Question about implementing VOIP

Lots more good info here. Allow me to take my own "cut to the chase"

Yes, this is true, especially in large shops where the FLU (forklift upgrade) is not a viable option. Bandwidth permitting, cross country IP trunking, as we are doing today with Mitel 3300's running between Houston, Boston, Nashville and Philadelphia, is another of these unique VOIP features that can provide rapid ROI, often in less than a year (eliminating multiple long-haul voice T1s by leveraging underutilized data bandwidth). It's also a way-cool way to cheaply establish remote extensions literally anywhere at a moment's notice, i.e., Houston sales agents working from offices in Boston but with Houston D-I-D numbers. We of course can do it from home too, but our Sr. mgmt still has misgivings about "maintaining control" over the workforce, so there won't be any telecommuting for a while.

I honestly think had the Cisco VAR proposed this (IP trunking) to us, instead of attempting to hollow out the account and start us down the road toward *their* goal of an eventual total FLU, we might have been more receptive. As it was, their concept of migration to a totally PC-based server farm was an instant turn-off complete with red flags, sirens and lights! I honestly think we'd consider using Semifore between the rooftops before we'd consider putting the corporate office phone system on any PC platform, whether Microsoft or Linux.

I've mentioned Linux twice now in recent postings because it has been brought to my attention that Cisco is considering migrating their Call Manager product to Linux as a way of addressing customer concerns with the historically unstable and unreliable Microsoft OS's. While Linux would be an improvement in that area, it still doesn't address the frailties inherent with most PC hardware in general. True there are "industrial strength" PCs designed for continuous commercial service, but those machines are expensive. If I'm going to spend that much gelt on PC hardware (that will in all likelihood still become obsolete in 3-4 years), I'm going to buy a full blown PBX, not some damned kit of components and software applications. The server farm concept isn't going to fly. That dog won't hunt.

Reply to
wdg
Loading thread data ...

in practise i try to use Diffserv marking (layer 3 IP)- that way you dont have to run over 802.1Q to preserve the QoS - but you need modern LAN hardware to make that possible.

It is more complicated, but if the kit you have supports it, it means that you dont have to worry about applying QoS on each hop across layer 3 switches and routers - that tends to reduce problems as the LAN gets altered.

Of course that's going to cost

/ Bias warning here - i work for a telco that uses cisco kit a lot, and i used to work for a cisco / nortel / ibm etc reseller /

this may be a sales channel difference - here in the UK Cisco doesnt sell direct, but only through resellers.

And the reseller training does stress get the environment right or Voip will not be reliable.

When it

i was just trying to make a point here - buying a bunch of cheap switches for voip only may be massively cheaper than upgrading an existing LAN to be QoS capable, and then configuring it correctly and keeping it that way.

not sure that is always true that converging the networks is a goal - it is just a tool to alter the cost model.

often Voip is actually put in for some of the fringe benefits rather than the fact that it is converged, such as phone logon and user migration / hot desking - and all of that stuff works whether the underlying data network is combined with voip or not.

i design this stuff for a living on WANs, and sometimes LANs. But there are lots of choices here, and many ways to get tripped up - e.g. you have suggested 802.1p - but that has limitations, since i dont plan on building layer 2 only data networks, and layer 2 QoS doesnt survive crossing a router.

And most manufacturers dont cover all the cases - e.g. how to set up layer 3 switches, how this stuff reacts to firewalls and address translation, WRED, MPLS and so on.

manufacturers

Default in Europe is that most big networks are multivendor - cisco is almost always part of it, but "pure" single manufacturer environments are the exception rather than the rule.

there are alternatives - you can run a chopped down version of Call Manager on a Cisco IOS router. Good for 100 phones or so.

The last thing on God's green Earth I want is my

Reply to
stephen

I've attempted to hedge that a little. True, almost 30 years in legacy telecom, but this old dinosaur is far from dead yet and he can also learn new tricks. In the past 18 months I've achieved certifications on two Mitel VOIP systems, the MN3300 as well as advanced certification on the SX-200 ICP. If I can do it, others can, but you've got to want to. It does no good to try and stand your ground with only a TDM background. If anyone with only a legacy TDM telecom background plans on staying employed much longer, they better get humping on broadening their skill set.

Of course, what else would you expect? Server farms are *their* bread and butter, their job security if you will. The more the better. The point is you've got to be able to get an audience with corporate superiors, not IT superiors to make the argument.

Like a pool of piranhas. Yes, I know. Cisco feels very much assured of winning because of their often not-so-veiled claim to ownership of the finish line. He who controls the "network" controls that which flows through it and connects to it.

Sort of like shoving warm butter up a wildcat's ass with your thumb; You're apt to get scratched up. But it's important that you prepare yourself (knowledge-wise) to face that battle. If we all roll over and go belly-up without a fight then we'll lose the right to say, "I told you so" when the fit hits the shan and customers (mainly the Admin Assistants) realize that they've lost some valuable features on their phones.

Next time you see one of the Cisco boys ask them if they've yet figured out how to deal with an old key system feature called "Common Ringer" and how they perform that stunt and extend it out to a Claxon Horn or a Yard Whistle. (i.e., I have 3 incoming lines, two internal tielines, and one internal DN for my voice mail zero point. I don't want a rollover, what I want is all 6 individual lines to blow the same loud yard whistle using a single pair of wires when any one (or more) of them is ringing so the guard at the gate, with a single line phone, can dial a pickup code and retrieve any 1 of the ringing lines and have the option of putting it into a call park orbit and take another call)

By the way, does anyone know if Cisco ever got plain vanilla MCR working on their sets? (MCR = Multiline common key appearance ringing, also used by Admins) - Word on the street is Cisco lost a 230-site national account bid to Mitel last year (2003) over that little feature. Cisco got the network side but lost the voice side. So much for claiming ownership of the finish line, huh?

Reply to
wdg

You do NOT need IPv6 for QOS within the same subnet.

Not so long ago those old PCs running Novell for years on end were also built in the USA using 1st quality components. Today's iron (now even the IBMs) are coming from China. Say what you will, I don't see that the quality is what it once was.

The issue with investing the bucks in a high dollar 'continuous duty rated' PC is that in the final analysis it's still a PC. Its hardware platform will be obsolete by the next presidential campaign. In our organization we LEASE the PCs and Servers partly for that reason. I do not want to see us get into a cycle of having to reengineer the damned phone system every 3 years when the two prior platforms have lasted for 14 and then 10 years, respectively.

For the sake of comparison look how quickly Cisco slaps "end of life" labels on their routers and switches. Look how often they release IOS patches. Look how many times you've had to buy more ram (at usurious prices) just to keep something going. Do you really want only a 90-day warranty on your phone system and then be forced into buying Smartnet contracts on every single component?

No thanks.

Reply to
wdg

I know (very little but) of only two ways to do QoS: one is to have the TA for a home/office system be between the broadband modem and the in-house router so the TA can literally throttle down the other data. Second is to have IPV6 that lets priorization be set.

Reply to
Rick Merrill

That 10 (or 20) years of reliability is interesting and implies other costs, like training costs.

Training costs (and the hidden cost of no/improper training) for the IT/Telephony staff are pretty high. Do your customers want to repeat them every 3 to 5 years? Of course, if you buy replacement gear from the same manufacturer, your training costs will be less (but the vendors will know that and may charge more for the new systems.)

But (IMO) the real expense is user training. Why buy a new PBX with more features unless your users are going to use those features? And how many users can do more than the "big four" functions--Hold, Conference, Drop and Transfer? (How many can even do those correctly?)

Reply to
Hank Karl

We're on the same page, but I'm not so sure about this comment. I'm guessing you're from a deep telephony background. But guys like you and me are getting left behind and/or simply weeded out. I've worked in national sales, mostly in the vertical market of higher ed, and I can tell you unequivocally that I'm working with fewer telephone guys and more IT guys. And IT guys often have no aversion whatsoever to the "server farm" concept. In fact, they are embracing the idea at an alarming rate. Used to be that telephony professionals were consulted (either internally, externally in the the forms of interconnects and consultants, etc) almost exclusively on telephony buying decisions. However, this is becoming less common, or at least the introduction of the IT professionals to the mix is becoming much, much more common. And it's here that Cisco looks poised to rule the world. I cannot imagine the legacy products like Mitel (excellent product), Nortel (excellent product, which has suffered a little), NEC (excellent product), Avaya (superb, over-priced product), Inter-tel (superb product, esp on the -100 lines platforms), Siemens (once the best product, still excellent) keeping pace with the marketing juggernaut that is Cisco. They simply are too well-embedded in the professional IT world, and businesses are no longer employing telephony staff to counter-balance the situation. I've been there - many, many times. I've sat with these people and told them the truth. I've presented compelling arguments for a hybrid solution based on architectures that have been performing solidly for decades. It's an uphill battle.

jakesnake

Reply to
jakesnake66

thing to remember is all QoS affects is which packet do you send 1st when there are 2 or more in a Q on a bit of kit.

If the network is lightly enough loaded that multi packet Qs dont happen (ie. say 100 IP phones on an 100 Mbps ethernet - less than 10% load and nice stready streams of traffic) - then QoS doesnt make any difference.

Or looking at it the alternate way - if the Qs stay small, then the amount of latency / jitter induced is within the acceptable range, and again QoS isnt needed.

Now - add in data flows, with very lots of packets, mix of frame sizes, spiked peaks of traffic, and QoS is more likely to be an issue.

run-of-the-mill

No - all you need is to raise the priority of the voip traffic, and leave all other types as default. It usually makes sense to pick out the exceptions (voip in this case) when setting QoS up - fewer and simpler filter definitions.

Reply to
stephen

Perhaps you missed the part about where I mentioned our Servers are LEASED meaning that they are automatically replaced, as a matter of policy, every

3 years. I have no choice nor vote in the matter. If I put my VOIP phone system on a PC I can look forward to a major service-affecting outage of indeterminate length every 3 years while we rebuild it from scratch. On a Microsoft platform I can also look forward to other service-affecting outages every few weeks when the latest malicious code patch of the day gets applied and the system has to be restarted. Folks, listen to me please, this is absurd! You don't do this with your phone system.

With Mitel at least the hardware side is of mature and stable design and never once have we had to add more memory or pay usurious charges to update the operating system or apply patches) or subscribe to a service contract for each piece of equipment. We also have never had to increase memory in order to add new features. Sure, the PBX system could be attacked, but it's far-far less likely to be if only by virtue of the fact that it is NOT running on a Windows/Linux/Intel platform. This is common sense, not rocket science. Once you figure out what the primary targets are (PCs running Windows & Linux) anyone who puts a critical application on that platform, directly "in the line of fire" deserves the result.

Unlike a lot of shops, our Mitels are all -48v DC powered with 8-hour battery banks backed up by dual/redundant diesel generators with ample fuel to run for several days. Even the new MN3300s, which are AC only, are powered by a large UPS which itself has a 48v battery system and similarly backed up. The picture I'm trying to paint here is that my phone system has simply GOT TO WORK and can *NOT EVER* be allowed to take an unscheduled dump. Our phones really are that critical, so much so that we even have redundant Local exchange carriers and redundant LD carriers, all fed into the building on separate SONET rings into separate dual/redundant Muxes which also are DC powered and tied to our battery plant. Telephone outages may be commonplace in your business but sure not in ours. We have not had an "unscheduled" service interruption in more than 8 years and I have the system logs to back that statement up.

Reply to
wdg

In article , stephen writes

Makes sense. - You do not need IPV6 for QoS, or am I missing something?

I agree of the Microsoft OS. BUT, why not a 'PC' as the platform? Not so very long ago, 'PC's' were running Novell, and were left gathering dust, grinding away in a corner for years. Many times not being rebooted year on year. - I do not think the hardware platform is the problem.

YMMV, Phil Partridge snipped-for-privacy@pebbleGRIT.demon.co.uk Remove the grit to reply

Reply to
Phil Partridge

I've got a machine that is primarily made of components manufactured in Taiwan, China etc. that has been in continuous operation for 4.5 years now.

So I won't fault commodity components. I will mention that at work our Windows NT 4.0 and Windows 2000 servers need to be rebooted monthly while the RedHat, Debian and SuSe boxes just hum along taking hundreds of thousands of hits per month for our web services.

Crucial.com does list relatively inexpensive DRAM for many Cisco products.

As to buying contracts etc. ever look at the maintenance bills for a Definity G3iV11 with 250 sets? Yikes!

Reply to
Tony P.

In article , wdg@[206.180.145.133] writes

Probably not. But, a decent branded server, used as the base should still run for many years.

Yes, it will be 'obsolete' in PC terms by the time it's delivered. But it will still perform its 'bought for' function. Let's face it, you don't *need* a P4 to write letters.

Even a Mitel, Meridian, What-have-you, will age.. Newer versions of the hardware will appear, but that doesn't mean your 'old' box will no longer work.

I have sites with the smaller Meridians. Every time the mains fails, they 'forget' everything. BT sort it out, then swap the unit because I complain it looses it's config. Next time the mains fails, guess what. - Yes, I know this is due to the incompetence of the maintainer.

I have a site with an Omni. Mains fails, the UPS keeps it up for a time, then it goes down. Mains comes back. It reboots and comes up with the correct config. Brilliant. - It's more-or-less a PC, but runs (presumably) a brand of Unix.

**Caveat** It is a box with a processor, memory etc. That's all I know of it.

Phil Partridge snipped-for-privacy@pebbleGRIT.demon.co.uk Remove the grit to reply

Reply to
Phil Partridge

Yes they do. Unfortunately Cisco will not "certify" secondary market DRAM and using it in your Cisco equipment will invalidate your Smartnet contract(s).

No, but I know firsthand the labor costs to be self-maintained with a network of 6 Mitel SX2000s, and (so far) 4 MN3300s + 4 brand-new SX-200 ICP (not yet deployed) systems with a total of approx 5500 sets. The two (yes, only two) techs that support these systems have many other jobs and duties so the abount of time spent on each task is tracked. I can tell you this; on such a large system it is far more economical to be self maintained. Mitel training is actually one of the more reasonable costs, because once certified on a given platform, future upgrade training is free and thereafter conducted online. That's a definite plus. Also, once you purchase your Mitel PBX software all future software updates, upgrades and subsequent releases are also free (from LW30 Rls 2 forward). The only thing you pay for are new features and only then if you want those new features.

Being self-maintained also helps to accurately identify equipment that

*really is* approaching the end of its service life without having to take the word of someone who has a vested interest in selling new gear.

No question, maintenance contracts are a ripoff, but are often the only way to budget support costs when you can't take care of it yourself. Many vendors are understandably reluctant to provide "good service" on a time & material basis because it makes it difficult for them to budget their staffing levels. It's kind of a Catch-22.

Reply to
wdg

I pretty much agree with everything you've stated.

I was once in a situation where we had a G3i and another agency got sick of being abused by Avaya and ripped out their G3i and all the 7406D's and 8410D's etc. We rolled that G3i cabinet down the street and had plenty of functional spare parts for our system.

All we kept maintenance on were the power supplies and the CPU. Everything else we could snap in by ourselves.

Reply to
Tony P.

In a clean, temp-stable (68º) environment, operating on clean, steady

-54.0v DC power, our Mitels seem to run forever. I honestly cannot tell you when the last time was that we lost even so much as a single port on a

16-port MC330 DNIC station line card. We're going on 5 years since we last replaced a hard drive and I can't recall having ever lost an MC-IIIE CPU card. Gosh, it almost broke my heart to retire the old black boxes SX2000-SGs in 1996, but we had no choice. They had long-since been M/D and R&D ceased, meaning there were finally no more Software upgrades available. BUT... there was a very attractive hardware upgrade path and MITEL would give you a brand new SX2000-Light redundant main control cabinet (w/dual processors and dual disk drives) free with each old "SG" retired. We were actually able to reuse 80% of our existing circuit cards and 100% of the existing instruments. From a financial perspective the equipment upgrade was almost painless. Staying with a brand you're happy with when it comes time to upgrade is always the least cost approach. These companies who yank out an M1 to put in a Siemens then yank that to put in an Eads must either have an unlimited supply of money or else sand for brains. Every time you do something that drastic you have to replace 100% of your proprietary instruments. That's crazy. Nortel has tried to get their foot in our door for years, but they never will because the cost of the different brand PBX is just the tip of the iceberg. Hell, they could give us the switch for free and it'd still cost us over a million bucks to replace the phones.

This is essentially the same crazy ideat the VAR trying to push the Cisco stuff onto us was hinting at. No, not all at once, but over a period of

2-3 years. That's still crazy.
Reply to
wdg

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.