The ten rules of building a small business network

Great article but...

why the piece on disabling NIC auto-negotiation? This is out-dated practise and causes FAR more problems than it solves. If you're going to advise people to do this, at least be sure to mention that auto-negotiation also needs to be disabled on each corresponding switch port, or (with fast ethernet) you get an instant and guaranteed duplex mismatch.

By neil_burton on 11 Aug 2009

You will forgive me if my view differs from yours: you have left out a number of vital pre-requisites, and made one basic error of logic. You're assuming everyone has managed switches; very simply, they don't. In a network of unmanaged gigabit switches, the only way to set the workstations so they share the servers fairly is to force their speed to 100 meg.

And it is simply untrue to say that you "get an instant and guaranteed duplex mismatch" - the problem I am describing comes when the switch & the NIC ALREADY disagree about those settings, because BOTH are autonegotiating, and disagreeing. Taking EITHER one and forcing it to a single speed/duplex setting allows the other end to successfully complete autonegotiation.

I am a little saddned at your dogmatic tone; I agree that it would be nice if everyone understood the benefits of a) matching kit and b) managed switches - but that's not the market this article is aimed at. Also, I think you need to be a little more careful with your criticisms when you pack rather more howlers per sentence into a short reply, than I managed while squeezing that piece of advice into the space available!

By Steve_Cassidy on 12 Aug 2009

Steve, I am not making any such assumption. Forcing NIC speed/duplex is even more likely to cause a headache with an unmanaged switch because the switch port will be hard set to autonegotionate and cannot be changed. Autonegotiation HAS to be enabled at both ends of the link, or not at all. If you force a NIC connected to an auto switch port, you are more-or-less guaranteed to end up with a speed and/or duplex mismatch. You mention gigabit switches - autonegotiation is defined and MANDATORY as part of the IEEE specification for gigabit ethernet. Many NICs and switches simply do not allow you to force speed/duplex and this contravenes the proper standard. If your SME readers are having to force speed/duplex to get around network problems then the best advise is to BIN the switch and PC NICS and get ones which adhere to IEEE802.3ab which, incidentally, was defined 10 years ago. Autonegotation compatibility problems are a relic of the 1990's yet the networking world still sticks to this bad and in your words dogmatic advice firmly. If anyone doubts this please read this excellent article published by Sun Micrososystems in 2004 http://www.sun.com/blueprints/0704/817-7526.pdf

By neil_burton on 12 Aug 2009

Steve, I am not making any such assumption. Forcing NIC speed/duplex is even more likely to cause a headache with an unmanaged switch because the switch port will be hard set to autonegotionate and cannot be changed. Autonegotiation HAS to be enabled at both ends of the link, or not at all. If you force a NIC connected to an auto switch port, you are more-or-less guaranteed to end up with a speed and/or duplex mismatch. You mention gigabit switches - autonegotiation is defined and MANDATORY as part of the IEEE specification for gigabit ethernet. Many NICs and switches simply do not allow you to force speed/duplex and this contravenes the proper standard. If your SME readers are having to force speed/duplex to get around network problems then the best advise is to BIN the switch and PC NICS and get ones which adhere to IEEE802.3ab which, incidentally, was defined 10 years ago. Autonegotation compatibility problems are a relic of the 1990's yet the networking world still sticks to this bad and in your words dogmatic advice firmly. If anyone doubts this please read this excellent article published by Sun Micrososystems in 2004 http://www.sun.com/blueprints/0704/817-7526.pdf

By neil_burton on 12 Aug 2009

I'm pleased to hear there is a standard that covers this topic. I am displeased to discover youthink that this standard represents what people actually do, or can do: I have no trouble doing what I have advised people to do - which is why I advise them to do it. I am perfectly aware of the requirement for 1000/auto as part of Gbit standards; that simply isn't the issue I see when I'm setting up or repairing a small business network.

You are also entirely confused about the reason why NICs don't allow you to force a speed/duplex setting - it's because they re being driven by a minimal driver, normally in the OS rather than specifically delivered by the manufacturer.

Advising people to bin a £200 switch and buy a £600 one, which conforms to a standard that no switch vendor uses as a sales tag one can successfully search on, is not useful advice: and it leaves out those who don't understand the benefit of such a move.

If you want to focus on the source of the knowledge problem in networking, can I suggest you try to figure out why people don't take ten year old advice while purchasing? That might help your cause and also clear up your evident misunderstandings of what's found out there in the field.

By Steve_Cassidy on 12 Aug 2009

Steve, I know perfectly well whats out there in the field, in small business environments just as well as big ones, and I know how often things are far from ideals and standards. However I can't keep count of the number of times I have encountered significant network degradation simply because somebody has done just what you've recommended, and fiddled with NIC speed and duplex settings, without understanding the full implications and the NECESSITY to match settings at both end of the network link. It is indeed illogical, though true nonetheless, that forcing a NIC connecting to an unmanaged switch will nearly always result in a link mismatch and considerable performance problems. If you don't accept this then you are sadly misguided and working to archaic principles which have been outdated for a decade, such is the assumption that autonegotiation often fails and by default should be turned off. Happy to take it offline than argue this out in front of your readers.

By neil_burton on 12 Aug 2009

If someone produces "significant degradation" because they make one change and then fail to test the effect of the change, then that's not a result of bad advice: furthermore, if the advice is, as you persist in baselessly claiming, "outdated", then how come you keep on finding the effect out there in the wild?

You persistently claim that autonegotiation works marvellously on NICs, but that a switch presented with a fixed-speed, fixed-connection device will produce an "instant and guaranteed duplex mismatch". This is a basic error of both logic and observation: if a switch does that, then whatever it is doing, it ain't "autonegotiation". More to the point, there are switches which *can* do it - and have to, since there's a perfectly common group of devices which present fixed speed interfaces. They're called Broadband routers...

I am afraid that you chose this forum to present your rathre peculiar version of reality: I cannot see why this should suddenly be "taken offline" when you are making material and damaging accusations, that you show no sign of retracting, in a dogmatic and illogical fashion. You wanted public notice for your opinions: now you have it. I think it's about time you backed them up with some hard evidence. Provide me with a list of the kit which demonstrates the behaviour you describe.

By Steve_Cassidy on 13 Aug 2009

You are rapidly looking like an argumentative tool...

Steve, considering your an IT journoulist im very surprised by your agressive attitude to a fellow IT professional who is ultimatley trying to stop fellow PC Pro readers from potentially making a mistake which you have told them to do.

You are clearly stuck back in another decade, displayed by some of your attitudes and ideas in your article.

You are also completely contradicting yourself - why would the kind of IT staff whom this article is primarly aimed at be looking to disable auto-negotiation by default and then diagnose any duplex errors which they may not even be aware of? They may also need to call on a contractor with correct network infrastructure knowledge as a result.

Personally, and even more so because of the network heavy applications these days, i would stick with gigabit speeds and autonegotiation for any new installation.

By Sonic2kuk on 13 Aug 2009

Steve, I suggested taking this offline because you are embarrassing yourself. I'll stand by what I've said from my first post. If you take an unmanaged gigabit switch (or virtually any other ethernet switch or broadband router running default settings) and force connecting NICs to 100/full you will very likely cause a duplex mismatch or link failure in almost every instance because the autonegotiating port will fall back to half duplex (hence guaranteed instant mismatch), or for a Gbit switch may fail to link at all. I don't need to list hardware here because the behaviour I am describing is inherent to every IEEE compliant ethernet device and NIC made in the last 10-15 years. I still come across these issues frequently out there in the wild, in 2009, because people just like yourself stick firmly to the mantra that "autonegotiation is bad so disable it where possible". Many of your readers following this particular piece of advice will most likely cause a problem which didn't exist previously. Rather than making a damaging accusation I'm merely calling BS on the dogmatic assertion YOU have made to your readers and suggesting that you revise or remove it. Read the article I referred to previously from Sun which applies to any ethernet link not just those comprising Sun hardware. There's also a good article from Cisco here:

http://www.cisco.com/en/US/products/hw/switches/ps
700/products_tech_note09186a00800a7af0.shtml

and a blogger who says everything I need to say and more here:

http://gdamore.blogspot.com/2008/08/why-you-dont-w
ant-to-force-link-modes.html

and good old Wikipedia, the section on Interoperability Problems sums things up nicely:

http://en.wikipedia.org/wiki/Autonegotiation

By neil_burton on 13 Aug 2009

more suited to a hairdressers

I subscribe to PCPro because of its generally good standard. However, after reading this article (The 10 Rules of building a small business network, by Steve Cassidy) and the follow-up, I am having a rethink. This kind of hysterical outburst would be embarrassing even in a hairdressers.
Here’s a little free advice:
1. If you are wrong, admit it.
2. If you are right, present your argument clearly and concisely. What did Neil Burton do to deserve this kind of personal attack? Did you catch him bonking your boyfriend?
This is the reaction of someone who is embarrassed and threatened. For me right now, your credibility is highly suspect.

By roseste on 13 Aug 2009

Steve, by saying this:

"This is a basic error of both logic and observation: if a switch does that, then whatever it is doing, it ain't "autonegotiation""

you are showing that you really dont understand how Autonegotiation works. You can only have a "negotiation" when both sides are talking about it. If the auto end detects the other end isn't using AN then it can get the speed but it can't get the duplex and will always assume Half, so if the other is set to full you are heading straight for a mismatch.

From a technical standpoint I don't have a huge amount to add to what Neil has already said, and the links he has provided provide all the background info you need.

The reason the trend of turning AN off started was because on the early 802.3u kit the vendors couldn't agree on implementing it properly and in the 95-98 era I would have agreed with you that turning it off was a solution but ONLY where there wasnt the option to run homogeneous kit. Since the IEE tightened the standard up in '98 this really hasnt been a problem.

On modern gear (which is pretty much anything < 10 yrs old) autonegotiation should be the default and you should only be forcing Link settings when problems occur. Otherwise you are just causing yourself problems trying to avoid a non-existant issue.

As regards seeing these issues "in the wild" the last time I had any such issues was between a Dell Powerconnect switch and a Cisco ASA because the tech had forced one end to a fixed 100 Full setup config while the other was Auto, it achieved a link but we had massive packet corruption issues, we set both ends to Auto and hey presto, problem solved.

By tiggertaebo on 13 Aug 2009

That simply does not line up with the experiments I have done with a real world mix of kit, and it doesn't tie up with what happens when one end of a LAN connection has a fixed network speed, either: Nor, finally, does it connect with what I've been told by vendors struggling with user support, who don't come across these issues because they don't test their kit widely enough with other devices. Did someone say "Alc*t*l"? Not me! You are dead right about the age of the kit that can crop up: everyone here is agreeing (however rudely they may express it) that re-investment is often sorely lacking in smaller business networks.

...but you're still coming in from too big a network background. PowerConnects are gorgeous things, but it's an uphill struggle to get small businesses to buy them.

By Steve_Cassidy on 13 Aug 2009

And just to sweep up: Neil,thanks for those links. You are however traducing what I have said, which seems to have been your behaviour from the word go. You believe I have advised people to fix settings to "full duplex". That's not correct. I advised them to *tune* their networks. That's what the copy says. It would be polite (though on this performance, unikely) to hear you acknowledge what was actually written.

Sonic2: In most networks with heavy throughput, gigabit is the starting point. However, the title of the article includes the words "Small network". If I had been asked to write about "heavily laden networks" then the advice would have been different. I do indeed agree, however, that many small businesses run out of knowledge quickly and hopefully have access to all manner of consulting resources - but leaving them completely broken because nothing is changed from the default is equally unforgivable!

Quite a few clients have bought all-gigabit kit and seen their network performance per-PC actually drop from the preceding setup where the workstations were choked back to 100 by an older central switch. The fix is to get a faster server: but then, where should the spending stop?

Roseste: Fortunately, Neil made his point without resorting to homophobia, which is just as well, since it is hardly relevant to the topic. Could you try to keep to the standards you find so praiseworthy, at least? I don't think my appraisal of his contribution was either all that different from his appraisal of mine - or especially personal. Unlike yours.

By Steve_Cassidy on 13 Aug 2009

(actually, one last request while I'm thinking about it. Can you chaps contributing just quickly roll-call how many servers & workstations you look after, either in your biggest clients, or across your portfolios? Bit of willy waving never goes amiss...

By Steve_Cassidy on 13 Aug 2009

Steve, I'm afraid I jumped all over the NIC remark because I literally spend half my working life troubleshooting issues which arise from networking experts who almost religiously peddle this myth that autonegotation doesn't work, and it isn't the first time I've whipped up some controversy in refuting it firmly. The last time I attended a training course (an architect level course by a prominent server-based computing software vendor) we were told to "always hard fix server NICs to 100/1000 full and never use autonegotation". This was written by the vendor in black-and-white, it wasn't just the instructor being a tool, as our friend here Sonic2 would put it. I thought I was going to be lynched by the instructor AND everyone else in the room when I raised my hand and dared to question this "hard fact". There was no advice to consult a networking team first, check switch port settings or to try autonegotiation first, all of which would be sound. A very good example of this problem in the real world was when I had a CFO with a financial database application which was taking 2 minutes to run a commonly-used query. It turns out his PC NIC had been manually set to 100/full and was connected to an unmanaged switch resulting in my infamous "instant and guaranteed mismatch". I simply set it back to auto and the query took less than 2 seconds. A very happy man indeed and I could have probably charged him a months worth of fees for simply reversing the very setting in question. Another customer had an ISA server in a DMZ with a NIC set manually and a switch set to auto, the Outlook Web Access login screen took 2 minutes to load (with various component images missing!) and they were just about to upgrade to a 40Mbps leased line costing many tens of thousands beleiving that to be the problem. I can go on and on with more examples. So hopefully that may explain why I raised my opinion firmly and in a way which may have seemed arrogant. For this I extend my apologies.

I do acknowledge the content of the article and the general market it was aimed at, and I would very happily print the article and hand it to my small business clients (with the offending remark tipp-exxed of course!), as in every other respect you offer good sound advice. This is what PC Pro is all about.

To answer your last question I work regularly with clients who have one or two servers, up to those with four or five hundred. I have encountered this issue in both the smallest and the largest environments, with switches and routers of all types ranging from £80k Cisco core switches to £25 Tottenham Court Road specials.

By neil_burton on 13 Aug 2009

Neil, doesn't that get the job done better than your first comment?

Firstly, full marks for sticking to your guns - but look carefully at what happens in this thread. Look carefully at what's happened with HP and their security announcemet; and then go and look back at what that vendor/trainer were doing. It's no more than about 2% a technical issue - teh remaining 98% is what people do with systems over a lifespan measured in years.

Yes; *if* a lot of pre-existing conditions are met and *if* the right investments have been made, and *if* there is sufficient in-depth local knowledge, and *if* all other factors are equal, *then* the Gbit autonegotiate setting can be used.

*but* some vendors get it badly wrong (Broadcoms still in active use now are an excellent case in point). Vendors themselves struggle with the issue - as you have agreed - and the fix for their struggling isn't to mount a multi-thousand-pound search for a better server platform, it is to find a hack that get the job done.

(I actually chatted with a guy who did do the said multi-thousand-pound search, when he got hit hard by the TCP Offload Engine problem in current Dells. He was at the BBC - his purchasing power and negotiating position bears *no* relation to what small businesses can do in that situation!)

(and I'm sure you don't quite mean it like this but if you have a source for £25 TCR special *managed* switches, I'd like to have some please!)

All your scenarios, by the way, you fixed by conforming to the advice in the copy - you *tuned* their networks. Though you are breaking Rule Two, by trying to apply rules from bigger (means managed) networks, to smaller ones!

The other issue you raise, namely how much one charges for a ten second tick in a box, is even hotter in network technology circles than the one about NIC card settings. You want to see a flame war; ask people what a "fair rate" is when you fix a hundred million pound network. It's a much more productive conversation than listening to people with "hairdresser hangups!"

By Steve_Cassidy on 13 Aug 2009

Steve, I think we can all agree that all networks both small and large can require "tuning" to produce an optimal setup. What I am arguing however is that disabling AN should only done once problems have occurred - and that this should be especially true in small low-cost networks where the level of control is less and the likelyhood of having a tech who understands how to properly tune these settings is small. You correctly said that you didnt explicitly say 100/Full - however if someone is following your advice and disabling AN where possible then imagine the following scenario:

A recent unmanaged switch (ie AN is forced and non-controllable) connected to a NIC where the tech follows your advice and disables AN straight off the bat, the first thing they are going to set it to is likely to be 100/Full. Chances are very high that you will get a Duplex mismatch and the associated poor performance so, the next logical step (following your advice) is to try dropping the NIC to 100/Half. This will work but will give you sub-optimal performance. However if they set to Auto first the chances are good that they will get a negotiated speed of 100/Full and therefore optimal performance.

In answer to your question about what networks I install/look after I would say that I actually spend most of my time dealing with just the sort of networks you describe - typically small companies of 5-25 users with kit varying from £20 Netgear swtiches up to mid-level Cisco/HP/DELL gear. If you are discounting co-location setups then the largest business network I've spend significant time with has approx 100 users. And in the past 5 or 6 years I can count the number of times I've had to force a link to fixed rate on one hand - and all of those have been instances of either legacy kit or some really random rubbish. In most of these cases I've made recommendations that the kit be replaced when feasible, and in all of these cases suitable replacement hardware was available for sub-£200 and in many cases significantly less. I'm not saying they always replace it straight away, as you say getting network investment out of a SME can be like getting blood out of a stone - But that in no way changes the advice I will give them.

Like Neil I thought the majority of the rest of your article provided good sound advice and I would certainly have no compunctions in using it to help justify the advice I would give to an SME client. Its just a shame that I have to give it with a caveat of "but I don't agree with this bit".

By tiggertaebo on 14 Aug 2009

You maintain that the most likely chosen setting is 100/full; I don't see why that's likely. It's dancing on the head of a pin, and it wasn't what I told people to do.

By Steve_Cassidy on 14 Aug 2009

I did say that you hadn't said that.. my suggestion for it as the most likely choice is because unless someone is aware of the potential for a duplex mismatch then naturally they will try the fastest setting first. Even if they go straight to 100/Half (for example) and don't experience a problem then that doesnt change the fact that they have a potentially sub-optimal setup vs using AN at both ends which would likely give them 100/Full

By tiggertaebo on 14 Aug 2009

I did say that you hadn't said that.. my suggestion for it as the most likely choice is because unless someone is aware of the potential for a duplex mismatch then naturally they will try the fastest setting first. Even if they go straight to 100/Half (for example) and don't experience a problem then that doesnt change the fact that they have a potentially sub-optimal setup vs using AN at both ends which would likely give them 100/Full

By tiggertaebo on 14 Aug 2009

Sorry; was sitting on a Voda stick connection there and had to press "submit" in a rush. Blame the ferry poert at Holyhead... first of all I am very pleased to see that we have some proper high quality contributions going on here, now that we have seen the back of those with nothing technical or relevant to offer.

Secondly, I stick to my guns (a bit like Neil...) - the dancing on the head of a pin here is all about how people percieve the performance they get from their network. A VAST number of people carry on using broken networks because they have never seen a fixed one, and bizarrely enough this situation hasn't moved very far forward for rather more than 10 years.

I would suggest that you're still dancing on a pin, guessing what people do in response to the suggestion to get in and do tuning. "duplex" is a rubbish term to give any indication of performance: in the US it refers to a type of apartment! That's nobody's fault, more's the pity - I don't think even Stephen Fry would defend the use of that term near regular users.

When once people try a bit of tuning they should home in on the fastest option with about fifteen minutes of work (assuming the LAN card drivers actually do what their dialog boxes claim they are doing, that is - which as we all know from our scars, sometimes they don't!).

We discussed the article a lot in house before it went to publication, and one paragraph got dropped in the name of the funky design (which I have to say I REALLY like - the in house team cover up my godawful graphics choices exceedingly well). The lost paragraph said (in paraphrase) "don't tae all these rules as dogmatic edicts - use them as a starting point for an internal discussion, which should help you to discover whether you are heading down the right road with your network". I suspct this thread would not have happened with that paragraph in place - but on the other hand, it's a very good thread.

By Steve_Cassidy on 14 Aug 2009

I would suggest that anyone who doesn't understand what the term "duplex" means in a networking context really shouldnt be changing those settings.

You are right in that people will find a setting that works pretty quickly out of the list - regardless of which setting they start with - and you'll note that I have never said that network settings shouldn't be tuned. What I objected to was your instruction that people should NEVER choose Auto, which can quite often be the best setting. You're probably right Steve - with that additional paragraph in place we would likely not have been having this conversation.

By tiggertaebo on 16 Aug 2009

Steve.. please...

I tell you. If this article was meant to generate column inches of copy dedicated to on-page SEO then well done Steve.

However, I'd say that as an avid reader of PC Pro, that you should calm down a bit matey and remember that if you stick your head above the battlements then you'll be shot at. Take it in the grace the comments were meant and stop taking it personally. You're losing credibility with every post. It's more like a forum than a blog post now - and you're sounding like a spotty teenager who's been flamed by someone. Whether they are right or wrong is not the issue here. Your attitude to their opinion is.

By CraigieDD on 17 Aug 2009

Alan B

Interesting read, I guess the saying "no two networks are the same" springs to mind here.

Steve C- I think your postings have remained consistent and non provoking unlike the other posters who should sit back and thunk before hit the submit button.

By a_byrne22 on 18 Aug 2009

And that includes me!

Correction-
"think before hitting the submit button"

By a_byrne22 on 18 Aug 2009

And that includes me!

Correction-
"think before hitting the submit button"

By a_byrne22 on 18 Aug 2009

And that includes me!

Correction-
"think before hitting the submit button"

By a_byrne22 on 18 Aug 2009

Too True.

Basically good advice, and Steve dares to state the unspoken and seemingly-taboo fact that many "Small Business" packages are totally unsuited to that role, being basically a budget version of corporate software rather than being designed for small sites.

Microsoft's SBS, with its requirement for the installation of Active Directory, specialist DNS, etc, etc. is an extremely ineffective and overcomplex way of managing a small site. Yet, I regularly see it being 'push sold' to sites with only three or five desktops.

The fact of the matter is that on these sites, far more IT time is spent fixing obscure problems created by the Active Directory, user profiles, group policies and all of their labyrinthine dependencies than is spent on providing services of actual value to the end-user.

A site this size would reap far greater benefits from a standard server providing basic file storage -but without all the Active Directory bloat- and a decent ISP-based email package.

By Anteaus on 21 Aug 2009

Too True.

Basically good advice, and Steve dares to state the unspoken and seemingly-taboo fact that many "Small Business" packages are totally unsuited to that role, being basically a budget version of corporate software rather than being designed for small sites.

Microsoft's SBS, with its requirement for the installation of Active Directory, specialist DNS, etc, etc. is an extremely ineffective and overcomplex way of managing a small site. Yet, I regularly see it being 'push sold' to sites with only three or five desktops.

The fact of the matter is that on these sites, far more IT time is spent fixing obscure problems created by the Active Directory, user profiles, group policies and all of their labyrinthine dependencies than is spent on providing services of actual value to the end-user.

A site this size would reap far greater benefits from a standard server providing basic file storage -but without all the Active Directory bloat- and a decent ISP-based email package.

By Anteaus on 21 Aug 2009

Please stop these articles

Wow... First the Jon Honeyball one...
Then this...

Some parts are useful (the test your backups) -

But on the whole - these articles are for the small business owner aren't they?

They certainly aren't for a techie. And certainly aren't for a techie in the small business.

Autonegotiating NIC's aside (OMG how much that made me laugh and cry) -
Anteaus - Is just plain wrong...

the article is only occasionally accessible.

The portrayal of the last two articles is that IT is a dark art... And it's not that dark any more folks...

Have you hit your market? Would any SmallBiz owner actually read that and gain anything valuable from it?

Please PC Pro - STOP this line of article, it's got way out of touch with the sector on which you are trying to address, and I would fear that if any of my clients ever read this, without their eyes glazing over, I'd worry for the damage it would do.

Fortunately, as the articles are as about as accessible as some remote island in the middle of the Atlantic... I don't have much to worry about...

Just stop.
Now.

I don't even want to look in the next magazine and see if this car crash series is continuing.

(And I know I've said stop too much already... But the articles are that bad!)

Regards

Jules

By gingerinc on 28 Sep 2009

I remember when autonegotiation was a problem.

Indeed I had to set a whole Disaster Recovery site to 100/half on both sides as the rubbish nic in the HP Pavilion PCs (running Windows 2000) would not talk to the Cisco 6500 Fast Ethernet blades at 100FULL.

But since that occasion I have not seen the problem once. Almost every nic now supports NWAY which is the technology behind the standards 802.3u and 802.3z and they all seem to work.

The point you should be making is that Gigabit may not be the best solution for every connection, if the major traffic is databases then the packets are small and gigabit ethernet has a default larger packet size so actual network response may indeed be improved by switching to 100BASET rather than 1000BASET.

A graphics studio on the other hand has large files and will gain from Gigabit speeds.

The standard for 1000BASET recommends autoneg for a reason, it allows both ends to drop down to 100FULL if the cabling is not up to scratch. But a gigabit switch (which in all cases should be managed I would agree) should not be left at autoneg because following my earlier argument some network traffic would be better at 100FULL or 100HALF.

Read Rich Sieferts "The Switch Book" for a thorough explanation.

Applications drive networks not the available technology.

As an aside, we once did some testing and we found evidence that Windows NT and 2000 could not send and recieve through the nic simultaneously. I have never been able to repeat the tests for XP, Vista and Windows 7, has anyone else?

By Techno2009 on 3 Dec 2009

I think that these articles, and possibly even more the comments, are for people like me who aren't IT pro's in the networking field, but avidly gobble up every bit of information and try them out when they can.

I've lost count of the number of times i've seen some useful titbit and used it, wondering where I'd read it, and then coming back to old magazines and online articles/threads.

I like PC Pro because i think it gets more things right than wrong, and if one of the contributors acts human and possibly (that's my caveat) makes a mistake, then hey - one is not going to lose sleep over it.

The "prove it or lose it" argument that Steve made would have been lost on me anyway, again, because i'm in the learning phase.

In summary, good article, feels a lot less technical than I would have expected, but that was more than made up by the comments, 'tool' notwithstanding... unless the tool was one of those awesome linkrunner things!

By khellan on 7 Dec 2009