ASTEC BBS V.90 Report
Updated January 8, 1999, new data in RED

For those Sysop's interested in direct dial V.90 support, for a small BBS, here is ASTEC's report on V.90 performance during the past 140 days, since its installation on our TF Server.

Method of V.90 Support

The 3Com/USR Courier ISDN I-modem V.Everything is used, running the latest firmware. It is interfaced to TF Server via a 230.4kbps DTE. The I-modem has a built-in NT1 connection and our telco connection is a CAT 5 cable directly from our service entrance box to the NT1 jack. No punch-down blocks are involved and all modems use their own cable... unused pairs within a cable are not shared.

The I-modem Configuration Manager for Macintosh appears to be a port of a Windows application. Pretty crude, but easy to setup for any Sysop type. ISDN is provided by Bell Atlantic in our area. The I-modem supports 56k X2 and V.90, plus all earlier protocols, including the old HST and V.FC 28.8. Unfortunately, it doesn't support K56Flex, so "V.Everything" isn't quite the case. We'd happily give up HST and V.FC for K56Flex, but I don't think USR would. ;-)

Observed Connect Speeds and File Transfer Performance

57 ASTEC clients (27%) currently have V.90 and others are considering upgrading from 28.8k or 33.6k modems, soon.

To date, 1781 V.90 calls have been monitored, with the lowest DCE being 40000 bps and the highest being 49333 bps. The low figure is from a client with known poor connect quality, who previously connected at 26400 bps with their 33.6k modem. We have disregarded their connections because of this. We have not seen any client connections above 49333 bps. Our local loop testing can negotiate 50666 and sometimes 52000, but these high DCE's are not included in our averages. The average caller DCE is currently 46666 bps. The average has increased due to two factors: 1.) Better modem initialization strings have been developed and 2.) telco line problems at some client facilities have been fixed.

In reviewing the 140 day period, it's interesting to note that V.90 users call more frequently, but are online for shorter periods of time because of the higher speed. Does this mean that V.90 users are better informed because of more frequent calls? More frequent calls are possible because less time is expended per call?

File transfer speed ranges from 4879 cps to 5493 cps for compressed files, depending on the file and connect speed. The average improvement is 53.8% in file transfer speed, over prior file transfer speed logs. The non-V.90 average for 31.2 and 33.6k connections was 3426 cps. The current V.90 average is 5279 cps.

Do 33.6k callers get better (or 33.6k) calling into the I-modem?

Not in our case, because our analog lines are so clean. However, it certainly could be a factor for other installations where the ISDN connection would resolve "the last mile" difficulties a BBS may be having, because of its physical distance to its CO, or other analog imparements POTS presents many BBS host modems.

User Opinion

The V.90 implementation has been well received and ASTEC clients who are using it have been very vocal about the greater speed and reduced file transfer times. ASTEC clients are not typically "vocal" about such things, but anything that gets them back to work quicker, is always appreciated.

There is surprising sentiment about ASTEC's file transfer speed, versus Internet file transfer speed, as observed by clients who are doing both. It's apparent that many of our clients did not realize how slow the Internet is during business hours, or took it for granted that this is all V.90 really had to offer. They are very surprised by V.90 thruput performance, with the Internet connection out of the picture.

Problems

The MDM# resource had to be modified to support the new connect message rates, since "Connect" was only coming up most of the time and users had no idea as to what their connect speed was, nor was TF Server telling us what was going on, DCE wise. Both the Server and User require the new MDM# resource and I understand Rusty will include this in TF 5.7 when it is final.

Connections clearly favor USR products, as "Rockwell" based V.90 modems don't connect as high, including our own tests in our local loop with our Motorola ModemSURFR V.90 models. Perhaps as V.90 matures, this will not be a factor.

Yes, I am aware of comments about USR's tendency to show high initial connect rates. We use Smartcom II v4.0 during these tests, since we've set it up to interrogate the modem and report call parameters as soon as we disconnect, looking specifically for speed shifts. We've done this into a number of ISP's, where speed shifts are evident with some calls, but shifts are small and go up as well as down, that we've seen. I think the overly high initial DCE reports are overly exaggerated, based on what we're seeing.

The I-modem has a problem answering and negotiating an ISDN call and we haven't figured that out yet. With only two callers with ISDN, we're not being pressed to solve it.

We don't call out on the I-modem, but use a second ISDN modem, a 3ComImpact IQ (ISDN only, can't connect to anything else), suggested by Ward McFarland (MegaWolf), which works perfectly with our Mindspring ISP account, actually PSINet since Mindspring leases their equipment in this area. A 2B 128k ISDN TCP/IP connection to SpiderNet in the Netherlands results in compressed file transfers of 15kcps. SpiderNet remains the fastest TF Server we have found, in spite of its European location.

Too bad the two ISDN modems can't be connected at the same time to the same ISDN line, like regular modems can, but they can't. The I-modem establishes its links with the CO quickly when plugged back in, but the 3ComImpact IQ sometimes takes several minutes to do the same thing. When we first started using it, we thought it was perhaps broken, but it's just simply very slow negotiating with the CO switch, compared to the I-modem.

Overal

I-modem performance has been flawless in negotiating all user calls (except ISDN), not just V.90 or X2. However, it should be understood that the majority of ASTEC clients are using modems recommended or supplied by ASTEC, which have been Hayes OPTIMA (up until early 1997), USR V.Everything and until they stopped producing modems, the Motorola ModemSURFR 56k. We are not subjected to calls from low quality, vanilla modems, thankfully. Also, ASTEC is only 2200 feet of copper to its CO, so its lines have always been very clean and involving ISDN presented no line length concerns.

As V.90 use increases, we'll have to add another ISDN line, or look at other options to support more V.90 callers. Right now, only the I-modem can do this economically, for a small customer support BBS such as ASTEC. And, all present indications are that it will do it very well.

Rod Paine
ASTEC Company, Inc.
September 7, 1998
Updated January 8, 1999