Tier-1 networks - top of the food chain

[thecoolh]

www.ssihost.com ... a friend's freehost... hosted on a cobalt box @ rackshack... box went down... just for nothing... rackshack says its "boot looping"... they define it as: boots up to say cobalt networks and begins to uncompress kernel and then suddenly goes back the cobalt networks part

i mean... whats up with that crap?!? now they want to reformat the box... and won't even put in an effort to backup the data... ! nor will they send the box out to the client to backup the data... (but then again, they don't have to cuz its legally their box and technically theres no way to stop the client from running off with the box after its been shipped to them for data recovery, hmm)... so... what does this mean? rackshack sells a defective box which dies and causes the client to lose all his data, and they don't even bother trying to help the client out.... ! what kinda crap is that!?

he can't post anything in here simply due to his contract with rackshack (and i don't think he has an account here)... he didn't even ask me to post this here... but i mean... when crap like this happens to friends, it pisses me off... now... if this is happening to people i know... i wonder how many other people its happened to... hmm...!

[thecoolh]

oh... btw... he's not getting a credit for all this service downtime caused by faulty hardware they provided him with... and the reputation of his sites are getting screwed in the process...

[hrmedclinic]

I saw your post.....5500/month for a 100mbps connection - WOW......
I mean is 128kbps = 40gigs of bandwith... its comes out
10mbps = 3125 GIGs....then 100mbps = 31250GIG/month...

thats like 17 cents per GIG.

I hope you are not some noneprofit organization :)

BTW you are not related in any way to Coolwebhosting.com? which later on became Bic Picture Inc..... and then cut off without warning 2000 some of their clients, packed their bags and left?

[ljames]

1) If you knew anything about RAQs you would know this happens. We have lost three boxes because of this. It is a pain in the a**. You just have to reformat and restore.

2)
and the reputation of his sites are getting screwed in the process...

How can you have a reputation on a $99.00/mo box??? That is like me driving around in a BMW and parking it out front of my trailer I live in. If you want a managed box, $99.00/mo isn't going to do it. That was just a bad business decision if you put something mission critical on a $99/mo box. Why would they refund a hardware failure? They told you the problem and how to fix it to where downtime is small. Also, with proper backups and a restore, you wouldn't be down more then 4 hours.

3)
I saw your post.....5500/month for a 100mbps connection - WOW......

Not really. Cogent is well under this. He is using "NON-Cogent" which means he is still using GIG-e either from Yipes, XO, Level3 or a few others. Nothing new. That price actually is a little high from what we have been quoted from other GIG-e's.

[hrmedclinic]

Okay,
So how much traffic per month is 100mbps is?

Most of the company I looked into give $ 2-7 gig/month.

May be my calculations are wrong, i'm not a tech guy - but 17 cents per gig is DEAL :)

[thecoolh]

How can you have a reputation on a $99.00/mo box???

Should it matter what the price of the box is? shouldn't it be dependant on the fact that its a box at all and that he should get atleast what he pays for? A server with 300GB/month traffic allowance and a bunch of diskspace... a server that stays up and the datas secure on...? if it were a colocation package, then uptime wouldn't be rackshack's concern... but that isn't the package, the package is where rackshack provided him with hardware that failed... and they won't even assist in recovering the data.


Why would they refund a hardware failure? They told you the problem and how to fix it to where downtime is small.

Now how would he fix a hardware issue with rackshack's server itself? Did rackshack give out console access to servers when I wasn't looking? And why should they give a service credit, not refund, on the downtime? because its managed hosting, not colocation! dedicated servers inherently imply that the webhost will take the responsibility of keeping that box up, period... I don't expect abovenet or exodus to give two shits if one of my boxes fail, because they're colocated! Not only does colocated mean they don't have to care, but it means they can charge me an arm and a leg to care! but dedicated, completely different story, the server is 100% their responsibility!

As for backups? isn't it a reasonable expectation that one's server will be stable and not have hardware failure causing total data loss...? I mean, thats what he paid for, isn't it? If they can't offer all they imply for the $99/month, they should kick the price up, not pretend alls good and move on.
----
as for my b/w rates... hehe... :D

btw... not related to "Coolwebhosting.com"...


Cogent is well under this.

cogent is $3000/month (for any commercial purposes other than a simple office connection which means no hosting, period, cogent = $3000/month)... add to that the price of an office (in a building thats cogent capable) or rackspace in some datacenter (carrier neutral thats actually cogent capable)... probably gone over $5500 already... now take into consideration, once you've bought that feed and the rackspace, you're only connected to one provider, cogent communications, at the tier-3 level... the 100Mbps i offer will be colocated directly in a tier-1 datacenter with a company that has their own private nationwide backbone (bought fibers off qwest a few years back, manages their own peering at all the major points including MAE-EAST, MAE-WEST and MAE-CENTRAL; this is not my own datacenter, obviously), not my office, not some 3rd rate colo... a tier-1 datacenter... :D and with all tier-1 datacenters, theres peering with virtually all the larger carriers, along, in this case, two hop, 0-1ms, access to MAE-EAST... :)

[ljames]

My point is this. If I go buy a Ford Pinto and it breaks down, who's to blame? Also, servers are machines. Servers run software. We all know this. We by now should also know that they break and we should back up everything that we can not recover. A $99.00 server and a $1000 server both will have issues. However one will usually have extra drives, RAID, redundancy, etc.. A $99.00 RS server has non of these. It is a single point of failure. Your friend gambled. He lost. It is nuts to say RS sucks because of this.

Also, it is not RS responsibilty to restore files. If you didn't back them up - how do they restore anyway?? If your friend was wise he would have ordered a second drive to run backups to. He would have also paid for management so if this happens they take care of it.

Listen. I agree it sucks. We have lost many Cobalts in 3 years. We have grown wise over the years. Custom backup scripts. Propriatary restore software, etc... Just don't say a fellow host sucks because your friend ran a business on a cheap server.



Okay,
So how much traffic per month is 100mbps is?

Most of the company I looked into give $ 2-7 gig/month.

May be my calculations are wrong, i'm not a tech guy - but 17 cents per gig is DEAL

Cogent and other GIG-e providers do not deliver bandwidth the same as ATM/SONET providers. Gig-e's use the latest in IP over Fiber Optics technology, meaning you get direct ethernet connectivity. This means no telco frustration and expense of T-1, DS-3, OC-3, etc. It is pure bandwidth at the 10, 100, and 1000 Megabits/second speeds. Not the standard tier1/baby bell Tx/DSx products which will always be more expensive.

SDColo gives you a 10 MB connection for $2000/mo with a full rack. $4000 gives you a full rack and a 100 MB connection.

thecoolh, is this standard bandwidth or ethernet? If it is a standard drop, I am amazed at the price.

[thecoolh]

If I go buy a ford pinto that was under full warranty (a dedicated server is meant to be fully managed by its provider) for everything, bumper to bumper, at no extra charge, then i sure as hell would expect the dealer to make good on their promise of making sure the car works... and if it does fail, i would expect they would deal with the matter in a timely manner... also, if I took the car into the dealer (after having used it for a while, so there are personal belongings in the car), i would expect them to atleast save the belongings to put into a new pinto (if the original pinto is damaged beyond repair), not just scrap it all... !

For my dedicated clients, if something happens with the box (thats caused by them), then its their problem... i'll happy do a drive reformat (but backup all data first) for a nominal fee... but if it happens because i provided them with crap hardware, then its my fault and i'd be "man enough" (not trying to be sexist/etc, but u ppl get the point) to take the blame and do everything in my power to get that box back up, with the client data in tact! Restoring data is not a matter of an actually corrupt hard drive... at the very most, i'd suspect a corrupt kernel/etc... personally, i'd just grab the hard drive and pop it into another server, copy everything over into a directory, reformat the cobalt, copy all the data back into a subdirectory for the client to sort out...

not just say "tough luck dude, we gotta reformat your box, your datas lost".... in my eyes, a dedicated server provider has the same commitments to a client's site/server uptime and data integrity as a virtual hosting provider... period!

------
SDColo is another one of 'em cheap "cogent providers"... ever wonder why rackshack dropped their cogent feed? from what i've been told, it was because of the poor quality of cogent bandwidth... i mean... think about it... they had a gigabit cogent feed, which they paid a hefty setup fee/etc... and yet they dropped it like two and a half weeks of using it... i really doubt they'll get their setup fee nor their first month's service fees back... they didn't even wait for a full month to go through... and in the process, they are running their remaining feeds quite high... i mean, something must have pissed them off enough about the cogent feed to drop it and run the remaining feeds at 80%+, eh? atleast rackshack cares about network quality...!

SDColo... they fall into the tier-3 category, wouldn't you agree?

as for "latest in IP over Fiber Optics technology"... its called Packet-Over-SONET (POS)... the main feature is that it gives about 30% better throughput over a OCx feed compared to ATM... but the handoff for this extra throughput is the loss of being able to QoS policies (giving packets priority levels for delivery, so less important data can wait like ftp, smtp, etc while more important data gets to go immediately like video, www, etc)... :P in addition to that, theres Dense Wave Division Multiplexing (but that has more to do at the physical/transmission layer than the data layer, but its DWDM that increases a fiber's capacity over 100x with absolutely no side effects except slightly more expensive termination equipment... and the requirement of higher grade fiber between points, but thats standard in virtually all SONET feeds nowadays)...

My service provider does use POS throughout their nationwide backbone (clients servers are directly colocated within their datacenter), but not DWDM yet (their backbone runs at less than 10% at present, no need to at all). I guess by your definition, its a "standard" drop, as their bandwidth comes in through OC3s, OC12s and OC48s... they're virtually in the same city as MAE-EAST and gains a lot of their peering partners directly from there... (for example, i get 2.6ms to w3.dcx.yahoo.com (9 hops), 2.9ms to goto.com (14 hops), 8ms to www.flying-hosts.com (9 hops)... most of these traceroutes leave the local network by hop 2, the rest are within the remote provider's network...

[ljames]

If I go buy a ford pinto that was under full warranty (a dedicated server is meant to be fully managed by its provider) for everything, bumper to bumper, at no extra charge, then i sure as hell would expect the dealer to make good on their promise of making sure the car works...

My point exactly - in a round about way. No one gives a bumper to bumper on a Pinto. RS NEVER posted or said they have a bumper to bumper on their servers. Also, I have been in the business long enough to know the difference between a dedicated server and a managed server. Your friend got a working server. It broke. That comes with the territory. RS offered to fix it. Only a fool (no offense) would expect a provider of a $99 server to take the time to do a full restore of all data. Was it RS's fault the drive failed?

I am not saying another provider might not have done so, as many will bend over backwards for people (which it sounds like you do for your customers), but a rack em - stack em for $99 sure isn't. Common sense should have dictated this.

I can see we do not agree on this, so I will bow out. :)


as for "latest in IP over Fiber Optics technology"... its called Packet-Over-SONET (POS)...

Sorry, but you are incorrect. Cogent does not use SONET. It is called IP-over-ethernet. If all you're going to do is route IP packets, it doesn't matter if you use IP-over-sonet, IP-over-ethernet, or IP-over-PPP. The only difference is that ethernet equipment is incredibly cheap because it is in ubiquitous use. The main reason that IP was ever routed over sonet was because the early ISPs didn't want to dedicate lines to their small IP business. In the beginning, all the IPv4 providers were also telcos, and were using bandwidth from their designed-for-voice-traffic networks -- which all ran SONET. With SONET it was easy to interleave voice and data traffic over the same optical fiber.

IP-only carriers don't have to worry about mixing with voice traffic, however; they have a raw fiber down which they can send whatever they like, in whatever format is convinient. So they use ethernet, because it is cheap, reliable, easily replaceable, and well understood.


<Edit>I understand SONET standards cover the base rate of a data stream over optical media. I was referring to Cogent and SONET. Just a quick point. </Edit>

[thecoolh]

I can completely see what you mean and definately respect your standpoint. Its what makes the difference between a large host and a smaller one (like me, although I may use more b/w than most of the hosts on this board, I am in no way a "larger" host)... the smaller ones are more willing to go the extra mile for their clients whereas the standpoint of a larger one is exactly as you stated... I've only found 1 large colo that goes the extra mile for their clients... 1, and i've got servers at about a dozen of 'em... (working to consolidate all my servers to that 1 colo btw)... Hell, i'm actually even trying to relocate myself and a few of my key technical staff to be local to that specific colo, I'm so pleased with their service i'm willing to go across the continent (west -> east coast) and cross into another country (Canada -> US)... Additionally, I could easily start a service to "compete" with rackshack and their prices... the only reason I don't is because I'm not local to the colo... I'd want to be able to go "the extra mile" for the client, which I obviously can't if I can't get physical access to the servers...

as for the whole debate on IP-over-ethernet vs SONET... I'll take your word for it... I was under the impression that all data over fiber (not counting short distance methods such as FDDI) would conform to SONET standards (SONET refers to the physical medium), Asynchronus Transfer Mode (ATM, the traditional method) being the cell-based method of sending data over SONET... and the new one being Packet-Over-SONET (POS, ethernet is a packet based method vs ATM's cell)...

Heh... well... regardless, I know my provider's transit doesn't come in via ethernet for sure... :D

[bitel.biz]

Heh, I have a friend that had problems with them too and he is reseller for them that sold more then 100 of their servers.
They said hard disk failure... who knows... maybe, but his site and his customers sites are down for more then two weeks.
Good way to lose customers :(

All of you people that have servers with RaskShack, do backups on your own as if something goes wrong you are on your own, don't expect help from them.

[hrmedclinic]

Hmmm may be i'll consider moving my server to your facilities...
Right now i'm with DataPeer. They are a little expensive but I havent been happier with noone esle. Do you have everything that they offer with their dedicated package?
We use multiple OC48 backbone connections running into our facilities.
All of our servers are monitored and protected 24/7/365

Virtual Protection.

1) Packet filter: Looks at each packet entering or leaving the network and
accepts or rejects it based on user-defined rules. Packet filtering is
fairly effective and transparent to users.

2) Application gateway: Applies security mechanisms to specific
applications, such as FTP and Telnet servers.

3) Circuit-level gateway: Applies security mechanisms when a TCP or UDP
connection is established. Once the connection has been made, packets can
flow between the hosts without further checking.

Physical Protection/Network Features.
Your servers will be located in the state of the art facility with following securities/features in place:
-Dual diesel-fuel backup generators
-Motion sensors
-Video surveillance 24/7/365
-Security breach alarms
-Fingerprint Scanners
-Liebert Climate Control Systems
-CiscoCore6509/Arrowpoint800/BigIron8000 Switches

Toll-Free 24/7/365 over the phone/e-mail or TroubleTicket system technical support is available for all of our customers.

I have an app server XEON machine with them with netscreen-5 as well as I am using their HITACHI SAN machine (only 15 gigs of it) to back-up/sinc my data.
Most important things for me are : toll free techsupport (24/7/365), good connection (at least OC48 from different backbone providers) and of course uptime. So far I havent been down even for a second.
If you can match all of that and get me a good price I'll consider switching to your coolhosing company.
Thanks

PS where is your datacenter phisicaly located?

[thecoolh]

umm... I'd rather not start stating prices in detail in here/etc... otherwise it starts seeming like too much of an advertisement... do you have ICQ? (me: 74440896). AIM? (me: NetDudeMcAffee). Phone Number? If you wish to call, can give me a ring at 604-318-7654. Or, as a last resort I guess, can drop an email by netdude@telus.net. As far as where the datacenter is: McLean, VA (very close to MAE-EAST)... As far as setup/etc goes, they'll go and install the hardware... reboots can be done in realtime at no extra charge (they have full 24/7 NOC staff; 1800 number to it)... commands typed into the console ("remote hands")/etc can also be done in realtime at no extra charge at anytime... but if you need hardware opened up and replaced (faulty network card/etc), that'll have to be done during business hours (but what i love about it is that if for example say a network card is faulty, they can install one of theirs for free until you get a chance to send one out)... so far, i've had two of my own servers fail on me... they've swapped my drives into backup servers I had there, afterhours at no extra charge (i did pay him as a show of courtesy; one of the head systems admins thats a personal friend happened to be online on AIM and went back to the datacenter just for me)...

So far, I haven't experienced a single second of downtime on the network level.

As far as network capacity, they're at the tier-1 level... beats out "having an OC48"/etc... :)

[thecoolh]

hrmedclinic, is your site: HRMEDCLINIC.COM ? i noticed that domain points to datapeer's network... but the sites down...?

[Technics]

thecoolh,

From your many years spent at FreeWebSpace.net you should know full well that a cobalt RAQ cannot handle the pressures of a free host. You should of told your friend this in the first place and pointed him in the direction of a server that had the power to handle this kind of service.

Does RackShack not state clearly in there terms of service that they do not take responsibilty for backing up users data? They will replace any faulty hardware but dont say anything about data.

I have a server located at RackShack, kinda wish id stayed with tera-byte as most the tech's at rackshack are thick as two short planks when it comes to dns issues. Apart from the poor knowledge of DNS they provide a very good service for $99 a month.

[ljames]

As far as network capacity, they're at the tier-1 level... beats out "having an OC48"/etc...

Again. This is not true either. EVER data center, regardless of if it is a tier 2 or tier one, will have drops that are lit. Verio is a tier 1, they only have OC 12s here in SD. They have OC48 drops ready, but they haven't turned them on. I almost guarantee you that your datacenter does not have ANYTHING greater then an OC-48 lit. Last I recall, not a single colocation has a lit OC192 in place. The cost is huge! and the bandwidth to fill an OC 48 is so large right now no one is filling those pipes.

Even if you are in a MCI building, you adhere to the same stipulations as "having an OC48"/etc... ". The bandwidth doesn't magically appear at tier 1s.

Also, in my opinion, teir 2s are better because they don't have single network points of failure. Verio will go down when its network is down, regardless of their peers. This is why I like Inflows or other tier 2s, they have multiple backbones dropping right into their network. So if their Verio pipe drops, their other ones will pick up.

Just my 2.

[thecoolh]

dude... lemme get you right: you think its actually possible for a tier-1 provider's network to go down? your speaking as if: a router fails and its all over for the entire network and all its peering and transit points. tier-2 providers are generally regional telcos, how could it be better to have a regional network than a national/global?


Verio will go down when its network is down... ... teir 2s are better because they don't have single network points of failure ...

You speak as if you theres a nuclear attack at one datacenter, their entire network is toast. Out of each of their POPs, not only do they have peering to outside carriers but they'd have connectivity out to other POPs of theirs, providing redundancy to all their connectivity period. And as for datacenters, well, not only would they definately have connectivity to all the POPs "within reach", but they'd also definately have connectivity to all the peering partners within their reach too.

How could a tier-1 network have a single point of failure, at all? That'd be more like a tier-3 network (but that doesn't happen much more nowadays as even most tier-3s have redundant connectivity).



Even if you are in a MCI building, you adhere to the same stipulations as "having an OC48"/etc... ". The bandwidth doesn't magically appear at tier 1s.

How extensive a tier-1 carrier's network is would determinte who it has the capabilities (from a "actually get approved by" standpoint) to peer with. Nationwide networks get to peer with virtually anybody, sure as hell beats paying for an OC48/etc. With sufficient peering partners, even if not at the OC48 level (usually, peering occurs on a on-demand basis, not just throwing in a HUGE pipe and praying it gets used), it'll beat having an OC48 to a few carriers anyday as long as none of the feeds are getting overused.

Having an OC3 running at 50% is no different from having an OC48 run at 50%, as long as theres ample b/w available for spiking/etc, it shouldn't exactly matter the total b/w in the circuit. Having OC12 peers with level3, glbx, qwest, exodus, uunet, sprintlink, etc/etc/etc beats out having an oc48 to uunet easily... Trying to deliver as much data directly to the target is always best, regardless of the connection speed it goes over as long as b/w is available.

----

If I'm not clear on anything here, don't mind me, I just woke up 5 minutes ago... LOL... just post a message complaining about it and i'll try and clarify.

[cory]

I finally had enough and am in the process of removing everything from their peice of crap server now.

beware of RackShack

they really suck the big one!!!

Spammers are able to get in through the /cgi-sys to use the mailform.pl to sprew their garbage.

[ljames]

dude... lemme get you right: you think its actually possible for a tier-1 provider's network to go down? your speaking as if: a router fails and its all over for the entire network and all its peering and transit points. tier-2 providers are generally regional telcos, how could it be better to have a regional network than a national/global?

No offense, but it is clear you have not enough knowledge in this arena and that you haven't been around long enough. If this were not true you would know about the MCI issues a few years ago. MCI updated their router software. Went bad. The whole!!! network (yes tier 1 - largest in the world with UUNet) went down. For days. It cost the lives of many small regional ISPs at the time. Verio.. post around. I have personaly experienced a 2 hour outage at the SD location, noit because of the DC, but because of the network. Do a little more research on this subject.


Having an OC3 running at 50% is no different from having an OC48 run at 50%, as long as theres ample b/w available for spiking/etc, it shouldn't exactly matter the total b/w in the circuit. Having OC12 peers with level3, glbx, qwest, exodus, uunet, sprintlink, etc/etc/etc beats out having an oc48 to uunet easily...

You got to be kidding me. Is it the same running cost to operate a Yugo at 50% and a 747 at 50%??? The hardware used to operate an OC12 and an OC48 are not the same. It involves a lot of money in upgrades. Some providers are smart and buy big first. But many don't. Also, do some research on the cost to operate a OC 192 into your DC. You would die. Again, I am not talking about bring one in for later, I am talking about lighting it. Imagine have a garden hose on full blast. What happens when you run it to a fire hydrant????


How extensive a tier-1 carrier's network is would determinte who it has the capabilities (from a "actually get approved by" standpoint) to peer with. Nationwide networks get to peer with virtually anybody, sure as hell beats paying for an OC48/etc. With sufficient peering partners, even if not at the OC48 level (usually, peering occurs on a on-demand basis, not just throwing in a HUGE pipe and praying it gets used), it'll beat having an OC48 to a few carriers anyday as long as none of the feeds are getting overused.

You are missing key points and are also losing the word "transit". Transit and peeering are two seperate things. Tier ones peer. Tier twos buy transit - they don't peer. Tier twos buy transit from multiple providers for redundancy. If your understanding were true, then answer this? Why do tier twos buy transit from other tier 1s if a tier 1 will never go down?? AT&T doesn't buy long distance service from MCI. Why would they buy transit?

[thecoolh]

MCI issues a few years ago.

Yes, i am quite aware of that issue... Along with the AT&T frame cloud collapse a few years ago that caused all of the financial services in the eastern united states to be pretty much screwed (including ALL atm machines/etc). Do you really think something like that is ever going to happen again with today's redundancy? Cisco surely learned their lesson... Just as, do you really think that something like Sept 11th is EVER going to happen again?


You got to be kidding me. Is it the same running cost to operate a Yugo at 50% and a 747 at 50%??? The hardware used to operate an OC12 and an OC48 are not the same.

I wasn't speaking of it in a network standpoint but rather of a client's usage standpoint! huge difference.


... operate a OC 192 into your DC.

Who the hell said I owned a datacenter? I think i've clearly stated in the past I colocate my own equipment at a much larger service provider as I won't put my client's uptime at my capabable yet financially restricted hands vs a tier-1 nationwide provider thats been around since '93.


Why do tier twos buy transit from other tier 1s if a tier 1 will never go down??

Because buying transit from just 1 provider would mean just one possible network path (whether it goes down or not)... and relying on the path they choose your data goes through. One would choose multiple providers simply so that one's data would be able to take the most efficient route, not just the 1 route thats available.

And... Yes, tier-1 providers out there DO buy transit. There are just some providers out there that can't bother and peer with everybody in sight, network infrastructure wise, it sometimes just comes out cheaper to buy a link for transit from another tier-1 provider that has all the network connectivity as desired, instead of going out and trying to light up fiber to 'em independantly.

For example, time-warner telecom (a tier-1 provider to the best of my knowledge) doesn't have a peering feed with the now defunct @home network, they're buying transit off of sprintlink to get through to them (@home's network still seems to be lit, scheduled to go down in feb i think). nor does twtelecom have peering with level3! they're both tier-1 providers...!

Yeah, you've probably been around the 'net for longer than I have, but I ain't no moron... :P I've got subscriptions to virtually every IT magazine available in north america, and its neatly stacked in my washroom as reading material... LOL... j/k... hosting/network+systems design has been my life for the last 2 years, a serious hobby for 3 years before that... that and my girlfriends... but thats it... dude... how many guys u heard of that can clearly think about network design while having sex? LOL... go over 3 days w/o sleep and not feeling tired as long as he's in front of a computer? (but the second i leave, gone like a light, lol)

SD location? san diego?

[thecoolh]

BTW... NEW INFORMATION ABOUT RACKSHACK

they do in fact have a feed to verio contrary to my previous beliefs... I have found a route that goes through their verio feed...

http://netdude.much-list.com/rackshack-verio3.gif

[ljames]

No offense, but remember this post. In a few months to a year you will look back and laugh as you learn more.


Do you really think something like that is ever going to happen again with today's redundancy? Cisco surely learned their lesson... Just as, do you really think that something like Sept 11th is EVER going to happen again?

This sums it up from my viewpoint. Were we ever going to see another world war after the great war? Listen, using "nevers" in this business leads you to failure. If you really believe this, then drop all your connections, sign up for one tier one and enjoy.

I am going to drop out of the peering vs transit, etc.. debate. It is going nowhere. Howvever, I just have one request ...


And... Yes, tier-1 providers out there DO buy transit.

OK, You posted it. Name one and where you found this information. I'm waiting ..........

:p

[thecoolh]

If you really believe this, then drop all your connections, sign up for one tier one and enjoy.
you stupid or just can't read? Didn't i just say why a tier-2 provider may use multiple tier-1 providers, even if the link may be stable!? and in network connectivity standpoint in general, a link may fail to a single tier-1 provider, but that has nothing to do with the entire tier-1 provider going down... !


OK, You posted it. Name one and where you found this information. I'm waiting ..........

i gave two examples: twtelecom + level3 (twtelecom buying transit from uunet to get onto level3) and twtelecom + @home (twtelecom buying transit from sprint)...

transit:
http://netdude.much-list.com/twtelecom-sprint-home.gif
http://netdude.much-list.com/twtelecom-uunet-level3.gif

peering:
http://netdude.much-list.com/ardent-home.gif
http://netdude.much-list.com/ardent-level3.gif

notice how in the peering, data goes from one network directly onto the other (ardent-home.gif and ardent-level3.gif)... whereas in the transit, it goes through a intermediary carrier... (sprint and uunet in these cases, respectively)... now... why in the hell would sprint/uunet be giving away free access to other networks it has access to!? they pay good $$$ in maintaining those feeds to those other carriers, and you just expect them to give it away!? hence: TRANSIT

[Tommy]

Great discussion but no name calling please. Let's keep it grown up :)

[thecoolh]

awwww... fine *slaps own hand*

:D

but i mean... i gave two clear examples in my previous post, yet he totally missed 'em... LOL

[Tommy]

I know I know. I think you are both going to have to agree to disagree because it's quite clear no one is prepared to give way on any of the points.

Still, it's been a usefull discussion so carry on......

:)

[thecoolh]

but I provide solid examples (along with traceroutes to show 'em) on atleast one of my key points (tier-1s buying transit)... something he challenged me to do... something he totally disagreed in...

but yeah... he's been around the block, i've been around the block... but two people can still see the same painting and see two completely different depictions... :)

but then again... networking is more of a pure science... hmm...

[hrmedclinic]

Holly cow! It feels like i just read the book :) and all i did is missed couple days reading this post :)

Well first: its not down - there is no website - i'm still working on it. This is my Dad's MedRehab - and there is about month and a half to the due date.

Fighting guys: I agree with both of you. Thing is it all could work perfectly in both case scenarios, and all depends on your prefferences and capacities.
Of course smaller hosting companies can not have a XXX million dollar datacenter and pay hundred thousand dollars just for the backbone. BUT, they can put their servers on a facility like that and get EXACTLY same connection, or get few servers mirroring each other from same or different places and that will get you even more reliable services.

Buttom line is : this forum is for discussion and finding out information - not to call each other "stupid".

[thecoolh]

aww... come on dude... me n ljames are just havin' some fun... :D n hopefully educatin' visitors along the way...

hard to come across ppl that are knowledgeable enough to be able to challenge us... heh... :)

[ljames]

maybe its the way I am typing that is not getting it accross. Also, we are talking about internet backbones, not colos right?

A Tier 1 Internet provider means that they peer (communicate) directly with all other networks on the Internet. This is important because latency (the time it takes your information to reach its destination) is dramatically lower with Tier 1 ISPs.

With a Tier 2 provider, your data must first traverse that ISP's network, and then cross its "parent" provider's network before it gets to the Internet. A second concern is that a Tier 2 provider will typically have only one connection to its "parent". Often, this link can become congested, increasing latency. Additionally, if that one link goes down, your connection goes down as well. This is why colos buy from mulitple providers.

Some small providers bypass using a "parent", and connect directly to small regional NAPs (Network Access Points). These regional NAPs are almost exclusively populated with other small, Tier 2 providers and suffer from the same resource allocation problems. Subsequently, if your destination is not on one of these networks, there is no guarantee that your traffic will be always be efficiently routed to its destination.

Granted I will admit this is older school thinking. Tier 1s do not buy transit. Why should they when they peer with all other tier 1??? I think the Internet has advanced so much that large companies/national guys are starting to call themselves tier ones. The fine line between the two is starting to gray.

Cogent is a fine example of this. They hardly peer with anyone. With the netrail buy and BW growth, peering has got a lot better. Would this be called a tier 1???

According to AT&T, UUnet, Sprint, etc.. NO. According to your definition - Yes.

Regarding the "stupid" thing - how old are you? I am guessing 17. Am I close? This is an honest guess.

[obfuscated]

Some things I found amusing...
http://www.rackshack.net/aboutus/tos.asp

(1) "All Payments To RackShack (Everyones Internet) Are Non-Refundable"
:idea: Good way to attract new customers.
(2) "Subscriber is aware that the Company may prospectively change the specified rates and charges from time to time." "contingent upon Company achieving and maintaining its cost of service goals"
** That's why they don't offer long-term contracts.
(3) "100% of the setup fee is subject to the 8.25% tax rate"
:confused: I've never heard of SERVICES being taxed! I've never seen any other hosting company charge tax for Setup Fees!!
(4) "If you dispute a charge" ... "you agree to pay RackShack an "Administrative Fee" of not less than $50 and not more than $150."
** Great customer service ethics.
(5) This one is my FAVORITE :rolleyes:
"We limit our technical support to our area of expertise."
:group: That's like a car dealership stating they will only repair what their technicians understand.
(6) They DO NOT BACKUP your data. You are on your own. In the event of a server crash, please place your head between your legs and ... well, you know the drill.

[thecoolh]

ljames... name a tier-1 network then... that "peers" with every other network in the world. while your at it: define "network"... just curious how it'd be possible to connect to every single one on the planet... there aren't that many global providers out there...

btw. i'm 19... :) can't gather that from the profile information? lol

[ljames]

ljames... name a tier-1 network then... that "peers" with every other network in the world.

Not everyone tier 1 peers will "all" other tier 1s. I Shouldn't use words like that. I hate it when I do that.


AT&T, CW, UUnet (MCI), Sprint, Genuity (GTE => BBNPlanet - the first backbone)- These are part of the backbone of the Internet. They laid the pipes, they sell transit. They have established the "peering relationships" with the other tier 1s in order to "share" bandwidth and increase their mutual reliability. If you need to buy transit - you are not a tier 1. At least by my definition.

Too bad there isn't anyone else out there who can stop by and post about this. :)

[thecoolh]

UUNet does NOT peer with AT&T... they're buying transit from FNSI to get onto AT&T... so you can remove UUNet (MCI WorldCom) from the tier-1 list... k? and since UUNet ain't peered with AT&T, AT&T ain't peered with UUNet... so there goes AT&T... hmm... any more removals to add? and if they ain't buying transit, and they aren't directly connected (peered)... how is data getting from one network to another? because I sure as hell don't think FNSI is giving the bandwidth away for free...

and I don't think that anybody that knows what they're talking about would argue that UUNet and AT&T aren't tier-1...

[obfuscated]

With all due respect, I think you gentlemen would have better exposure if you restarted (or moved) your conversation to a new thread. I doubt many would find it by chance since it's listed under "rackshack".

[thecoolh]

thanks dude... good suggestion... i'll email one of 'em moderators 'boutz it... :)

[Tommy]

The original thread is in it's full glory here but under the suggested title by thecoolh.

Enjoy!

[ljames]

and if they ain't buying transit, and they aren't directly connected (peered)... how is data getting from one network to another?

OK, hold on. You seem to be getting a little too tense, but here it goes.... they peer with someone who does hop to that network! Oh my god.... That is new??? Wow??

AT&T ==> lets say Sprint ==> UUnet. "Oh, that is just crazy, impossible. I can't believe it. But my cute images I took showing tracerts confirm my theory. I can see it in a graphic that was dispalyed to me. It must be true......"

Bottom line. Believe it if want, I know no longer care as you have turned this from a discussion, to instead a 19 year acting like a newbie insult contest. I guess all the years experience you gathered with tier 1 providers once you left high school as far surpassed my experience. Here it goes -

Tier 1 = pier
Tier 2 = transit

It is like this with you. "Hey, vanilla taste like chocolate" - That is impossible - "But I ate some white ice cream and it tasted like chocalate" - Could it be white chocolate? - "No, I read it in a magizine vanilla is white, it has to be true". - OK, good luck.

I now leave hopeing someone smarter then both of us can come by here and clear it up. My work with tier 1 backbone providers has not been enough.

Good luck with coolwebhost.

[thecoolh]

dude... i ain't pullin' my facts out of thin air, i've confirmed my standpoint with two sr admins at ardent and a friend in network ops at uunet canada and two other friends in network ops at primus telecom canada. (i was bored, lol)

go ask someone who knows what they're talking about for a second opinion. every time I state some facts, it seems your reply posts try and do everything possible to "ignore" it... why? I address every issue you bring up immediately and directly.

[thecoolh]

ljames...

olm (http://www.olm.net/networkOverview.html) claims they're tier-1... u can confirm with me that they are not? cuz i just hate hosts that claim stuff like that with absolutely no truth to it... especially when they're big... hmm...

also venturesonline.com (http://www.venturesonline.com/hosting/colocation.html) ... hmm...

eh?

[ljames]

OLM is not saying they are a tier 1 provider. They are saying they *connect* through multiple tier 1 providers. They have no middle tiers.

If you look back a few post ago you will see I thought this is where we were getting mixed up. A tier one colo/DC is different then a tier 1 backbone provider.

VO is an old Verio colo host. They just recentally moved to their own datacenter. Again, they are just saying they have direct connections.

http://www.venturesonline.com/network/network.html

[s_mack]

I just read through every post.

I now know even less about tier-1 then i did before (which was nothing)

I still don't know if RackShack sucks.

I 100% agree with the ORIGINAL post, that faulty equipment should be the responsibility of RS in this case... no doubt! and I don't care what the price is.

"Buyer beware", my friends, only goes so far. If a company promises dedicated hosting, there is a certain level of expectation that the equipment being provided is their responsibility... EVEN IF THEY CHARGE $1/yr! They promise it, they deliver. They don't deliver... well that gives you the right to at LEAST come onto a forum such as this and say they suck.

It reminds me of the last host I was with... package said 10GB space and 16GB traffic for a paltry $75/year. When I approached 2GB space and a dozen GB traffic, they shut me down because they didn't expect anyone to use it. They were a local company and I took the to court... i lost, but only because I refused the refund they offered (it was more fun to take them to court).

I'm not a computer genius... far from it. I do hold a degree in marketing, and I serve on the Advertising Standards Council here. Businesses have to stop promising the world and delivering nothing.

Why am I making such a long, boring post?? BECAUSE YOU GUYS MADE ME READ THROUGH 7 NOVELS TO GET HERE!!!

:) totally kidding.

Very informative posts guys.