🚀 go-pugleaf

On Tue, 09 Jul 2002 15:11:17 GMT,
spammme@yeahright.lanboyo.homeunx.net (Komalley) wrote:

>Back in the days when netbui was the default stack on NT, and TCP/IP
>was on its own diskette, NT did a really crappy job with the IP stack
>and SMB shares.
>
>By 4.0  I think you would find that tcp/ip-netbui  performance was
>very similiar.

Not..  NT 3.5x and 4.0 had roughly the same performance for both
protocols.  Actually, 4.0 lost a few percentage points when compared
to 3.51 for a single file xfer session.  

P.S.   I still have my dual boot NT 3.51/4.0 networks in place..
  Both workstations & server..  

Planning to retire the server and switch to linux.. But the Samba
implementations don't support a Netbeui mode. :-(  sniff. 

>The main problem with Netbui in my eyes is that Windows will try to
>use it for name resolution, even when it should know better.

I found in practice the exact opposite occurs.  Netbios over tcp get's
connected first (Wins resolution occurs before reception of Servers
Netbeui broadcast during IPL/login.)

Winds up using a lot more CPU cycles(30x) while running 10x slower for
all subsequent file xfers.. Yuck.. 

>Sniff on the line and watch all the requests for the netbios name
>www.google.com  and you will see what I mean. This results in odd
>pauses when accessing the net, and of course, ulcers.

Not really.. Netbeui Idle state is way less than .1% of traffic of a
busy network.   A server announcement once or twice a minute(~200 byte
packet)  and netbios poll(s)/final(w) sequence every 30 seconds. (60
byte packet). 

Big deal.  

If you don't believe me.. Then, I suggest you start copying very large
files to/from your server.. Be sure to watch your cpu utilization(s)
while timing the xfers. Use files that are at least 50% bigger than
RAM on your local workstation.. Can't have any cheating using a local
cache.   

>On Sat, 06 Jul 2002 05:26:58 -0400, Tim Keating
><NotForJunkEmail@directinternet11.com1> wrote:
>
>>On Sat, 06 Jul 2002 05:46:54 GMT, "Warren H." <wholzem@hotmail.com>
>>wrote:
>>
>>>Gunter Herrmann wrote:
>>>>
>>>> Since it is about using a router NetBEUI is completely useless.
>>>> Unlike TCP/IP and SPX/IPX, NetBEUI is NOT routable.
>>>
>>>You are correct that NetBEUI is not routable, but the "broadband routers"
>>>available on the market today use a router only for the LAN to WAN
>>>connection. The LAN is still serviced by a switch or a hub, so non-routable
>>>protocols still pass on the LAN.
>>>
>>>HOWEVER, just because it does work is no reason to install it. TCP/IP will
>>>work on these networks as well. If it doesn't, then something is
>>>misconfigured. Adding a second protocol doesn't fix the problem. It just
>>>works around it. One or the other protocol is going to be the default
>>>protocol. Let's look at the options:
>>>
>>>If TCP/IP is the default protocol, then TCP/IP communications will be
>>>attempted first, and until they fail or time-out, NetBEUI won't be used. If
>>>NetBEUI is the default protocol, then NetBEUI communications will be
>>>attempted first, and TCP/IP won't be attempted until NetBEUI fails.
>>>
>>>So while previous posters have noted that NetBEUI is a great little protocol
>>>for small LAN's, it was designed with the assumption that it would be the
>>>primary protocol, or perhaps secondary to IPX/SPX on a Novel network. The
>>>Internet was something that existed, but most LAN's weren't even connected
>>>to the Internet. In those days, NetBEUI was the perfect solution to simple
>>>networking.
>>>
>>>The situation today is different. TCP/IP is so integrated in Windows XP that
>>>it cannot be uninstalled through the GUI, and it cannot be made to be the
>>>secondary protocol to NetBEUI. NetBEUI is so much out of the picture that
>>>they buried it on the CD, and didn't make allowances for it to be installed
>>>through the GUI.
>>>
>>>The solution to having a LAN where TCP/IP is not working for internal
>>>communications is to fix TCP/IP. Installing NetBEUI is like putting a sign
>>>on your front door telling people to go around back instead of fixing the
>>
>>Not quite.. NetBeui has many advantages..
>> NO need to worry about direct hacking from outside..
>>   Non-routable protocol.. 
>> It's also a LOT faster and more efficient protocol..
>>
>>>way the front door sticks. It may be a temporary work-around, but it's not a
>>>solution.
>>
>>1st. Last time I measured it.. NT 3.5x/4x land.. 
>>   Netbeui was 20 to 30 times more CPU efficient than it's TCP based
>>cousin.   It was also ~10 times faster over 100BaseTX connection. 
>>
>>   I.E. Netbios over TCP  400KB/sec.. 100%  of 100 Mhz P5 cpu.
>>         Netbeui 4MB/sec.. 35 - 40% of 100Mhz P5 cpu.. 
>>
>>2nd. NT tended to use the Netbios over TCP first.. Why? only M$
>>knows.. It still is a real pain to force it too use the more efficient
>>Netbeui stack.  I.E.  Break the bindings Between Netbios and TCP/IP.
>>
>>3rd. You can peacefully coexist TCP and Netbeui on the same lan.. You
>>just got to know which bindings to break.     
>>
>>----------
>>
>>Now, you may think protocol efficiency doesn't matter with today's
>>faster CPU's.. Until you scale up to a 1Gb or 10Gb/sec Lan, then you
>>can watch all your server&client CPU cycles get sucked up, handling
>>the TCP/IP stack, while your super fast Lan remains mostly idle.  :-(