From: Vincent Poy (vince_at_victor.MCESTATE.COM)
Date: Sun Oct 03 1999 - 18:39:07 EDT
On Sun, 3 Oct 1999, Mark Bell wrote:
> > Yep, a modem doesn't even use up 5% of the bandwidth of the ISA
> >bus.
>
> Let alone even requiring the elder XT bus technology. POTS lines have
> serious limitations, and as fast and as furious as modems have advanced, it is
> still a drop in the bucket compared to a LAN connection. I routinely go to
> gaming conventions, and the one thing I love seeing is people who have never
> played games or transfered files on a LAN before (They literally buy a network
> card the day before or hope to buy one at the CON). They don't realize how
> slow their modems are! :)
Oh well, I just hope PCMCIA modems have decent buffered UARTs. It
seems like people today still thinks that modems are rated in Bauds and
not Bps since 2400 Baud is the fastest modem available. Even a 56kbps
modem is 2400 baud x 24 bits. :)
> >> In addition, based upon the PC99 spec whipped up by the WINTEL monopoly
> <g>,
> >> ISA is slated for total removal in the next wave of PCs (Meaning your next
> >> PC will NOT have ISA slots).
> >
> > That alldepends since I think there will still be atleast 1 or two
> >ISA slots since there are some cards that are still ISA now like my
> >Terratec EWS64 Soundcard which is supposed to be high-end.
>
> No, the new spec calls for total abolishment of the ISA slots and only PCI
> slots from now on. Now, of course, manufacturers can break rank, but this is
> the first time that Microsoft and Intel have officially banished ISA to the
> land of dead technologies when drawing up their yearly PC spec. All previous
> specs included allotments for inclusion of ISA slots. Yep, ISA cards are
> still
> produced. My Sportster 128k ISDN Adapter is ISA.
I know, what I meant was what will happen to people with ISA
cards? Perhaps someone may come out with some kinda converter or a
external box. I guess eliminating ISA is a good way to fix the problem of
no IRQ's available.
> > You forgot that there is half duplex and full duplex.
>
> On a good day, no traffic, full duplex, even a switch connection if you
> want (non gigabit). Theoretical, maximum bandwidth so we can ascertain if
> peak conditions are containable within bus standards. >;P I really was
> not targeting differentiating conditions of the NIC vs. shared bandwidth,
> switch connection, or duplex mode. If someone wants a breakdown on network
> topological design, email me. <g>
Heh, but I mean Full Duplex does make a difference since even my
network at home is congested if it was half duplex.
> > The bottleneck isn't in the bus but with the speed of Hard drives
> >since no HD can write at 12.5MBytes/sec sustained. 3Com never warned
> >anyone about the 10/100Mbps not giving any speeds better than 10Mbps for
> >the PCMCIA version.
>
> Sustained vs. Burst. Burst conditions *are* realized from time to time and
> the network would feel the pinch during those moments. You can always
> construct scenarios to utilize full bandwidth (RamDisk comes to mind along
> with burst, and large cache controllers (SCSI)). On a network I consulted,
> an Access Database MDB was on the back end (~80megs in size) in a RamDisk for
> optimal speed for client access. The point is to realize the maximum limits
> and construct
> similar, potential loads (load balancing/load min/load max). If we were
> to network a bunch of PCI machines running 8-bit SCSI HD adapters vs. 512meg
> RamDisks used for optimized network traffic or controllers with large caches,
> we could see bandwidth waste vs. full saturation quickly. When I found out
> they
> opted to cache their entire MDB in a RamDisk on their server for maximum
> speed, I was pretty amazed. I thought it was an interesting approach to
> client/server applications! (They had a UPS just in case along with a software
> controlled procedure to backup the MDB if power went out :)
I think what you say may work except that in a network with high
capacity servers, ram disk will get expensive so normally, the bottleneck
will be where the HD is.
> I always wondered why 3Com never warned users of their PC Card solutions, yet
> warned them of the 3C515TX series 10/100 ISA adapters. The 16-bit PC Card
> spec has a theoretical limit of ~7.84MB/sec.
Maybe because the PC Card solutions were all bought from
MegaHertz/USR and not part of their original Etherlink card family?
> > Yep, that was why I sent my 3CCFEM556B which is PCMCIA and got
> >a 3CCFEM656B after a year. I actually made $100 from 3Com in doing that
> >since the PCMCIA version costed close to $400 last year and the CardBus
> >version is $229 with rebates.
>
> Hehe. :) That's an interesting way of doing things! I should look into
> evaluation/trade-up programs more closely.
It wasn't a trade-up program. I had to use the One Year
Satisfaction guarantee on the side of the box on them.
> > I'm actually a fan of Telebit more than USR since the PEP days
> >back in 1987 when I was one of the first persons who tested Telebit, USR,
> >Hayes modems before there was even a 9600 standard. Telebit was more
> >robust and 18k compared to HST which was only 9600. That's why even in
> >Austraia when there were regulations that all modems were Telebit.
> >Concord was the one who came out with v32 in 1989 or 90. I agree with you
> >that RealPort is a neat idea except for the size of the thing and also the
> >modem isn't as good as 3Com's. Intel has the best networking cards though
> >but the only problem is that their PCMCIA offerings are based on Xircom's
> >guts and their modems are Xircom's as well.
>
> Only reason I used Telebits and their PEP protocol was to retrieve
> Usenet/email traffic for a BBS back in 1991-2. Wasn't PEP rated at
> 19.2k? (Maybe I missed something, but I wasn't the biggest Telebit fan
> in the world. You're probably way more up on the spec than I....) My
> traffic carrier didn't offer HST connections for the traffic, and
> after seeing how LARGE the downloads were, I could completely
> understand why. I agree with their superiority (Telebit), but all the
> local BBS's and even the LD ones I called were USR/HST (HST was
> asymmetrical, ugh). Wow, that's interesting you were one of the first
> people to test early implementations of their modem products back
> then. In 1987, I couldn't afford anything close to top of the line. To
> be able to beta test modems from the big boys (Hayes was huge then,
> but their Ultra 9600bps proprietary protocol failed badly, IMHO). I
> relied on my 300bps modem of power for years until I saved for a more
> powerful USR/HST tower of power. On a BBS level, USR reigned surpreme
> for years. I couldn't believe how fast that 2400/9600HST "round light"
> model of modem I had was! I've stuck with USR products ever since
> then. My last HST dominant connection period was their their 16.8
> variant of the protocol. That was their last hurrah. Heck, I use a
> Courier V.Everything now when I have to do modem communications with
> my desktop.
Telebit was 18k at the start and then I think they went to 23.6k
or something with TurboPEP. HST started out as 9600 then 14.4k when v32
was out then the 16.8k HST before 28.8k was out. Then they got out of the
HST naming game and went with X2. hehe. The only reason USR was superior
in SysOps was because Telebit's SysOp program costs $695 while USR only
charged $250. I still have my Telebit TrailBlazer on the bookshelf that I
haven't touched for like 10 years. I had a wide footprint USR 14.4k Dual
Standard that I traded for the v.small Dual Standard 28.8k v32bis with Fax
and then did the X2 upgrade. This is slightly better than the
v.everything since the v.everything didn't have daughterboards like mines
did. The Telebit seemed to handle any bad line condition without
problems. Were you on FidoNet before because your name sounds familiar.
Cheers,
Vince - vince_at_MCESTATE.COM - vince_at_GAIANET.NET ________ __ ____
Unix Networking Operations - FreeBSD-Real Unix for Free / / / / | / |[__ ]
GaiaNet Corporation - M & C Estate / / / / | / | __] ]
Beverly Hills, California USA 90210 / / / / / |/ / | __] ]
HongKong Stars/Gravis UltraSound Mailing Lists Admin /_/_/_/_/|___/|_|[____]
This archive was generated by hypermail 2.1.3 : Thu Jan 23 2003 - 09:55:27 EST