Networking crazyness

[RedRed]

I just got to tell you all this.....

One of my sub-offices has a fiber link running to it. The IT staff have almost EVERYTHING going down the 'pipe'....

For the past 3 weeks we have had poor and intermittent performance to this location (about 6 staff based here, AND its a hub for 6 other sites.... perhaps 30 PC's and 50 phones (VoIP), 80 staff....

We had NTL in (obviously), and had to implement a full disaster contingency plan for comms in the whole node.... (MEGA disruption!)

NTL (the 'fibre-Provider') worked to replace the routers, hubs, even some of the electrical wiring going to the comms room. No Joy. The system would work some times, but spend hours with bugger all happening..... Mail was up, then down.... transfer rates varied wildly when it was working....

THEN
A 'spark' (electrician) who was in servicing the wiring to the building, accidently disconnected the main fibre link to the building..... He blew over the end and re-inserted the cable..... BINGO!

Worked fine for the last 3 days......


Stupid or what?

RedRed

[Sasq]

Sigh, that should have been the first thing they checked.

(says the ex telco fibre tech)

[RedRed]

Yeah!

the Ironic thing was that it WASNT the comms engineers who fixed it - it was some contracted Spark instead......

D'oh, double D'oh!

RedRed

[Kurt]

duh....knowing how important it is to have good clean ends on fiber... you pay them (NTL) vast sums of money I guess?

[Helevitia]

Whoever your network admin is should have been checking the routers and switches for layer1 errors. Probably tons of alignment errors along with fragmented and undersized packets. Seeing these kind of errors would have pointed to the fiber at some point.

Dave

[RedRed]

Helevitia:
The admins are in IT. I am a development project manager for an IT project, but not actually in the IT department!..... There is some 'politics' round this.... I cant go 'BOO!' about the network my app runs on.... They are SOOOOOOO precious.......

Example: The middle teirs and Oracle server I use to service 200 users (its early days in the project - there will be 1600) we kitted out with gigabit network cards (my recommendation).... the B*st*rds had the link between them throttled to 10 megabits! Took me ages to work out what was wrong - I am 'not allowed' into the 'production' server rooms! It was only by sneaking in that I found out!

RedRed

[MultimediaMan]

If there was a problem on the fiber, as Helvetia said, there was probably a major trunk problem that would have shown up as CRC errors. If you run Cisco gear, it would have been Screaming "Fix Me, Stupid!"

Perhaps a better question is why there was only ONE Fiber connection from your closet to the Switch or the hub...another fiber card (or a single 4 port Fibercard) would have left an alternate route for the traffic to go over. Don't let your IT staff say they would need to run another fiber, either...most indoor fiber cables have 3 Pairs of Fibers, and only two are used.

[Kurt]

Originally posted by ex RedRed
Helevitia:
The admins are in IT. I am a development project manager for an IT project, but not actually in the IT department!..... There is some 'politics' round this.... I cant go 'BOO!' about the network my app runs on.... They are SOOOOOOO precious.......

Example: The middle teirs and Oracle server I use to service 200 users (its early days in the project - there will be 1600) we kitted out with gigabit network cards (my recommendation).... the B*st*rds had the link between them throttled to 10 megabits! Took me ages to work out what was wrong - I am 'not allowed' into the 'production' server rooms! It was only by sneaking in that I found out!

RedRed

They probably considered you were just one single PC user

What's the point of using gigabit anyway? Can you manage that much data (~200Gbit/s)?

[RedRed]

sorry Kurt, I wasnt being clear.....

I have 2 (comissioning a third) middle teir application servers (meant) to be connected to an oracle Unix server... Using a CORBA brokering system... (Its DCOM at the moment, the developers plan to move)..... I have my middle teirs talking to the Unix box through a gigabit (dedicated) line - the application servers also talk to our intranet through 100 Mbit cards.....

I will have perhaps 10 middle teir servers running by the end of the project (8 serving clients, 1 for external links to third party apps, 1 for data sharing project with some regional hospitals.... these will connect (eventually) to 2 IBM 630's (or a 650, if my budget can swing it!)

Regards
RedRed

[Kurt]

aok

noob question: what's faster 1 650 or 2 630? do you get more diminishing returns with 8 chips or with 2x4? is the network link between the two the bottleneck? software?