Why Didn’t Nortel Do Better ? Cisco Wasn’t Always the Top Dog.

In response to Omar Sultan at Cisco on ‘Why you want this switch ?’ . In my view, Cisco IOS was buggy,slow and the hard­ware product was a poor design, but Nortel got the usab­il­ity and tech­nical sup­port very wrong. Customers chose Cisco any­way because the Cisco TAC made the prob­lems not seem so bad.

There were a num­ber of things about the Nortel BCN that were ahead of their time. And a num­ber of reas­ons why I think the mar­ket didn’t take to the product over­all. I used to work for a com­pany that sold many vendors of equip­ment in the late nineties and early noughts and remem­ber how the engin­eer­ing team stead­ily moved away from Nortel /​ Bay /​ Wellfleet to Cisco.

Multiprocessor OS with Modular Kernel.

When you bought a Nortel BCN you got a back­plane, and lim­ited pro­cessing power. Every blade that you installed then added CPU pro­cessing to the chassis. The oper­a­tion sys­tem dis­trib­uted itself across all the CPU’s. The sys­tem grew in per­form­ance as you added inter­face /​ line cards. Each line card had a num­ber of CPU’s that were needed to pro­cess. BayRS was also mod­u­lar. You could actu­ally take pieces out of the image that you did not use. Cisco has only imple­men­ted ker­nel mod­ules in the last few years

This meant that I spent just as little cap­ital as pos­sible to get what I needed. The Cisco approach (by com­par­ison) means that I buy a mon­ster rout­ing /​ pro­cessing /​ super­visor engine and then hope that I will need all that per­form­ance. Thus all the cap­ital is required up front. I always felt that this was a good model for Cisco, (and not so good for the cus­tomer) since Cisco gets all the rev­enue early in the life­cycle. This also has the effect of encour­aging cus­tom­ers to buy more than they need (Aside: shout out to all those people with Cisco-​​standard Power over Ethernet — just wanted to say: I told you so).

Reliability

The obverse of an SMP approach that Cisco IOS used was always bound by the cent­ral CPU. As a mono­lithic OS you could never increase the per­form­ance by adding CPU’s later. And the search for bet­ter per­form­ance led to new CPU’s, hard­ware archi­tec­ture and crufty soft­ware hacks (such as fast switch­ing) and spawned a vast amount of soft­ware releases, and a lot of bugs until Cisco improved their test­ing in ²⁰⁰³⁄₂₀₀₄. BayRS was much more reli­able than IOS. And I mean a lot.

I remem­ber that it was com­mon prac­tice to upgrade the IOS as soon as some prob­lem was encountered, and would often fix the prob­lem. (Note: this is not the case in the last few years where Cisco has fixed the testing).

Long life­cycle.

The BCN was a viable product for nearly fif­teen years.This was also true of the ASN router, which could be stack to provide 50K pps to 200K pps in the final con­fig­ur­a­tion. At that time an sim­ilar class Cisco product had a life­cycle of less than five years before you had to fork­lift it out. Almost for­got, the Nortel mod­ules were com­pat­ible through all that time. (Unlike Cisco mod­ules which were dif­fer­ent for every chassis at the time).

Site Manager

Lovingly described by engin­eers as Site Damager or Site Mangler. This GUI con­sole used SNMP to con­fig­ure all areas of the sys­tem in a nice graph­ical inter­face. One hand the GUI made con­fig­ur­a­tion bet­ter and you could more clearly under­stand what your choices were. Compare with Cisco IOS where the defaults were often invisible.

Site Manager had to match your BayRS ver­sion because of the SNMP depend­ence. In some cases exactly match. This meant installing many ver­sions of Site Manager on your laptop. Sadly, MS Windows was not up to the task (over­lap­ping DLL’s, drivers etc) and it was an engin­eers night­mare to turn up at a Nortel site at 4 o’clock in the morn­ing with the wrong Site Manager on his laptop.

They also had a retarded num­ber­ing scheme for both BayRS and Site Manager — some­thing like SM V5.1 was for BayRS V9.3 and so on. There was no easy way of know­ing what was the right ver­sion. Queue more angst and stress.

Oh yes, how could I for­get that Site Manager crashed a lot on Windows. Which was prob­lem when you mak­ing a crit­ical change on a router and it snarfed. Unhappy cus­tomer. Unhappy engin­eer. It was bet­ter on Unix though (but we didn’t have Unix laptops in those days.)

Awful Diagnostics

Its true that BayRS had truly atro­cious dia­gnostic and debug­ging tools. I never got over that.

Nortel Support

Nortel Tech Support was good in the early days but ran down to laugh­able. When com­pared to the Cisco TAC, you just would not choose Nortel.

US cent­ric

Nortel was always focussed on the US mar­ket. Getting Nortel to focus on cus­tom­ers out­side the US was pain­ful. Stock would often be diver­ted to US without warn­ing. Since I was always ‘rest of world’ it was just another negative.

Its all about the Operation

You will notice I have not noted any neg­at­ives about the Nortel hard­ware. That is because, pound for pound, Nortel always stomped all over Cisco for per­form­ance, cap­ab­il­ity and cost. Where Nortel lost out was the lack of focus on usab­il­ity and oper­a­tional excel­lence but mostly the tech sup­port. Cisco had a big edge with fea­tures, I think because the mono­lithic OS was easier to develop for.

It was easy to recom­mend Cisco when you have just had a tough night unwind­ing a Nortel net­work using a second rate inter­face and tool­set after Site Mangler cor­rup­ted the con­fig­ur­a­tion. You don’t for­get the bad times eas­ily. Using the Cisco IOS CLI was bet­ter than Site Manager. So when the cus­tomer asked the engin­eer what he recom­mends, the word was Cisco. When the engin­eer pro­gressed to be in pre-​​sales he said the same thing.

Cost doesn’t always matter

One of the les­sons I learned is that cost doesn’t always mat­ter. Cisco had a more expens­ive product that was more unre­li­able and slower than the com­pet­itor. Cisco over­came these prob­lems with bet­ter tech­nical sup­port both in terms of dia­gnostics on the box, and in terms of the TAC.

Someday, it might be inter­est­ing to con­sider that the qual­ity and cap­ab­il­ity of the mod­ern TAC is a busi­ness response to fact that their product needed a lot of sup­port. The world class cap­ab­il­ity might have been a neces­sity to survive.

Conclusion

None of this is really rel­ev­ant to today. Cisco has picked up their test­ing and product devel­op­ment so that IOS is not usu­ally buggy or flawed. I stopped work­ing on Nortel kit a few years ago so I can’t make com­ments about whether they have improved. But if Nortel was doing a good job then I think com­pan­ies like Extreme and Foundry would not exist.

I recently worked on Alteon gear, and was sur­prised at how little the product had developed. Alteon was the pre-​​eminent mar­ket leader to Arrowpoint when they were both by Nortel and Cisco respect­ively. Cisco has developed load bal­an­cing extens­ively but the Alteon seems unchanged from five or more years ago.

Who knows if Nortel can come back, but it wise to remem­ber the les­sons from yes­teryear before con­sid­er­ing Nortel again.

Please rate this post:

(No Ratings Yet)

 Loading ...