Question about T1/E1 connections

[Scottf66]

So I am new to the networking field, and still in school. However i do have an internship at a very large company, and we supply bonded t1/e1 setups to our outside locations. If memory serves, a t1 is 1534 megB or something like that, but bonded gives you essentially 3mb.

Knowing this, some co-workers say it's very expensive to purchase a T1, much less bonded T1's. While I never got a price, i see Time Warner selling 100MB for 100 or so bucks for business. Why would you pay for 3mb when you can get exponentially faster?

I understand broadband is a shared media, and you may not actually get 100 meg, but even if you got 50 meg, thats still 15x faster. Not to mention the modem from TW is probably much cheaper then a router from Cisco.

I am sure I am missing something here, but I was just curious.

 

[Fardringle]

When reliability is more important than speed, as it is for business critical systems, you get a T1 (or an equivalent faster connection). T1 typically has 4 hour or less guaranteed FIXED time for any outage, and usually better than 99% guaranteed uptime.

A 100MB connection from Time Warner is nice, but when it goes down (not if) it could be days before it's fixed again depending on the scope of the problem, and while that is annoying for a home user, it won't put them out out business. A business critical system down for days could result in a huge amount of lost revenue.

 

[Gryz]

T1 and E1 (the equivalent outside the US) are part of old technology. Very old technology. And expensive technology. Very very expensive technology.

T1 is part of Sonet. E1 is part of SDH.
What Sonet and SDH do is lay out reserved bit-pipes over an underlying telcom network. In the old telcom days, everything was about predictability, accountability and reliability. If it couldn't give you guaranteed 99.999% uptime, it was considered useless. No matter the cost. Result: a network that was slow as hell, only a little more reliable than cheap packet-switched networks (like the Internet on a large scale, or Ethernet on a small scale). And expensive. Really really really expensive. Really expensive.

Telcoms loved it. Lots of complexity and cost. But they could charge the costs to their customers anyway. The Internet killed it.

There are still people who require old fashioned bit-pipes. God knows why. Because someone highup in the customer's organization thinks it is worth the money ? Because the customer has legacy systems that require a pure bitpipe (old IBM stuff). Or because something is really life-critical, and you can't afford hickups that could happen once a year in an Internet based VPN network. And like Fardringle says, T1/E1 connections are probably sold with more guarantees (SLAs) then regular cheap Internet connections.

And I am sure there are customers out there that have old systems that work, and they don't want to touch anything. Or they are clueless, and pay for something they don't need.

I live in the middle of nowhere. (Actually, 1.5 km outside a small village, and 3.5 km outside a little bigger village where the telephone colocation is (with DSLAM). I get 6.5Mbps ADSL for 30 euros per month. If I wanted a E1 connection, I would be charged for the distance where that T1 would terminate. It would be only 2 Mbps. And it would probably cost me hundreds of euros per month. Or more.

It's old technology. Like X.25. At one time it served a purpose. But not anymore. It's good to learn that it existed. But now you can forget about it.

 

[imagoon]

First T1/T3 may be delivered via SONET but they are not required to be. It is not required to be "part of" SONET etc and you will see many T1 lines delivered via other techs. T1 / E1 is still in use because alternative techs just arn't as reliable still. Time Warner / Comcast etc just can't seem to do it for whatever reason. I could dump my reliability logs on any day and the Comcast "business" lines are typically in the 99% a week if I am lucky while the T1s are at 99.995% or more per year.

In reality until something better comes along and is actually available (like ATT PNT as an example) we are stuck with slower lines because we can't have things like our phone systems randomly dropping out on a daily basis like Comcast does.

FYI T1 is normally delivered via SDSL now.

 

[thecoolnessrune]

T1's have very few genuinely beneficial use cases in today's world. Back when T1's were going up against dial-up and ISDNs they made sense, but most of their benefits have eroded tat this point.

If you have a traditional on-premises PBX, then T1's are still a very relevant part of a business. Since T1's can be used as PRIs or as Data channels, they offer a lot of flexibility for customers by allowing them to handle both. Keep in mind this sort of capability was a big deal before the days of VOIP. When you needed both data for your operations, as well as voice numbers for call center and offices, T1's and their T3 big brothers made a lot of sense.

Nowadays, T1's make very little sense for the vast majority of businesses, but the connection type persists mostly due to old network admins who don't / won't know any better, and Telcos doing a good job keeping the marketing hype that T1's are the way to go.

One of the big things that people bring up (and Telcos) is SLA, but all SLA does most of the time is protect the Telco from you, not the other way around. Most of the time, it includes statements that keep you locked in a contract, and limit the amount of retribution you can receive when a circuit goes down. It basically excuses a Telco to provide bad service.

The other thing people bring up anymore is that Cable is "Shared" while T1's are not, which doesn't make any sense because while the last mile of a T1 may be dedicated, so are the first several feet of your Cable connection. Whether a few feet or a couple thousand yards away, eventually you end up in the same massive IP equipment as everyone else. Heck, companies like AT&T provision both their T1's, and their U-Verse off the same exact fiber. Why? Because AT&T knows that they can sell T1's per unit of bandwidth for far more money, and they know some businesses won't accept it if it *isn't* a T1, because those admins haven't continued to follow the industry and still believe that a Tx connection is the end-all be-all of connectivity.

And T1s getting repaired quickly? Goodness no. I've worked in a NOC, and our customers will drop their precious T1's all the time. We have *days* old outages at any one time with AT&T, Verizon, and CenturyLink, all in various states of "We really need to roll a truck, but we just haven't been able to dispatch yet". Level 3 is usually pretty good, but all falls apart when the circuit passes through a LEC (Local Exchange Carrier) that simply refuses to act. The Telco has SLA agreements with the LECs, and they'll pursue compensation (on their end) for the downtime, but that does *nothing* to get your circuit back up.

That being said, I do find that our Cable circuits have outages more often than our Tx circuits, but when we do, the outage is usually resolved within 5 minutes to an hour. In my time in the NOC, I've never had circuit that required Telco intervention that was resolved in under 2 hours. Heck, if it is less than an hour, they won't even find out why, they'll just say the circuit "came clean while testing" and leave it at that.

T1s have very specific advantages such as low jitter and predictable latency that makes them a better experience for VOIP services and certain applications that rely on such predictability.

At this point, if you do not require high uptime (think small outages, that could happen a couple of times a year, most Telco's only rate cable up to 95% uptime, not 99%), then Business class cable is a great option. Most applications can run just fine over it.

If you have high upload-bandwidth needs (greater than 5Mbps), or need low-latency variation for VOIP, or need high uptimes, then instead of Tx lines, consider Metro Ethernet. Metro-E is now less than the cost of multiple bonded Tx lines. While a 3M bonded T1 line may cost $700-800 a month, the same cost can get you a symmetrical 10M Metro-E over Copper.

Either way, unless you need predictable latency to remote sites, and only a small amount of bandwidth, then T1s can still make sense. Otherwise, most of its use cases have been replaced with viable alternatives.

 

[Red Squirrel]

T1s are probably mostly for phones now. With a t1 you have 23-24 lines depending on how it's setup. I think the typical setup is ISDN so you get 23 voice channels and the D channel which is basically for control. PBXes typically get a couple T1s going into them for outside lines. Rail, hydro etc use them a lot too for communication. It is really old and overly expensive though but lot of equipment is designed to use it so it just sticks around.

To me what needs to happen is 100mbps internet needs to just become wide spread, and it needs to be treated like T1's where the SLA is 4 hours etc if you want to pay for that level of service. The problem is a logistical one and not a technical one.

What gets interesting is DS0's. Some customers may not actually need a full T1, so a T1 is split and sold in chunks to multiple customers. An old style setup like this involves a DE-4 channel bank and that's about the size of a small microwave oven... all for 1.5mb.

Now days a compromise is just regular internet with VPN tunnels to your sites. You may not get the same reliability and level of service though, but it all depends on the actual ISP. There's no reason why an ISP can't provide high end SLA on internet, they just choose not to. My ISP's new FTTH service is actually considered fairly high SLA. If you call in middle of the night they'll work to fix it. If there's a major outage impacting more people, techs are called out to fix it.

 

[Scottf66]

Thank you guys for the enlightenment. I was just curious why they would have the expense of slower lines.

 

[drebo]

To me what needs to happen is 100mbps internet needs to just become wide spread, and it needs to be treated like T1's where the SLA is 4 hours etc if you want to pay for that level of service. The problem is a logistical one and not a technical one.

That's just not feasible for DOCSIS.

The only solution that can do that is fiber. Even things like EoDS1 and other EoC solutions don't have the same reliability (though they mask it well due to the overtly redundant nature of the tech.) Fiber isn't viable in a lot of places for a lot of reasons.

The reality is that if you have a business critical path, there has to be carrier redundancy in place. Whether it be via VPN from relatively low cost Internet connections, cellular, whatever.

A lot of people have this misconception about T1s and the SLAs they provide. There is no guaranteed response, even if you buy directly from the ILEC. What the contracts always say is "if it's not up in 4 hours, you get some percentage of a month back." And if you buy from a CLEC, there is no way they can guarantee 4 hours because they have no guarantee from the ILEC. This may not be the case everywhere, but it is the case in what AT&T labels their "22-state agreement."

Doesn't matter if the T1 is delivering MPLS, EoC, DIA, or whatever.

Also, even though T1s are a building block of SONET, they do not need to be delivered as such and don't even need to be backhauled as such anymore. As stated above, nearly all T1s are deployed with DSL tech. From there, they can be terminated directly at the CO (increasingly popular for a variety of reasons) or they can be transported at L2 via a variety of techs (some telcos even still use ATM for it) back to a POP.

Long and short of it is that T1s still have a place in the market today, as defined by business requirements, and it's not always just because people are set in their ways. A T1 gives you predictable and reliable latency and a relatively large increase in reliability as a whole.

 

[imagoon]

Basically what Drebo said. The only other tech that I have found to be decently reliable has been MAN's. Unfortunately that was only available @ 10 of the 31 locations we just lit up last month. So 21 sites are lit up by bonded T1's while the others are ATT fiber Ethernet. It just isn't available even in the middle of Chicago in some places. Mean while in the last couple hours since my last post, I had two 5-10 minute outages notices from the network monitors at 2 of the VPN/Comcast sites....

 

[thecoolnessrune]

The only solution that can do that is fiber. Even things like EoDS1 and other EoC solutions don't have the same reliability (though they mask it well due to the overtly redundant nature of the tech.) Fiber isn't viable in a lot of places for a lot of reasons.

Every EoC under our management, whether our infrastructure, or our customers, has better uptime stats than every single T1 in our management. In fact, I can only remember one time off the top of my head that we had an EoC circuit go down that wasn't part of maintenance or customer issues, and it was due to a mis-handled work order putting the circuit on the wrong VLAN. Even that sort of bureaucratic mess-up is better than our T1 issues that we've seen, like mistakenly decommissioning a customer's circuit, which seems to happen a couple times per year, almost always by AT&T. When that happens, the circuit is down for several days, because once decommissioned they can't bring it back up, they have to provision a whole new circuit.

In fact, the only thing that comes to mind that has had better uptime stats than our EoCs was our OC-48 SONET ring, which we recently replaced with 10G EoF. Only reason it waited until recently was the number of customers who wanted to keep their T1s.