Fiber

Only novices daisy-chain switches

If you support so many network nodes that multiple switches are required, technology consultant Erik Eckel encourages you to do it right.
There may be times when you're tempted to daisy-chain switches, or needlessly interconnect network switches because you're too lazy to run new Ethernet drops to a location where a single drop exists but two or more are required. This amateur practice is a quick fix, but it's no long-term solution.

Some offenders suggest the redundant switch is temporary, and they'll return to properly run additional Ethernet cabling, but that rarely happens. Instead, the client is left with interconnected and undocumented network switches, more unnecessary hops, poor network performance, and additional devices to fail following power outages and lightning strikes.

If you support so many network nodes that multiple switches are required, do it right. Purchase stackable switches designed to work together that scale to match the site's requirements. Cease interconnecting switches via an Ethernet cable whenever possible. Seek switches featuring high-speed stacking ports (such as via an HDMI interface). Then power the switches using a proper battery backup.

The real world

Unfortunately, best business practices aren't always followed in the real world. I've lost track of the number of times my consulting office has inherited a client complaining of intermittent connectivity issues or just plain poor network performance.

On a few occasions I've even been called in to troubleshoot poor performance after the previous IT provider deployed a new and potent RAID 10-, 32 GB RAM-, SSD drive-, and dual-CPU-powered server only to discover performance issues still remained. Following a simple on-site assessment, the cause becomes obvious, at least to a professional.

Frequently I find a network closet with a 24- or 48-port switch. If the client's lucky, the switch is powered by a battery backup. Then I might find a 16-port switch in a corner office that shares a single Ethernet drop with a half-dozen switches and a handful of printers. Then there's a five-port switch hidden behind a file cabinet in the receptionist area and another five-port switch in a nurse's station that shares a single Ethernet cable with a pair of machines and a network printer. Oh, and these extra switches are all plugged straight into surge protectors, too -- there's no battery backup in sight. How surprising.

Every time the power browns out, there's a thunderstorm, the electricity fails for just a fraction of a second, or some other blip occurs, those (and in some cases hidden) switches suffer corrupted memory. Subsequently, certain pockets of the office can no longer connect. Rebooting the main switch in the network closet doesn't fix the problem. Frustration rises.

I've even seen rogue switches hidden behind cubicles built a decade earlier no one knew existed malfunction and flood the network with bad traffic. Try troubleshooting that problem when you can't even physically locate the switch.

Avoid becoming an offender

You should follow best practices and insist on deploying additional Ethernet drops when they're required. Unless you're caught in an incredibly rare case in which the site is a historically registered building and a cable just can't be run, you should run new cabling.

Here's the cost justification: If you don't run new cable and you depend upon daisy-chained connections, trouble will arise -- it's a question of when, not if. This is true even if you go so far as powering the extraneous switches using a battery backup, which helps reduce the number of network failures. Each time a failure arises, the client's going to suffer an outage, which means staff can't work and that translates to lost revenue. Then the client calls you, which results in a service call. Maybe you power cycle a switch or two or even replace one of the extraneous switches that's been fried by a lightning strike, resulting in another $75 in parts and $125 in labor. Consider that scenario may repeat itself three of four times over four or so years, and the client's looking at having lost hours of staff productivity and maybe a thousand dollars in switches and service.

A new cable drop might run $100 or $200, and it will provide a much more reliable network experience. It's simply more cost effective. Clients won't always understand the cost efficiency at first -- that's your responsibility. As a consultant, it's your job to help educate the client and help the client make better, more informed decisions. Point to your years of experience, and then finish with the simple truth: only novices daisy-chain switches.

About

Erik Eckel owns and operates two technology companies. As a managing partner with Louisville Geek, he works daily as an IT consultant to assist small businesses in overcoming technology challenges and maximizing IT investments. He is also president o...

25 comments
txrx
txrx

I work in an AV company - we often have a main hub with a data switch and then 1 or 2 home cinemas which also require around 8 Ethernet connections. A small 8 port switch is normally implemented to provide data.

Is this article suggesting that we should be running 8 cables direct from the main switch for each of these devices?

A93
A93

Although I agree with this post I feel that the writer has not been considerate when talking for small businesses with low budget. And secondly for building codes and restrictions. Apart from these two points I don't have any other reason to opposing cable drops when compared to Daisy chaining.


reisen55
reisen55

A co-consultant I work with had a law office completely re-wire his environment for about $5,000 and it still did not solve issues, and the check was not sent either. Bad client indeed. In some offices, such as one of my best, hanging a small switch for a few machines is a necessity of cost. These economic times stink!!! My clients want it cheap and fast too, and I would love to do more cable drops but that costs a ton of money. Nobody is looking for spend right now. So calling us LAZY or, worse, a NOVICE, is an insult directly aimed at hard-working, suffering consultants. Shame on you. Oh, my clients don't withold checks.

fgranier
fgranier

One should never: Use copper to go across buildings. Use copper to go between floors. And if one is in an industrial plant one should also aply the same rules. So use managed switches with redundant communications betwwen them if you need high availability. Use a management tool that allows you to manage all your switches. Do not create fire hazarads and avoid grounding problems.

CharlieSpencer
CharlieSpencer

Having a UPS or other battery backup is no good if the batteries had died. Create an appointment or task in your calendar to order replacement batteries every three years. Use a permanent marker to write the date of installation or replacement on the unit, so your successor will know how old the batteries are.

llsdigitek1350
llsdigitek1350

Lazy? How about dealing with a client that is two cheap or does not have the money to pay for the extra runs in this down economy?

tbmay
tbmay

You clearly have a better customer base than I did. I didn't have the privilege of "insisting" anything. The vast majority of mine want it now and cheap. They would like "good" but they weren't willing to pay for it....so good enough is good enough. Congratulations, you've built a business of good customers if they're willing to let you do things right.

Charles Bundy
Charles Bundy

If by "daisy chaining" you refer to multiple switches in a rack where A - B - C. Then you are foobar if B dies as it breaks the interconnect. I agree that is a bad practice but you jump more than $200 when you start buying with stack/backplane capability. Of course you only cry once when you buy quality! The other issue is hanging one switch off a port. That would not be "daisy chaining" but oversubscription. Not really a performance problem at the desktop except as you point out documentation of the beastie. Of course you start trying to do that in a packet dense environment and you are asking for trouble (e.g. the server room.) I've never had an unmanaged desktop switch get addled by a brownout. Actually I've had more problems having desktop switches on UPS get confused when attached devices that aren't go south for the winter! (e.g. laser printers) Thus I'm not real big on stackability unless you are going to need more than 48 ports in a floor closet (get a single device.) In that case use a fiber switch and oversubscribe per floor. As far as desktop switches behind panels go you let facilites do that, they are hiding more than switches! Power, ethernet, phone, J. Hoffa structural anomaly #1 come to mind... :)

Gumm
Gumm

So, the justification for not using five port switches is lightning strikes? Get real.

TechKid0107
TechKid0107

You hit the nail on the head with this one. Very good article

dcolbert
dcolbert

Thanks for sharing. Timely and relevant to my line of business, I'll be sharing with my staff.

muto
muto

Your remarks are spot on. But don't get too upset with the opinion of the Tech Republic writter... after all you get what you pay for and last I checked Tech Republic is free ;) Besides, the author is clearly not familar with the cost of installing a cable in an A Class New York City office building, with restrcitions on time of day for installations, insurance, and union membership all at issue. It is hard to justify a single cable run for $1,000 when a $30 5 port gigabit switch could do the job... or what i have ocassionally done, putting an 8 wire cat-5 splitter on each end (to create two ports without the use of any electronics that can potentially fail).

strachan.william
strachan.william

The NOVICE bit is quite insulting! I have a client who is using several desktop switches to provide extra connectivity, because they're too cheap to rewire the facility. I've also had to daisy chain these switches off of their POE VoIP phones, because they're too cheap to buy power adaptors. This is not ideal, by any stretch and I hate supporting these environments, but...it is what it is....end of the story!

oldbaritone
oldbaritone

One should never Use copper to connect areas fed by different electric service entrances. Some large old industrial buildings had a "second service" installed to increase capabilities or provide for some new machine or process. Do yourself a favor. Run fiber to that area, even if it's only a few feet away. You'll save yourself a lot of headaches. Got a tee-shirt. Don't need another tee-shirt. Trying to save you from getting a tee-shirt... ;-)

NickNielsen
NickNielsen

One installs what the customer is willing to pay for unless one is willing to suck up the additional cost of meeting standards....

tbmay
tbmay

I've run my share of cable as an employee. As an independent, anything over the simplest of task that doesn't require drills and wall penetration is an absolute no-no. That needs to be done by people who know what the heck they're doing, and are properly equipped and licensed.

llsdigitek1350
llsdigitek1350

Amen... I too have clients that want it now and cheap

dave
dave

I have done extensive testing on the so called 5 port (8 port, anything unmanaged xx ports, sub $100) puppies that you buy from Best Buy, etc. These have a single Ethernet chip, no buffer memory, no 802.1d (or even original STP) and are unmanaged. I tested file transfers (large and small) between 2 PCs against 2 different Nortel Switches (24 port and 48 port). These have large memory buffers, 802.1d and the rest of the alphabet soup and are manageable. The Nortel switches (or any other commercial vs residential) were half the time completing file transfers and stable. I had experience with a Netgear 8 porter taking down an entire L2 network consisting of 20 floors and 1,000+ people. Only a router stopped it from spreading further. Oh and by the way, I will NOT engage with anyone about the pros or cons of that architecture at that point in time. So here is my challange to Gumm. Give me a 5 porter that you use and 5 minutes alone in your building. I will take down your entire network so badly that all your manageable switches will be using all their cycles for forwarding traffic such that you will not (via Telnet or SNMP) be able to query ports to see where the traffic is coming from. Even a direct console connection will be painfully slow or even unresponsive. You will have to manually disconnect each switch until you find that the traffic subsides. Only then will you be able to start getting warm as to finding where I hid the unit and what I did. The above scenario was accidently created by a advanced help desk person who was trusted with adds, moves and changes. Get a proper switch and leave the 5 porters at home.

tbmay
tbmay

I also put the idea of running a full time business to rest. The small business customer is honestly not much better...as a market for it services...than the residential customer....and you REALLY don't want to go down that road. The truth of the matter is I pretty much agree with Erik on most things, but there is something about "computers" and "IT" that makes many many people think it's not money well spent. I still work on specific projects for larger organizations or specific needs. Primarily software related....system integration, etc. Networking is often part of it but we're talking about vpn's, firewalls, etc. I'm done with trying to convince Joe Business Owner we should do things the right way. I'm simply not a good enough salesman to tell someone we should drop the money on a quality switch when he's already barking about the $200 unmanaged dlink. Especially when that dlink will probably run for years and be relatively inexpensive to replace.

oldbaritone
oldbaritone

because those cheapskates will go hire a shyster when something goes wrong. Sometimes it's better to face the customer and say "I'm sorry, but I won't do that. It's not safe to do that." I had a customer who wanted me to throw a small switch on top of a suspended ceiling above a work cluster, and string zip-cord extension cords to it. I told him "Absolutely not!" - in writing, mentioning that this would be unsafe and a violation of local codes. So he found some dumb kid who "install[ed] what the customer was willing to pay for". I don't do work there any more, and I know it's for the best. The cheapskate found a chump, and frankly this customer was more trouble than he was worth. He always took the cheap way out, and howled about the bill or refused to pay when things went wrong. I'm better off because that customer is gone, and I still have my professional integrity.

Gumm
Gumm

And what I would say back is this, I work in the UK, unlike America we have an awful lot of listed buildings and I happen to work in one where we're not even allowed to hang things up on the walls. I would take on your challenge but you're obviously unable to give reasonable grounds for such an engagement. In larger businesses with a healthy IT budget (or any sort of budget at all) supplying the kinds of equipment needed is probably no problem, but in businesses like mine, they would never shell out that kind of money and would prefer to run the risk of a collapse (as utterly rare as they are) and I can quite understand that. By that standard, smaller businesses have fewer switches and making it easier to pin down. Being a network admin for seven years and never having switching problems I think the 'gamble' so far is worth it. Rather than replacing the switch in your case, you should probably train your staff a little better eh?

NickNielsen
NickNielsen

That was an ethical situation and a violation of building/fire codes rather than a violation of best practices. I wouldn't have installed the thing either.

dave
dave

The staff member was trained and had been doing this work for years. I'm just saying that even a properly trained individual can accidently make a mistake. Happens to everyone sometime. As much as I hate to say it you could look into the grey market and get some manageable switches that have some basic standards. The fact that your network has been running as well as it has over the years is a good thing. No everyone is so lucky.

Editor's Picks