Re: [81attendees] sucky Delta hotel network (and bufferbloat)

Jim Gettys <jg@freedesktop.org> Tue, 02 August 2011 00:48 UTC

Return-Path: <gettysjim@gmail.com>
X-Original-To: 81attendees@ietfa.amsl.com
Delivered-To: 81attendees@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id AA1631F0C49 for <81attendees@ietfa.amsl.com>; Mon, 1 Aug 2011 17:48:09 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.599
X-Spam-Level:
X-Spam-Status: No, score=-3.599 tagged_above=-999 required=5 tests=[AWL=0.000, BAYES_00=-2.599, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.30]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ho+lxoX1R8K5 for <81attendees@ietfa.amsl.com>; Mon, 1 Aug 2011 17:48:08 -0700 (PDT)
Received: from mail-vw0-f44.google.com (mail-vw0-f44.google.com [209.85.212.44]) by ietfa.amsl.com (Postfix) with ESMTP id E7EF421F851A for <81attendees@ietf.org>; Mon, 1 Aug 2011 17:47:49 -0700 (PDT)
Received: by vws12 with SMTP id 12so5860011vws.31 for <81attendees@ietf.org>; Mon, 01 Aug 2011 17:47:56 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=sender:message-id:date:from:organization:user-agent:mime-version:to :cc:subject:references:in-reply-to:content-type :content-transfer-encoding; bh=HYeGRlYNzhV/YIicjfBZFf02vBz99m7M3YTs+1HOd3o=; b=vvmKrr2wSPNvPK00yqCGuGoGbQUnT5qf8WdjuFdJEruU2bQM1B+UGLdvsL3lRtiMQe mobA0LRuGIpGEUouJiD7uX/NivUTPb66G+7gjTVxH4SIoK+SHX+6sWLh+HpeOL5EvwKb 0xvbSWsrOieIyxPfSMUihX9sH36AFZFMJHZf4=
Received: by 10.52.117.99 with SMTP id kd3mr4912032vdb.388.1312246076497; Mon, 01 Aug 2011 17:47:56 -0700 (PDT)
Received: from [10.0.0.3] (c-24-218-177-117.hsd1.ma.comcast.net [24.218.177.117]) by mx.google.com with ESMTPS id dn12sm551811vcb.40.2011.08.01.17.47.53 (version=SSLv3 cipher=OTHER); Mon, 01 Aug 2011 17:47:54 -0700 (PDT)
Sender: Jim Gettys <gettysjim@gmail.com>
Message-ID: <4E374939.4090903@freedesktop.org>
Date: Mon, 01 Aug 2011 20:47:53 -0400
From: Jim Gettys <jg@freedesktop.org>
Organization: Bell Labs
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:5.0) Gecko/20110627 Thunderbird/5.0
MIME-Version: 1.0
To: John C Klensin <john-ietf@jck.com>
References: <4E2CC532.3090209@sunet.se> <4E2CCE72.3010500@sunet.se> <4E2E2236.2060808@sunet.se> <87EC2139-C7C2-4113-97E2-9EB9DA2406EA@juniper.net> <alpine.BSF.2.00.1107262259400.80739@joyce.lan> <9D00CDB7-2C04-446E-8D53-0552BAEE22EE@juniper.net> <633EAF31-607E-4FEF-B502-5FFAD89BF01A@juniper.net> <4E371370.40804@freedesktop.org> <00E8AAF99E25FF49A1F55E9A0CD19EBCB93DF4E65B@SGSINSXCHMBSA2.sg.alcatel-lucent.com> <CE4DBD9C-E366-400B-9B39-F86D591F25AB@juniper.net> <98F2DACCFC679BFB8A8E3592@PST.JCK.COM>
In-Reply-To: <98F2DACCFC679BFB8A8E3592@PST.JCK.COM>
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7bit
Cc: 81attendees@ietf.org
Subject: Re: [81attendees] sucky Delta hotel network (and bufferbloat)
X-BeenThere: 81attendees@ietf.org
X-Mailman-Version: 2.1.12
Precedence: list
List-Id: IETF 81 Attendee List <81attendees.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/81attendees>, <mailto:81attendees-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/81attendees>
List-Post: <mailto:81attendees@ietf.org>
List-Help: <mailto:81attendees-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/81attendees>, <mailto:81attendees-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 02 Aug 2011 00:48:10 -0000

On 08/01/2011 05:37 PM, John C Klensin wrote:
>
> --On Monday, August 01, 2011 22:18 +0100 Maciek Konstantynowicz
> <maciek@juniper.net> wrote:
>
>> Thanks for the pointers, will digest.
>>
>> The hotel manager I spoke to, didn't even know what IETF is
>> about, and didn't expect any excessive WiFi demands breaking
>> hotel's Internet access service. Maybe as a community, going
>> forward we could provide more candid feedback re Internet/WiFi
>> performance during IETF gatherings. Or better give hotel mgmt
>> heads-up about coming tsunami of Internet users... so that
>> hotel could address it pro-actively..
>>
>> The business incentive for the hotel would be to avoid giving
>> 20% discounts if things don't work out as per my case :) +
>> customer satisfaction.
> Sigh.
> I used the network in the Delta quite a lot and didn't
> experience much of a problem, but kept my expectations set for
> "hotel" rather than "IETF".  I also mostly used the wired
> Ethernet -- it may be just anecdotal, but I've generally found
> that wired Ethernet is much less likely to be problematic than
> WiFi if neither is set up by experts.
>
> While warning hotels that we are a collection of heavy Internet
> users and that it is a good idea to make sure that their IP
> vendors know that they need to tune things up to avoid
> embarrassment, the bottom line is that the only times folks seem
> to have zero complaints about hotel networks are where the hotel
> infrastructure starts out above average and is then hooked into
> the IETF network.  It simply isn't plausible to believe that
> interconnection will occur with other than the primary hotel.
> It would be good to know in advance how likely it is even with
> the primary hotel but those opting for other hotels should
> arrive with the assumption that they will be dealing with a
> generic hotel network and What You Get is What You Get quality
> (whether it is free and worth what one pays for it or not).
>
>

Yes, exactly.

You suffer a lot less on ethernet, as you take one possible point of
bloat out of the path.

Windows, by default, won't saturate a 100Mbps connection (it's tuned to
stay just below 90Mbps, as far as I can see). Mac and Linux will
saturate anything to a gigabit without thinking with even a single TCP
connection, given enough time for them to ramp up.  Even if Mac and
Linux saturate the ethernet link and fill those queues in the
driver/router, at least the amount of time they represent at high
bandwidth is much, much less.

So your queues build in the hotel's router, rather than backing up in
the wireless router and your laptop, at least when plugged in usually.

Wireless, going through multiple walls and also being shared often
guarantees your queues build in that wireless hop, where bloat is often
much worse than the uplink, and shared among others to boot.

So good practical advice until this is sorted out is to plug in to
ethernet whenever possible, and/or carry a high speed router of your own
to pug into that ethernet in your room, to try to guarantee the
bottleneck link is at the hotel backhaul.

Same thing goes at home: the home router/laptop link in your house is
probably more bloated than your broadband connection (and you can
bandwidth shape that to keep those queues small), so if you ensure your
wireless bandwidth is always above your broadband bandwidth you can have
reliable low latency.  But in houses like mine with chimneys, this can
be easier said than done.

There is a lot you can do about bloat once you understand it by moving
the bottleneck link to where it may be less problematic.
                        - Jim