbers but there are more factors involved. Your assuming that every host is only hosting. For many of the smaller hosts they generate money by running their own sites. For example: using 10/Mbps of Cogent to generate $7,000 of revenue. 10/Mbps costing you $300 on a 100/mbps line. That's a very nice profit margin. Or take a 6/mbps site generating $20,000 in revenue a month, even better. It really comes down to the hosts business plan, if they know what they are doing they will be fine with or without Cogent. And $3800 gross profit is A LOT of money to a 1 man host. So there's nothing wrong with $3,800 gross profit.

For any "new" host to think they can sustain themselves on Hosting alone without a big bankroll is not going to make it. They have to have other revenue streams besides hosting to pay for the new hardware. Or a nice Credit Limit with Dell Business.

I also think you can't deny the fact that there is a market for Cogent bandwidth. NO mater how many people bash on Cogent, people are willing to pay for it. As an oportunists it only makes sense to sell what people want. If people want Cogent then I'm going to sell it to them [period]. The beauty of a router is if Cogent goes away tomorrow, we'll just reroute the traffic through Verio or whomever. So the threat of Cogent dissappearing tomorrow doesn't really exist to us. And if the host has a solid business plan, making a 500+/Mbps commitment to a Tier 1 provider won't be a problem. With that kind of commitment your cost per meg won't hurt the profit margins too badly.

So until Cogent goes away or gets bought out, enjoy the profit margins. Thats my philosophy.

p.s. Denis, you better have the AMG package on that bad boy.

I've read that post three times now..... it rules....

I vote it for Post of the Year. :thumbsup

I am a LAN/WAN guy by day...so I don't get much into internet hosting for the adult business.

However, I am paid to design VPN systems using Citrix, IP enabled frame relay and other similiar technologies. I can vouch for the WAN part of Brads post. I see Verio and shudder because I know the problems that are going to occur. It is a rare week when they don't have a router problem...that is the usual excuse given. Tier one or not...they are typically a problem and we move the client to a backbone that is more stable and reliable.

As for the hardware argument, I love to see Cisco stuff all the way through the system...you start hearing alot about Nortel or Linksys...or any other type of non Cisco router...well, you are in for some long days.

On the other hand, you guys with these giant dual processor boxes are using a cannon to kill a fly....typically your computer is going to cache the website first thing and deliver from that cache until something changes. A webserver is basically a file and print server with some bells and whistles. RAM is a great thing to have but let me ask you..how much processing does a webserver do...a little parsing and that is it. The exception would be if you are running applications over the web...but those utilities are typically relegated to VPN systems.

Right on Brad, tho...I just changed hosts and drilled the guy for an hour about hardware, connectivity and throughput ...and the origins of all.

Things are not always what they seem...

Here's the beginning of the problem, back when Cogent started announcing their super low pricing, everyone was wondering how the fuck could they offer bandwidth so fucking cheap. Yeah, we knew they'd do a large portion via peering, but how would they get peering with the tier 1 providers? Where was their infrastructure, etc.

To add even more to the confusion, their web server was hosted off an ExecPC DSL connection in Seattle at the time, and their two nameservers were gatekeeper.cogentco.com (colocated machine at some facility in Oregon on a 10Mbps Ethernet link) and monet.titania.net (machine sitting off of a silly ass dedicated 56k dial-up connection owned by Joseph T. Klein, original person who was in charge of peering at Cogent).

Anyways, network finally rolls out, and they're trying to sell their asses off (took months before they started doing double digits of traffic in major cities) but, after each ISP is pondering how they're still going to pull this off, they start receiving peering applications in their mailboxes.

Ok, so they have the little guys covered, how are they going to get the big boys to peer with them? Around this time, UUnet released their peering policy, and at the time, Cogent was nowhere near of meeting those requirements. So what do you have now? Tier 1 and 2 network operators that are definitely sure that Cogent is going to fail due to their lack of getting their bandwidth cost down.

Tier 1 NetRail announces that they're filing for bankruptcy, Cogent sees them as an opportunity to achieve Tier 1 status, as does new startup, Aleron. Cogent and Aleron go into a bidding war, and Cogent comes out on top with NetRail, they also go into a bidding war over bankrupt Telia, and Aleron comes out on top in that one.

Anyways, these purchases made the news, but what didn't make the news was the layoffs. Before the purchase, Cogent laid off nearly half of its engineering staff which was made up of ex-Digex employees, bringing the total count down to 8-10, then after the Netrail acquisition, the entire engineering staff of NetRail was given the boot. Some see this as Cogent running a tight ship, or as Cogent tightening their wallet.

So now Cogent has all this peering, should be a decent network, right? Not when Netrail's original peering connections were established at DS-3 (45Mbps) and OC-3 (155Mbps) speeds, all it takes is one or two Cogent customers pushing line-rate to flood that pipe, and most of the pipes were full, i.e. their only connection to the Level3 network was out of San Francisco, and on traceroutes, you'll see this full pipe most of the time during the day, example, today:

12 gigabitethernet4-1.core1.SanJose1.Level3.net (64.159.2.69) 184.200 ms 185.866 ms 184.009 ms
13 so-2-1-3.pr1.SanFrancisco1.CA.us.netrail.net (205.215.1.229) 365.807 ms 360.171 ms 362.528 ms
14 GigE2-1.tr1.SanFrancisco1.CA.us.netrail.net (205.215.12.1) 359.351 ms 361.219 ms 359.307 ms
15 p13-0.core02.sfo01.atlas.cogentco.com (66.28.28.57) 345.223 ms 353.359 ms 340.751 ms
16 p15-0.core01.sfo01.atlas.cogentco.com (66.28.4.69) 485.132 ms 428.902 ms 488.983 ms

8 nr1-p350.paix.ip.att.net (12.123.221.2) [AS 7018] 72 msec 72 msec 68 msec
9 205.215.1.169 [AS 4006] 72 msec 200 msec 204 msec
10 gige2-1.tr1.sanfrancisco1.ca.us.netrail.net (205.215.12.1) [AS 4006] 200 msec 200 msec 200 msec
11 p13-0.core02.sfo01.atlas.cogentco.com (66.28.28.57) [AS 16631] 204 msec 208 msec 204 msec
12 p15-0.core01.sfo01.atlas.cogentco.com (66.28.4.69) [AS 16631] 200 msec 404 msec 204 msec

This brings up another issue, why won't Cogent establish an interconnect in Washington, DC or in other locations with Level3 or other providers? The reason is that, Cogent didn't meet larger networks' requirements, plus these people are already pissed off for Cogent taking their customers, and thirdly, Cogent just forced their way into their network with the acquisition of NetRail who had previous interconnects with these networks.

So now, you have REALLY pissed off networks who want to retaliate, so what do these networks do? They simply say, no additional interconnects into our network or circuit upgrades, and they have a right to do so because it's their network, but they do honor the current interconnects due to the original contracts with NetRail. Rumor is that Sprint and other providers have already voiced this to Cogent, and I'm pretty sure many others will because they're tired of hearing "$30 bucks per meg" floating around the office.

So what do you have now? A network that's lagged as fuck during normal US business hours and that can't be upgraded without raising prices (which they're doing), but long story short, the plan was to get customers in the door, and jack up prices, so watch for phase 2 of the operation (assuming they don't go bankrupt beforehand).

So now your Cogent powered network has customers who are pissed off at these large fucking round trip times, slowness during the day, you might be making a profit, but you also are now operating a revolving door hosting company (group comes in due to low pricing, group walks out due to shitty network).

But, hey, you get what you pay for.

http://missnglnk.com/way2fast-hosting.gif

go for some place that has multiple feeds into the same building. some may call it a "colo hotel" - you'll benefit because their network is fed from different carriers and since it's all in the same building, the hosts you go with don't have to pay pricey local loop costs and can pass that on to you.

places like fci are housed in the westin building up in seattle, which houses most of the telco carriers on the west coast. they only have to pay for the amount of bandwidth they want, no local loop (normally the line ran from building <-> telco) - those type of companies are where your money is well spent. because they don't have to pay local loop they can buy bandwidth in more quantities as well as diversify their composition of carriers, it's a win/win.

great post missnglnk!!!

Lots of great posts in here :)

Brad

Wow - the first post I have printed for future reference from GFY... Absolutely valuable info.

I have been paying over 2$ per gig for the latest two years... I have seen how prices gone down down and down.... but I have been with my lovely host. They have nerver been cheapest ones, but... I have really got value ( no worries, minimum downtime ) for my hosting bills.

Switch from $2/ gig to $0.50/ gig and blow your business :)

just a small "devil's advocate" post.

Dont let your host get away with high bandwith prices.

the price of bandwith has come down and his still coming down. Use that to your advantage. Mention Cogent to them ( that always piss them off)

Dont be a sucker and go for underpowered, overcharged hosting companies. But dont be a sucker and overpay just because your server isnt coming down every couple hours.

;)

Indeed. I still think Cogent was the best thing to ever happen to the hosting industry (at least from my perspective, hehehe).

The lowered prices have nothing to do with cogent, it has to do the fact that the market is having a huge decline..

also true

i think there are a lot of good points made about cogent and some of the hosting companys who use them, but you knuckleheads take 'you get what you pay for' way too literally and too often as your mantra. This 'slogan' u guys make it has a grain of truth to it but in a technical industry it doesn't neccesarily apply.

How many of you host your servers on linux and freebsd? using a FREE webserver, a FREE ftp server, a FREE ssh distro, FREE monitoring tools (mrtg cricket blah blah), FREE database, FREE compilers/build environment, FREE X server and on and on

How many of you regret this horrible decision and are in the process of paying big $ to "upgrade" to windows servers instead? Or better yet Solaris?

Just because its sun and microsoft it must be worlds better, right? It's the same thing with applying logic like "the first thing i look for is if they have all cisco gear" if you've ever really worked with cisco, extreme, foundry, arrowpoint's equipment ETC you know that frankly cisco is WAY overpriced and their performance and features SUCK in comparison. Cisco was an industry leader at one point but since then they've been living off their name. For the past 3-4 years most major internet companys have drifted away from cisco and are using a lot more foundry and extreme, because they offer more features, a better interface, and often times phonomenally better performance at 60% of cisco's pricepoint.

Cisco is not that great or brilliant of a company, have you ever used their first generation of any product line? Even their first set of dsl equipment a few years ago? It's always horrible under quality assured pieces of shit because they know people will buy it since it has the word cisco on it, and will just wait for the firmware upgrades over the comming years to smooth out the problems. All the great values that have come out of cisco the last couple years like their loadbalancers the "localdirector" that was a $20,000 pii-300 which would max out at 1000 connections before they bought out arrowpoint and restamped 'cisco' on them. How about cisco's PIX firewall? You really get your money's worth if you spend $10,000 to upgrade a piece of hardware on your switch just so you could route between vlans when every other vendor on the planet does this with software for 20% of the price, and with less hardware still manage to do true wire speed. Yahoo, exodus, earthlink you name it, all using foundry and other non-cisco or cisco-rebranded equipment for truly massive amounts of traffic.

This isn't intended to be a rant against cisco, but rather to make a point that because something costs a lot or a company has a recognizable name does not mean they have a superior product.
It's the same thing with saying "only use dell or compaq equipment", what the hell is that? You like paying a 100% markup?

You think theres some magic motherboard in the machine made by an ancient taiwanese kungfu motherfucker in a strawhat holding a bag full of moondust in one hand and a soldering iron in the other? They buy the exact same components you can, put them in a case, and charge you for a support contract, when any of the parts you buy yourself will have a 3 year warranty on them anyway. Now you can pay an extra $3000. Woopie! You get what you pay for, because $599 for a gig of ram, $99 for an upgrade from a celeron 850 to a piii 850 (no joke, check their webserver prices), $270 to UPGRADE from a 20gb drive to a 120gb drive is sold to you inside of a case that has a plastic "dell" logo on the front. WHOHOO BARGAIN CITY!@!# And hey, if you want to upgrade to a second* p3 processor, on their dual systems, its only $500 for a p3 1.13 ! You get what you pay for though because this processor was flown in personally via helicopter from the hollow center of the earth by a cyborg who stamps them individually with his testicles before they're blessed by 3 priests baptized and packaged in the reborn baby jesus's foreskin.

Speaking of value, to upgrade from a p3 1.13 to a p3 1.26 its only $100 difference, for a grand total of $599! GOOOO DELL. It's the same deal with compaq. Except for some extenuating circumstances (say you have so much money you really dont care and would just rather be raped in fees so that someone can come on site and swap in new ram for you once a year and you never have to think about it or whatever) it makes plenty of sense to buy the components saving a SHITLOAD of money and put the server together yourself (takes all of 20 minutes). Scoffing at people who do this is ignorant.

There seems to be almost a sort of social conditioning here where people who are more cost-conscientious
are ridiculed for being 'cheapskates' and whatever and the old chant is mindlessly repeated pay more pay more pay more. It looks to me like it just feeds on itself and allows people to try and justify guaging those who don't know any better. For instance, I asked jason from cologroup the other day how much he would charge for 1u of space. He offered $100. $100 for 1u of space a month on a rack that cost him less than a grand. There is air conditioning, power, all that to be taken into consideration too of course, but still thats $40 higher than the most expensive price i've ever seen for 1u of space including big colos like level3. Am i getting what I pay for with jason because hes got the bomb ass inch of space for me? Wheres the added value? Must be because that rack is made of adamantium?

No, he's just used to dealing with people who live blindly by the philosophy "You Get What You Pay For".
:2 cents:

DMR,

as with all things, there is no hard set rule, its a case by case situation.


The 'you get what you pay for' doesnt apply to servers made by Dell and such, true.

But the price difference between a IDE and SCSI system falls in the 'you get what you pay for'

or the Cogent Vs Other Tier1 situation

its all relative...

p3 vs Athlon, etc...

WOOHOO! ! ! ! http://www.opticalamplitude.com/assets/rockon.gif

fuckin' DAMN dmr! I think I may cry! That post kicked ASS!!!!

This topic was good, I agree. I can say though, our dedicated customers have been very happy with Cogent and our virtual customers are about to experience the same.

It may kind of kill the compairson, but so be it. :) I am not in this business for short term, it is the long term I am focused on.

my favorite part.... http://www.opticalamplitude.com/assets/rockon.gif

i've always had less problems with custom built pc's. not sure how custom built servers would be, but i'm assuming if the tech knows his shit it'll be better than any overpriced dell system.

not to say that they suck or anything, they're just on the pricy side.

yes, i too liked that part

drguile don't get me started on ide vs scsi :warning
:winkwink:

The best way to use cogent is to have all your equipment in a cogent lit building ... and then have every surfer that hits one of your hosted sites also in a cogent lit building. This way everything stays on-net. On-net cogent traffic is very speedy. It's only when the rest of the internet wants to push or pull anything on or off cogent's network that there are issues. :eek2

This site is for netrail customers only. If you aren't using netrail, its gonna be slow! :)

Nice info guys, thanks

Another issue I have come across with a previous host that has not really been covered here is badly compiled software, apache, php, OS patches etc.

Really I am not in a position to tell if a server has a well compiled version of apache on it, gotta rely on someone following through with their promises.

I've not looked at the facts, so I could be wrong, but.. in all likely hood this was some 'funny' $$ where cisco invested hardware to own there place inthe network

Cyberpunk

in case anyone was wondering what cogents been up to lately: http://biz.yahoo.com/prnews/020318/dcm024_1.html
heres their latest press release.

uhhhh... what Jamie said!

Nice read... :thumbsup

DMR rocks, what he said is very true. If your TGP posting and 99.9% of them are freeloaders, they deserve Cogent bandwidth. If you run a actually paysite, your customers deserve better =P

Cogent bashers will become Cogent lovers once they actually have a server on them. I know someone that use to bash it like fuck and now lease a full 100Mb connection off them. Allot of people just use a Cogent reseller.

We like any good host tested cogent, and I have to say i am no cogent lover..


400ms to one state away just does not sound very good :)

Hmm, when did you last test Cogent? I think they are better now. My results are not too bad. What do you guys get?


Pinging www.verisexy.net [66.28.252.5] with 32 bytes of data:

Reply from 66.28.252.5: bytes=32 time=149ms TTL=240
Reply from 66.28.252.5: bytes=32 time=186ms TTL=240
Reply from 66.28.252.5: bytes=32 time=143ms TTL=240
Reply from 66.28.252.5: bytes=32 time=144ms TTL=240

Ping statistics for 66.28.252.5:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 143ms, Maximum = 186ms, Average = 155ms

Pinging www.greenapple.verisexy.net [66.28.252.3] with 32 bytes of data:

Reply from 66.28.252.3: bytes=32 time=144ms TTL=49
Reply from 66.28.252.3: bytes=32 time=143ms TTL=49
Reply from 66.28.252.3: bytes=32 time=143ms TTL=49
Reply from 66.28.252.3: bytes=32 time=155ms TTL=49

Ping statistics for 66.28.252.3:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 143ms, Maximum = 155ms, Average = 146ms

I get great ping times to way2fast.com but there's a price for it. I got these from your site for colocation.

1-2 mb/s - $400
3-5 mb/s - $390
5-7 mb/s - $375
8-10 mb/s -$370
11-13 mb/s -$365
14-15 mb/s -$360
15-17 mb/s -$350
18-20 mb/s -$340
21 mb/s -$338
22 mb/s -$336
23 mb/s -$334
24 mb/s -$332
25 mb/s -$330
Over 25 mb/s -
Call

I used to be a Cogent lover, then I became a customer on a Cogent network, now I hate Cogent.

Right now I will drop you a 100meg feed for $6500.. thats $65 per meg, not more expensive then cogent.


This month i will be glad to deal with anyone who wants to host with us, Ignore the prices on our site and come directly to me.

That's pretty good pricing. I think Williams has similar pricing if you buy 100Mb ethernet from them.

I did, he made me an offer I couldn't understand, so I took it :thumbsup

Ok kidding, he did make me a great offer, I took it, getting ready to move shit now.

:mad:
100M for $6500, $65 p/M.

Right now, you are PRAYING that your customer utilizes MAYBE 10M of it. HOPING he bursts once or twice to 30-40M...then, in essence, you just sold a 10M pipe for $6500...Not bad.

And if the customer MIGHT sustain 75M constant...then you lose your ass...

See, you buy pipe from GlobalCrossing (who are Chap11, may I add). You have an OC-3 from GLBX...So, you just sold an uplink to share your network with all the other people you sold a 100M uplink to...So maybe you have sold 6-7 people at $6500 each to share a 155M pipe. (See folks, this is what happens). You PRAY each and EVERY night that your customers MAYBE hit 5M. All in the while, your customer boasts to his friends "I BOUGHT 100M FOR $6500!"

Let me share a quote direct from the site.
Hi-End Media hosting clients are placed on a multiple OC-9 connection. Our servers are in faciluties that uses less then 50% of available bandwidth, at peak times.

Can you answer me something? What in the world is an OC-9??? I would LOVE to know! Three OC-3's do not constitute an OC-9 by the way.

This comes to my conclusion SinEmpire's case in point on the difference between a host using Cogent's business model and a host riding TRUE ethernet from a Level3/Verio.

It's like getting buy one get one free at K-Mart...Two of shit, is still shit.

MrBandwidth
:mad:

Wow! Looks like Mr Bandwidth knows his stuff. See people, this is what I'm trying to drive home to you. Alot of hosting offers are NOT what they seem. As he points out, many hosts that sell bandwidth profit only by overselling their network and having clients that don't use all the bandwidth they're buying.

Brad

:glugglug
Well, my whole point is this:

Mercedes and Kia both sell cars. Both cars get you from A to B.

The Kia will break down and take a shit every 5000 miles, but the Benz goes like a champ for 100k.

You replace the Kia every 50k miles, while the Benz lasts for well over 300k miles.

The Kia costs $10k, while the Mercedes costs $50k.

Do the math...

This works the same for just about every busness out there...

Kia and Mercedes can co-exist in the marketplace, and both will have a substantial customer base...it just all depends on how much "garage" work you want to do in between tune-ups.

MrB
:glugglug