[26778] in resnet
Re: Bandwidth Management - Dynamic Allocation?
daemon@ATHENA.MIT.EDU (Doughty, Marc)
Tue Sep 27 17:37:36 2011
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary=bcaec52163e91ee68804adf30d25
Message-ID: <CAEPWjzvL0HanQqnApTrHY2uWL0F8d_Bfuab0TE01AJMiFq+bow@mail.gmail.com>
Date: Tue, 27 Sep 2011 17:35:15 -0400
Reply-To: Resnet Forum <RESNET-L@listserv.nd.edu>
From: "Doughty, Marc" <marc_doughty@BROWN.EDU>
To: RESNET-L@listserv.nd.edu
In-Reply-To: <CAOAqwrdnGTUyWqZ4rE7xLUGseDnGu4FLfBFrO8BTr+r_fSXiYQ@mail.gmail.com>
--bcaec52163e91ee68804adf30d25
Content-Type: text/plain; charset=ISO-8859-1
Cool, this is what I would expect. With 'equal allocation', you can
'overcommit' by quite a bit, right? If you have 1gbit and 1000 users, I
imagine you probably shouldn't limit each user to 1mbit, because all the
users are never at all the computers at once. Has anyone experimented with
this yet? 'Equal allocation' at overcommited rates, so instead of
'bandwidth/users' something like 'bandwidth/(users/5)'? I'll bet there's a
sweet-spot between 3 and 20 'shares' of bandwidth that keeps speed-demons
happy without letting common tasks starve the community.
NetEQ does much 'smarter' things than I could ever figure out for myself.
Their stuff sounds like it intelligently prevents
bufferbloat<http://en.wikipedia.org/wiki/Bufferbloat>and fairly
divvies-up unused bandwidth using various weighted queues.
- Marc Doughty
"If you aren't sure who is the give-way vessel, you are the give-way
vessel."
On Tue, Sep 27, 2011 at 2:16 PM, Greg Briggs <briggsgb@plu.edu> wrote:
> You have stated your reservations, but I am still going to have
> to recommend the NetEqualizer. Carving up the bandwidth equally will mean
> that the user perception of the internet connection will be poor even when
> you have bandwidth to spare. It makes more sense to have a device that can
> maximise the users perception of a connection. Here are some example
> scenarios.
>
> NetEQ when utilization is low, and it is not doing anything:
> User perception of Skype like services: Good
> User perception of Netflix like services: Good
> User perception of large file downloads: Good
> User perception of "ajaxie" webpages that constantly update some doodad on
> the page:Good
> User Perception of games: Good
>
> Equally allocated bandwidth when utilization is low:
> User perception of Skype like services: OK as long as the user is not doing
> anything else.
> User perception of Netflix like services: OK as long as long as the user is
> not doing anything else.
> User perception of large file downloads: Slow all of the time regardless of
> where the user is downloading the file from.
> User perception of "ajaxie" webpages that constantly update some doodad on
> the page: OK
> User perception of games: OK as long as the user is not doing anything
> else. That is until the game needs to download custom content from a server,
> then the user has to wait to enter the next round because of the hard rate
> limit.
>
> NetEQ when utilization is high and penalizing the top flows:
> User perception of Skype like services: Good
> User perception of Netflix like services: Good - The caching bar at the
> bottom should be slightly delayed, but the video shouldn't skip. The user
> is unlikely to notice.
> User perception of large file downloads: Good - The file is delayed a bit,
> but will still download relatively quickly compared to a hard bandwidth cap.
> The user is unlikely to notice.
> User perception of "ajaxie" webpages that constantly update some doodad on
> the page: Good
> User perception of games: Good downloading content between rounds might be
> a tiny bit slower, but fast compared to a hard rate limit.
>
> Equally allocated bandwidth when utilization is high:
> User perception of Skype like services: OK as long as the user is not doing
> anything else.
> User perception of Netflix like services: OK as long as long as the user is
> not doing anything else.
> User perception of large file downloads: Slow all of the time regardless of
> where the user is downloading the file from.
> User perception of "ajaxie" webpages that constantly update some doodad on
> the page: OK as long as the user is not doing anything else.
> User perception of games: OK as long as the user is not doing anything
> else. That is until the game needs to download custom content from a server,
> then the user has to wait to enter the next round because of the hard rate
> limit.
>
> As far as the P2P thing is concerned. While I too realized that
> theoretically P2P would be favored, in practice it wasn't really noticeable.
> If you wish, you can use connection limits to deal with this.
>
> On last thing to note: On Obama's inauguration day, the NetEQ at PLU was
> able to tame the ridiculous number of live streams of the event without
> me intervening to change settings. The only problems reported turned out to
> be bandwidth problems on the other end.
>
> I hope you find this useful.
>
> Greg Briggs
> Network Engineer
> Information & Technology Services
> Pacific Lutheran University
>
>
> On Fri, Sep 23, 2011 at 3:09 PM, Randall Watanabe <randallw@hawaii.edu>wrote:
>
>> Hello All,
>>
>> We currently have a Packetshaper that is at the limits of its capabilities
>> and we are looking to upgrade. I've searched through this and various
>> other
>> lists for information on solutions from Procera, Exinda, and the
>> NetEqualizer, but was wondering if there is a simple solution out there
>> that
>> basically just takes all the devices online, and carves up the available
>> bandwidth equally.
>>
>> We currently use our Packetshaper in this manner via dynamic partitioning
>> and if there is a low cost solution that manages bandwidth in this way
>> without doing the layer7 application analysis, that would be very
>> tempting.
>>
>> I had originally hoped that the NetEqualizer took this approach, but my
>> understanding is that it allocates the bandwidth equally based on flow,
>> not
>> device (IP). It seems to me that that approach would favor P2P type
>> applications that open up a bunch of connections at relatively low
>> bitrates.
>>
>>
>> We have ~4000 users and anticipate managing somewhere around 1Gbps in
>> bandwidth within a few , so any solution we look at would have to be able
>> to
>> manage those kinds of numbers.
>>
>> Thanks in advance for any responses.
>>
>> Randall Watanabe
>> ResNet Computer Specialist
>> UH Student Housing Services
>>
>> ___________________________________________________
>> You are subscribed to the ResNet-L mailing list.
>>
>> To subscribe, unsubscribe or search the archives,
>> go to http://LISTSERV.ND.EDU/archives/resnet-l.html
>> ___________________________________________________
>>
>
> ___________________________________________________ You are subscribed to
> the ResNet-L mailing list.
>
> To subscribe, unsubscribe or search the archives, go to
> http://LISTSERV.ND.EDU/archives/resnet-l.html___________________________________________________
>
___________________________________________________
You are subscribed to the ResNet-L mailing list.
To subscribe, unsubscribe or search the archives,
go to http://LISTSERV.ND.EDU/archives/resnet-l.html
___________________________________________________
--bcaec52163e91ee68804adf30d25
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Cool, this is what I would expect. With 'equal allocation', you can=
'overcommit' by quite a bit, right? If you have 1gbit and 1000 use=
rs, I imagine you probably shouldn't limit each user to 1mbit, because =
all the users are never at all the computers at once. Has anyone experiment=
ed with this yet? 'Equal allocation' at overcommited rates, so inst=
ead of 'bandwidth/users' something like 'bandwidth/(users/5)=
9;? I'll bet there's a sweet-spot between 3 and 20 'shares'=
of bandwidth that keeps speed-demons happy without letting common tasks st=
arve the community.<br>
<br>NetEQ does much 'smarter' things than I could ever figure out f=
or myself. Their stuff sounds like it intelligently prevents <a href=3D"htt=
p://en.wikipedia.org/wiki/Bufferbloat">bufferbloat</a> and fairly divvies-u=
p unused bandwidth using various weighted queues.<br clear=3D"all">
<br>- Marc Doughty<br>"If you aren't sure who is the give-way vess=
el, you are the give-way vessel."<br>
<br><br><div class=3D"gmail_quote">On Tue, Sep 27, 2011 at 2:16 PM, Greg Br=
iggs <span dir=3D"ltr"><<a href=3D"mailto:briggsgb@plu.edu">briggsgb@plu=
.edu</a>></span> wrote:<br><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
You have stated your reservations, but I am still going to have to=A0recomm=
end=A0the NetEqualizer. =A0Carving up the bandwidth equally will mean that =
the user perception of the internet connection will be poor even when you h=
ave bandwidth to spare. =A0It makes more sense to have a device that can ma=
ximise the users perception of a connection. =A0Here are some example scena=
rios.<div>
<br></div><div>NetEQ when utilization is low, and it is not doing anything:=
</div><div>User perception of Skype like services: Good</div><div>User perc=
eption of Netflix like services: Good</div><div>User perception of large fi=
le downloads: Good</div>
<div>User perception of "ajaxie" webpages that=A0constantly=A0upd=
ate some doodad on the page:Good</div><div>User Perception of games: Good</=
div><div><br></div><div>Equally=A0allocated bandwidth when utilization is l=
ow:</div>
<div><div>User perception of Skype like services: OK as long as the user is=
not doing anything else.</div><div>User perception of Netflix like service=
s: OK as long as long as the user is not doing anything else.</div><div>
User perception of large file downloads: Slow all of the time=A0regardless=
=A0of where the user is downloading the file from.</div><div>User perceptio=
n of "ajaxie" webpages that=A0constantly=A0update some doodad on =
the page: OK</div>
</div><div>User perception of games:=A0OK as long as the user is not doing =
anything else.=A0That is until the game needs to download custom content fr=
om a server, then the user has to wait to enter the next round because of t=
he hard rate limit.</div>
<div><br></div><div>NetEQ when utilization is high and penalizing the top f=
lows:</div><div><div>User perception of Skype like services: Good</div><div=
>User perception of Netflix like services: Good - The caching bar at the bo=
ttom should be slightly delayed, but the video shouldn't skip. =A0The u=
ser is unlikely to notice.=A0</div>
<div>User perception of large file downloads: Good - The file is delayed a =
bit, but will still download relatively quickly compared to a hard bandwidt=
h cap. =A0The user is unlikely to notice.=A0</div><div>User perception of &=
quot;ajaxie" webpages that=A0constantly=A0update some doodad on the pa=
ge: Good</div>
</div><div>User perception of games:=A0Good downloading content between rou=
nds might be a tiny bit slower, but fast compared to a hard rate limit.</di=
v><div><br></div><div>Equally allocated bandwidth when utilization is high:=
</div>
<div><div><div>User perception of Skype like services: OK as long as the us=
er is not doing anything else.</div><div>User perception of Netflix like se=
rvices: OK as long as long as the user is not doing anything else.</div>
<div>User perception of large file downloads: Slow all of the time=A0regard=
less=A0of where the user is downloading the file from.</div><div>User perce=
ption of "ajaxie" webpages that=A0constantly=A0update some doodad=
on the page: OK=A0as long as the user is not doing anything else.</div>
</div></div><div>User perception of games:=A0OK as long as the user is not =
doing anything else.=A0That is until the game needs to download custom cont=
ent from a server, then the user has to wait to enter the next round becaus=
e of the hard rate limit.</div>
<div><br></div><div>As far as the P2P thing is concerned. =A0While I too re=
alized that theoretically P2P would be favored, in practice it wasn't r=
eally noticeable. =A0If you wish, you can use connection limits to deal wit=
h this. =A0</div>
<div><br></div><div>On last thing to note: =A0On Obama's=A0inauguration=
=A0day, the NetEQ at PLU was able to tame the=A0ridiculous=A0number of live=
streams of the event without me=A0intervening=A0to change settings. =A0The=
only problems reported turned out to be bandwidth problems on the other en=
d. =A0</div>
<div><br></div><div>I hope you find this=A0useful.</div><div><br></div><div=
>Greg Briggs</div><div>Network Engineer</div><div>Information & Technol=
ogy Services</div><div>Pacific Lutheran University</div><div><br></div><div=
>
<br>
<div class=3D"gmail_quote"><div class=3D"im">On Fri, Sep 23, 2011 at 3:09 P=
M, Randall Watanabe <span dir=3D"ltr"><<a href=3D"mailto:randallw@hawaii=
.edu" target=3D"_blank">randallw@hawaii.edu</a>></span> wrote:<br></div>=
<div>
<div></div><div class=3D"h5"><blockquote class=3D"gmail_quote" style=3D"mar=
gin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hello All,<br>
<br>
We currently have a Packetshaper that is at the limits of its capabilities<=
br>
and we are looking to upgrade. =A0I've searched through this and variou=
s other<br>
lists for information on solutions from Procera, Exinda, and the<br>
NetEqualizer, but was wondering if there is a simple solution out there tha=
t<br>
basically just takes all the devices online, and carves up the available<br=
>
bandwidth equally.<br>
<br>
We currently use our Packetshaper in this manner via dynamic partitioning<b=
r>
and if there is a low cost solution that manages bandwidth in this way<br>
without doing the layer7 application analysis, that would be very tempting.=
<br>
<br>
I had originally hoped that the NetEqualizer took this approach, but my<br>
understanding is that it allocates the bandwidth equally based on flow, not=
<br>
device (IP). =A0It seems to me that that approach would favor P2P type<br>
applications that open up a bunch of connections at relatively low bitrates=
.<br>
<br>
<br>
We have ~4000 users and anticipate managing somewhere around 1Gbps in<br>
bandwidth within a few , so any solution we look at would have to be able t=
o<br>
manage those kinds of numbers.<br>
<br>
Thanks in advance for any responses.<br>
<br>
Randall Watanabe<br>
ResNet Computer Specialist<br>
UH Student Housing Services<br>
<br>
___________________________________________________<br>
You are subscribed to the ResNet-L mailing list.<br>
<br>
To subscribe, unsubscribe or search the archives,<br>
go to <a href=3D"http://LISTSERV.ND.EDU/archives/resnet-l.html" target=3D"_=
blank">http://LISTSERV.ND.EDU/archives/resnet-l.html</a><br>
___________________________________________________<br>
</blockquote></div></div></div><br></div>
___________________________________________________
You are subscribed to the ResNet-L mailing list.
<div><div></div><div class=3D"h5"><p>
To subscribe, unsubscribe or search the archives,
go to <a href=3D"http://LISTSERV.ND.EDU/archives/resnet-l.html" target=3D"_=
blank">http://LISTSERV.ND.EDU/archives/resnet-l.html</a>
___________________________________________________
</p></div></div></blockquote></div><br>
___________________________________________________
You are subscribed to the ResNet-L mailing list.
<p>
To subscribe, unsubscribe or search the archives,
go to http://LISTSERV.ND.EDU/archives/resnet-l.html
___________________________________________________
--bcaec52163e91ee68804adf30d25--
