Gigabyte Network "Teaming" and other Dark Arts!

[MadAd]

Oh im a quick and dirty kind of guy no need to go to high tech methods just to get a rough outline of what to expect there Mr Beagle.

 

[Madwand1]

iperf is free, and has been used by some organization for Internet landspeed records in the past.

E.g. using version 1.7:

server: iperf -s
client: iperf -c server -l 64k -t 15 -i 3 -r

Tests single transfers using 64k buffer size for 15 seconds, with results every 3 seconds, from client to server, and then in the reverse direction.

I think I'd find some time during a break from cleaning the garage (if I ever started it) to run such a simple test.

 

[TheBeagle]

Good Evening Mr. Madwand & Everyone.

Mr. Madwand: May I suggest that you take the time to engage in the testing if you desire to have those types of results. Obviously, you portray yourself to be a very knowledgeable network type individual, and you also seem to have a burning need to solicit others to do you bidding. However, in case you didn't hear about it, in the USA, President Lincoln freed the slaves a long time ago - thank God! Therefore, since I don't have any current need for the additional information that you seem bent on obtaining, I wish you success in obtaining that material on your own.

As far as my choice of pursuits is concerned (cleaning my garage versus running tests for you), I'll reserve the right to make those personal judgments concerning the allocation of my available time. Good luck with your forthcoming testing endeavors!

Best regards to everyone. TheBeagle :beer:

 

[Madwand1]

Mr. Beagle, I am somewhat offended that you portray my offers for help and knowledge as slavery. I suggest that you reconsider your words and and what I have said.

I and others have done this sort of testing, and it on the basis of this testing that we declare that you are wrongly portraying the benefits of teaming. My suggestion if for your and others' benefit to better understand the benefits and non-benefits from such a setup. If you wish to continue to justify your efforts under false assumptions and a lack of understanding, it is your right. Just don't think that what you've posted in this thread without having done the tests to confirm them is correct.

 

[Blazer7]

Hello Mr Madwand1,

I was under the impression that this thread had run its course. AFAICR it was never the point of this thread to promote teaming and I can't remember anyone here trying to do so. The point of this thread was to make clear of what teaming requires in order to function and that has already been done.

What's this ?I and others have done this sort of testing, and it on the basis of this testing that we declare that you are wrongly portraying the benefits of teaming.??

Who is portraying the ?benefits? of teaming and who's ?we??

If you know of any of ?the benefits and non-benefits from such a setup? that haven't been mentioned in this thread please feel free to share with us and if you have indeed tested teaming as you imply please post some numbers. That would be beneficial.

 

[Madwand1]

Originally posted by: Blazer7
I was under the impression that this thread had run its course. AFAICR it was never the point of this thread to promote teaming and I can't remember anyone here trying to do so. The point of this thread was to make clear of what teaming requires in order to function and that has already been done.

No, this thread has made clear claims about performance benefits given specific configurations, and these claims are false, and based on misunderstandings.

Besides, if there are no implied or hoped for performance benefits in this setup, then why bother? Why be interested in teaming? Why spend money on managed switches? Why tell people this is what you need to do if you want this work? For what end? For no good reason, just because you can and have money to spend, or because you think it will give you a performance benefit?

http://www.alacritech.com/Supp...Documents/Teaming.aspx

There are a couple of common misconceptions about link aggregation that cause people to have unrealistic expectations of resultant performance.

 

[Blazer7]

Hello Mr MadWand1,

I do not fully agree with you. In fact I cannot agree with you and that's because of the following things that have already been posted here.

Originally posted by: Madwand1

No, this thread has made clear claims about performance benefits given specific configurations, and these claims are false, and based on misunderstandings.

Originally posted by: TheBeagle
Good Morning Mr. MadWand1 & Everyone.

Mr. MadWand1: The object of my lengthy endeavor was to get the Teaming function to become operational.

Originally posted by: TheBeagle
Good Morning Mr. MadAd & Everyone.

In reality, I was just damn glad to find out HOW to get Teaming to function, let alone start some additional comparative testing routine.

I believe that it was already mentioned that performance wise there is little to gain from teaming so I really do not understand your obsession with this.


------------------------

Originally posted by: Madwand1
Besides, if there are no implied or hoped for performance benefits in this setup, then why bother? Why be interested in teaming? Why spend money on managed switches? Why tell people this is what you need to do if you want this work? For what end? For no good reason, just because you can and have money to spend, or because you think it will give you a performance benefit?

Here are some more exerpts from previous posts that answer your questions.

Originally posted by: Blazer7

the topic is not just about networking but also about the mobo manufacturers policy on not disclosing all info related to this feature to their customers.

Shedding some light on this may turn out to be quite beneficial for some ppl here in Motherboards. Understanding how a specific feature works along with its requirements and overall functionality may change how ppl look at teaming and how this specific feature affects their final decision when purchasing a mobo.

and again....

Originally posted by: Blazer7

We are not talking about server hardware here so why would anyone need teaming ?

Teaming is not the only issue here. Another one is the fact that mobo makers do not provide their customers with all the needed info on this specific feature while they do advertise their products as being capable of teaming. Letting ppl know of what teaming requires, what it's real use is and what they can expect from it makes it much easier to evaluate the need of that feature.

and lastly.....

Originally posted by: SolMiester
I think the best function of your teamed motherboard NIC's, will be fault-tolerance, especially if you have mapped drives back to the server.

I believe that this topic has been adequately covered and I really cant see any reason why we should discuss the same things over and over again.

I sincerely fail to see what new your last posts bring to this.

As far as I am concerned this thread has run its course.

 

[Madwand1]

Originally posted by: Blazer7
I believe that it was already mentioned that performance wise there is little to gain from teaming so I really do not understand your obsession with this.

If I've been repetitive on this point, it's because from what I've seen, it hasn't really been absorbed. It's an easy mistake to make, and I think people have been making it and not realizing it. So I've been somewhat pedantic on this point.

If you agree with my point, that there's typically nothing to be gained from teaming in home environments, that this is a fruitless expenditure of time and money, then we're done.

If however you like the OP posted advice about getting teaming to work and what additional acquisitions you'd need to make for that, then I think the question would be -- why bother if it doesn't provide any performance benefits in the end? Or if it does, then what are the benefits? How can these be seen / shown?

 

[Blazer7]

Originally posted by: Madwand1
If you agree with my point, that there's typically nothing to be gained from teaming in home environments, that this is a fruitless expenditure of time and money, then we're done.

It depends on how one values redundancy. Even if the performance gains are negligible the addition of redundancy will carry some weight for many users, your truly. So no. In that aspect teaming does have some practical use especially if you have a home server.

If however you like the OP posted advice about getting teaming to work and what additional acquisitions you'd need to make for that, then I think the question would be -- why bother if it doesn't provide any performance benefits in the end? Or if it does, then what are the benefits? How can these be seen / shown?

As I've already mentioned, repeatedly, this doesn't have to do with performance only.

Like the OP I too believe that people need to know of what teaming requires in order to function properly. Since teaming is advertised by mobo makers it is only fair that the public should learn of the cost, requirements and benefits of teaming in order to evaluate the need of that feature or not.

This thread has already covered all of the above.

 

[Madwand1]

Originally posted by: Blazer7
Like the OP I too believe that people need to know of what teaming requires in order to function properly. Since teaming is advertised by mobo makers it is only fair that the public should learn of the cost, requirements and benefits of teaming in order to evaluate the need of that feature or not.

I agree with the spirit here.

As it happens though, much of the material in this thread is misleading and incorrect.

It isn't even true that you must use a switch with teaming support in order to team NICs, although this varies with different implementations, and using the standard technique with a managed switch is preferred for standardization.

 

[Blazer7]

Originally posted by: Madwand1

As it happens though, much of the material in this thread is misleading and incorrect.

I do not understand your comment. Most of the material here is on target and anything that?s not has already been corrected by those that post here. AFAICS there?s nothing left to correct.

It isn't even true that you must use a switch with teaming support in order to team NICs, although this varies with different implementations, and using the standard technique with a managed switch is preferred for standardization.

Since we are talking about mainstream mobos and not server hardware I wouldn?t expect any manufacturer to implement teaming in a more elaborate way than GB does. That said, and since GB requires a switch with LAG for teaming to work I would expect that the same goes for ASUS and the rest.

The good thing here is that a router/switch with LAG is probably the easiest way that the average joe can setup and use teaming. The Linksys SLM2008 that the OP was talking about is a very good solution at a very reasonable price ($110) and as I can recall there was also mention of an even cheaper 5-port switch for $20 less.

However, if you do know of any solutions out there that will allow teaming with the use of a normal switch/router then please share with us. This will undoubtedly be both helpful and enlightening.

 

[Fallen Kell]

Glad I bought a 16 port SRW2016 2 years ago People called me crazy buying a beefy piece of network hardware like that, but now I get to tell them "I told you so"