John,
That's a great post which works quite well with my post titled FW: setting mtu size on
a 2611".
A year and a bit ago, I wrote two papers - one on jumbo frames ("Extended Frames,
Uncovering the Myth: 'Is Bigger Really Better?'", and another on "ATM vs. Gigabit
Ethernet". The jumbo frames paper led quite well into the ATM/Ethernet debate with
the overhead associated with ATM playing a big role.
ATM is definitely useful in environments transporting small data frames (or cells) -
and is ideally suited for (and is more efficient than Ethernet) voice transmissions.
Many "housekeeping" functions of networks have small payloads, and again, ATM is a
good technology offering lower overhead than Ethernet **when the data to be
transmitted is small**. For larger payloads like file transfers, it becomes grossly
"top-heavy" in overhead.
For those of you unfamiliar with jumbo frames, the idea was to increase the MTU of
Ethernet from 1500 bytes, to 9000 bytes. The concept was developed to address several
issues, perhaps the most problematic of which was the capability of PC systems to
receive and process Gigabit speed data. It wasn't uncommon for a PC to have a maximum
throughput capacity of only 400 Mbps - lacking the capability to utilize the
full-bandwidth of the technology.
A 1 Gigabit per second Ethernet network, is theoretically capable of accepting
1,707,650 (64-byte) frames per second down a 1000Base-T Category 5 cable. System bus
and processors were struggling to keep up when Gigabit technology was first
introduced. By increasing the maximum transfer unit of Ethernet to 9000 bytes, a
Gigabit pipe could handle 14,000 frames per second (a 1500 byte frame would transport
80,000 frames per second). With the overead on large frames being less, PC's would
have fewer frames to deal with "per second".
Jumbo frames is an interesting technology - I'm somewhat surprised it hasn't received
more attention to date. Perhaps some of you who work for network product
manufacturers would like to take a stab as to why this isn't receiving more
attention... especially since in October of 1998, Alteon Networks announced it had
developed a way of making jumbo frame technology transparent to networks that don't
support jumbo frames.
-- Leigh Anne
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
> John Nemeth
> Sent: Wednesday, June 07, 2000 4:55 PM
> To: Mark Holloway; K Sacca
> Cc: [EMAIL PROTECTED]
> Subject: Re: ATM
>
>
> On Oct 29, 6:28am, "Mark Holloway" wrote:
> }
> } You honestly think ATM is going to take over Gigabit? ATM had
> its chance in
>
> I don't think it has a chance.
>
> } business. ATM on the WAN is still growing rapidly, but ATM on
> the LAN is
>
> Actually with technologies such as 10GbE (expected in the next
> year or so), [CD]WDM, MPLS, lamba switching (switching of light rays),
> etc. I expect that ATM will soon start to die in the WAN as well,
> except for legacy networks. ATM is just too expensive, complicated,
> and inefficient. It has a 9.4 per cent raw overhead per cell and when
> you add in all the gobbledygook necessary to actually use it, the
> overhead supposedly comes to 24.5 per cent. On the bottom of page 298
> in TND, Priscilla says:
>
> "One disadvantage of ATM is that the overhead for transmitting ATM data
> is much higher than the overhead for transmitting traditional LAN
> data. The 5-byte header required in each 53-byte ATM cell equals 9.4
> percent[sic] overhead. When segmentation and reassembly and ATM
> Adaptation Layer (AAL) functionality are added, the overhead can grow
> to 13 bytes or 24.5 percent[sic]."
>
> That means that your nice OC3 connection only has a throughput of
> 117mbps. That is absolutely rediculous.
>
> }-- End of excerpt from "Mark Holloway"
>
> ___________________________________
> UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
> FAQ, list archives, and subscription info: http://www.groupstudy.com
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
___________________________________
UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
FAQ, list archives, and subscription info: http://www.groupstudy.com
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
