Thursday, March 18, 2010

Network Dictionary — Mythinformation

July 15, 2008 by Greg Ferro · 6 Comments 

Mythinformation is the term used to describe some form of know­ledge that is based on apo­cryphal, false or old know­ledge. A play on the the word mis­in­form­a­tion.

A clas­sic piece of myth­in­form­a­tion is the belief the Ethernet auto-​​sensing does not work. The stor­ies of a mis­con­figured Ethernet ports caus­ing major prob­lems are either myth­ical or based in the past when there were no accep­ted stand­ards for auto-​​negotiation.

The only instance for lock­ing duplex/​speed is when using obse­lete equip­ment, or when you do not know the con­fig­ur­a­tion of the remote side. In fact, for­cing speed and duplex reduces Ethernet func­tion­al­ity. For Gigabit Ethernet par­tic­u­larly, dis­abling auto-​​negotiation will almost def­in­itely cause poor per­form­ance and out­ages when the Gigabit Ethernet line gets heav­ily congested.

Please rate this post:

  Why Rate Posts?
1 Star - It\\\'s Crud2 Stars - It\\\'s Tosh3 Stars - Something\\\'s missing4 Stars - Needs works5 Stars - Good Enough6 Stars - Good7 Stars - Excellent8 Stars - Brilliant9 Stars - Astonishing10 Stars - Awesomely Godlike? (No Ratings Yet)
Loading ... Loading ...

Comments

6 Responses to “Network Dictionary — Mythinformation”
  1. Jeff Darcy says:

    I think you could have picked a bet­ter example. Autoneg *does* fail on a lot of equip­ment which is argu­ably either obsol­ete or cheap or both, but still quite com­mon even in ser­i­ous data cen­ters. I’ve seen it with my own eyes, it’s no myth, and don’t go all No True Scotsman on me by say­ing any­thing that fails autoneg is obsol­ete by defin­i­tion. If you want a real myth, how about the one about the “vast major­ity” of serv­ers not being able to gen­er­ate 1GE worth of HBA traffic? That’s apo­cryphal, false, or old for you.

    • Greg Ferro says:

      I have a dif­fer­ent exper­i­ence. The only autoneg prob­lems I have seen in the last 3 years where when someone else wasn’t using it, or when the cabling was faulty. Hard set­ting bypassed the faulty cabling, by in the long term it did cause errors and was replaced.

      And yes, I and oth­ers, still believe that. However, it gonna be pretty sub­ject­ive, your exper­i­ence may be with high qual­ity serv­ers with good HBA /​ TOE /​ setup para­met­ers and that will make the dif­fer­ence, but most people don’t have that setup or that exper­i­ence so I still stand by my statement.

      I also know that it will not true in the future, but, hey, its a mov­ing target.

  2. Jeff Darcy says:

    There’s a pat­tern here: you haven’t *per­son­ally* seen some­thing so you dis­miss it as a myth. Well, I’ve seen the things you claim are myth­ical. I have sev­eral dozen I/​O nodes in my lab at work, each of which is a real-​​life counter­example to your “1GE worth of HBA traffic” claim. The kicker is that, on an indi­vidual basis, each of those nodes is not very expens­ive or power­ful. This laptop could beat one on most tasks. Clearly it’s the *non*existence of such things that’s a myth. The tar­get has *already* moved.

    Yes, you have a dif­fer­ent exper­i­ence. You can stand by your state­ments all you like, but that doesn’t make them true. My exper­i­ences aren’t myth­ical; I prefer to think of them as legendary. ;)

  3. Ivan Brunello says:

    Just two notes:
    1) Never dealt with OOOOOLD cabling, which still uses 2-​​pairs patches on PC110 con­nect­ors. Well, we had to manu­ally force our gig­abit switch to run 100 FD.

    2) On the other hand, talk­ing about Cisco equip­ment, I found old FastEthernet switches were really a mess w/​ auto­sensing; it worked no more than 50% of cases.
    BUT, autone­go­ti­ation on gigabit-​​class ones (not only against gig­abit hosts, but also down to FastEthernet) was indeed really working.

Trackbacks

Check out what others are saying about this post...


Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!