Ascend Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
(ASCEND) 5.0Ap33 results here...
We've just now completed 24 hours of 5.0Ap33 on 3 of our Maxen.
2 are CT1 (one has 3, the other has 2) and with 2 PRIs.
All modems are Rev F K56 and all Maxens are connected directly to
Cisco 2501 routers, no RIP or OSPF; all static routes.
All the CT1s have LQM enabled, 150 MIN 6000 MAX. The PRI unit has it
disabled because I have some WIndows CE users on that PoP.
We upgraded from 5.0Ap23 which had been stable, if not spectacular.
Existing problems with Ap23 were dropped connections and burp-tones
screwing up cheap modems, affecting less than 1% of our user base.
We also know that some of the dropped connections are due to our
Friends At The Telco Who Won't Upgrade Their Damned Switch- we're
actually swamping voice channels. In some areas at peak times you
can't get a dial tone on a voice call, so they start kicking our
users off.
What we've seen:
Upside:
Across the board, one jump in connect speed. The fringe-area users
(k56 or 33.6 modems, too far from the telco for K56 speeds) got a
1-place jump: i.e. 21.6 connects went to 24, 24 went to 26.4, 26.4
went to 28.8, 28.8 went to 31.2- 3 users that had ZOOM K56 modems
and were getting 31.2-35 went to 45-48.
Users already getting K56 speeds have noticed little or no difference,
although we are getting steady 50K in downtown Oxford now; it used
to vary between 40-48K. These are all CT1s. The PRI PoP has shown no
difference, nor did it need to. It's been a steady 48-50 for months.
Super-cheap 33.6 and K56 modem users have reported less trouble.
Throughput seems boosted by about 3% on our pre-compressed test
files.
Downside:
On the downside, we have quite a bit more dropped connections, but I
can't be sure it's not Our Friends At The Telco as usage has been
high since the upgrade, and ALL of our x2 users have experienced
problems. Turning off v.42bis has fixed them (so far) but it was not
necessary at Ap23.
My connection from home, which I usually leave up all the time, has
dropped 6 times today. It usually drops about once every 2 days or so.
I'm running Zoom K56, August flash. There is a September-dated flash
on their website, but the rev looks earlier than mine, so I don't
know whether to try that or not.
On an issue that I've not seen discussed:
On our stuff, I see a 1-jump difference in connect speed (on marginal
connections) if I have Win95 dial out automatically after starting some
TCP/IP program. If I just dial, with no other software running, I get
say, 26.4, and if I fire up Netscape and let that "trigger" the DUN, I
get 24. Completely repeatable. Repeatable on more that 15 users that
we have as our "smart user base"(tm). Could some of you folks try
this and see if it works for you as well? I'm extremely curious about
this as I can see no reason why it should matter, unless DUN is even
more broken than I suspected..... BTW, I have the newest one from
MS, but all Win95 builds that we've tested exhibit this behavior.
More data as it comes in; we're watching all this stuff.
'Shot
++ Ascend Users Mailing List ++
To unsubscribe: send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd: <http://www.nealis.net/ascend/faq>
Follow-Ups: