[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: bandwidth mobility



Dana,

My understanding from talking with IESG members is that use of RSVP is being
discouraged except for limited cases. Have you heard anything to
this effect? 

		jak
		
>From: "Dana L. Blair" <dblair@cisco.com>
>To: "tim clifford" <tjc@lacunanet.net>, "James Kempf" <James.Kempf@Sun.COM>, 
<more@psg.com>, <E.Crane@motorola.com>
>Subject: RE: bandwidth mobility
>Date: Tue, 21 Aug 2001 07:16:53 -0400
>X-Priority: 3 (Normal)
>X-MSMail-Priority: Normal
>X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200
>Importance: Normal
>
>>   -----Original Message-----
>>   From: owner-more@ops.ietf.org [mailto:owner-more@ops.ietf.org]On Behalf
>>   Of tim clifford
>>   Sent: Monday, August 20, 2001 11:28 PM
>>   To: Dana L. Blair; James Kempf; more@psg.com; E.Crane@motorola.com
>>   Subject: RE: bandwidth mobility
>>
>>
>>   all
>>   ok, now i think i've finally lost the thread of where the
>>   requirement lies.
>>   as a quick, high level statement (i'll try to do a better job or please
>>   someone save me  ;-)
>>   1. at some general level the "system" (subscriber, mobile device, access
>>   net, mobile core, peer networks and applications) must have a
>>   way of being
>>   aware of available bandwidth for a subscriber at any time.
>>   2. the elements of the system should have a way of exchanging
>>   this status
>>   information
>>   3. the elements of the system should have a way of reacting to available
>>   bandwidth and changes in what's available
>>   4. we probably need to define other metrics beyond bandwidth
>>   (e.g., delay,
>>   jitter, loss?)
>
>Implementation of the RSVP protocol between network elements satisfies the
>requirements you have just stated.
>
>Applications should use an RSVP API provided by the host
>operating system whether client or server.  So the source
>of RSVP requests are hosts.  Windows supports RSVP API.  I don't
>know about other OS's.
>
>Switches and Routers within the network react to the requests
>to (a) authorize an application to access the network with
>a certain QoS capability, and (b) manage bandwidth over congested
>links by either (b.1) avoiding congestion or (b.2) prioritizing
>packets over the congested link.
>
>Avoiding congestion can be done by over-provisioning the network
>or denying RSVP requests that would congest the network.
>
>Prioritizing traffic can be done with queuing out the output
>interfaces of hosts, switches, and routers.
>
>thanks,
>Dana
>
>>
>>   a few observations
>>   1. much of this is implied in the architectures out there, at
>>   least in mwif
>>   there are policy managers, resource managers, etc. that seem to have
>>   bandwidth mobility in mind as one type of capability to
>>   support.  however,
>>   its not clear to me that the architectures necessarily support the many
>>   ramifications of 1-4 above
>>   2. Operators may want to think about what this really means to
>>   them.  i'm
>>   not sure the subscriber or web site, for example, care about
>>   the intervening
>>   services a great deal, they might just measure actual goodput and react
>>   accordingly
>>   3. this discussion of cookstoves is interesting but i'm not sure how it
>>   affects the requirement.  if in fact i can't cook a turkey on a
>>   camp stove
>>   then various participants need to recognize the change and react
>>   accordingly - maybe terminate the session and influence the bill
>>   accordingly.
>>   4. finally, i can't resist, imho the point of mobility is that available
>>   wireless bandwidth per user will often be INFINITE compared to available
>>   wired bandwidth per user, i.e., i'm mobile and don't have the option of
>>   stopping, waiting for that brilliantly architected (;-)
>>   operating system,
>>   windows, to boot up, and connecting to a higher speed wlan or
>>   wired port.
>>   so the ratio of some available bw to none is infinity last time
>>   i checked
>>
>>   > -----Original Message-----
>>   > From: owner-more@ops.ietf.org
>>   [mailto:owner-more@ops.ietf.org]On Behalf
>>   > Of Dana L. Blair
>>   > Sent: Monday, August 20, 2001 7:13 PM
>>   > To: James Kempf; more@psg.com; E.Crane@motorola.com
>>   > Subject: RE: bandwidth mobility
>>   >
>>   >
>>   > >   -----Original Message-----
>>   > >   From: owner-more@ops.ietf.org
>>   > [mailto:owner-more@ops.ietf.org]On Behalf
>>   > >   Of James Kempf
>>   > >   Sent: Monday, August 20, 2001 1:08 PM
>>   > >   To: James.Kempf@sun.com; more@psg.com; E.Crane@motorola.com
>>   > >   Subject: RE: bandwidth mobility
>>   > >
>>   > >
>>   > >   Eric,
>>   > >
>>   > >   I agree that the low end will always be attractive to
>>   someone. What
>>   > >   I was trying to say is that I think there are possibilities for
>>   > >   bandwidth management that are not there right now because
>>   of a lack
>>   > >   of mechanism for intertechnology handover and a lack of support
>>   >
>>   > Just want to point out that inter-technology handover is achievable
>>   > for Best Effort data with current Mobile IP standard, rfc2002.
>>   >
>>   > Dana
>>   >
>>   > >   in the mobiles. Whether or not software defined radios ever pan
>>   > >   out is an open question. Intertechnology handover is an  easier
>>   > >   problem to solve I think, but there are still issues. What I'm
>>   > >   trying to say is that I don't think we should structure the
>>   > >   requrirements based on an assumption that bandwidth is always
>>   > >   cheap and limitless, or that it is always constrained and
>>   expensive.
>>   > >   If the currently "hot" 802.11 provider scenerios play
>>   out, bandwidth
>>   > >   will be lots cheaper and more available in certain cases, but
>>   > >   still limited in others.
>>   > >
>>   > >   		jak
>>   > >
>>   > >   >From: Crane Eric-ATCM39 <E.Crane@motorola.com>
>>   > >   >To: "'James Kempf'" <James.Kempf@sun.com>, more@psg.com
>>   > >   >Subject: RE: bandwidth mobility
>>   > >   >Date: Mon, 20 Aug 2001 12:00:46 -0500
>>   > >   >
>>   > >   >James,
>>   > >   >I like your ideas for automated maximum bandwidth.  I can
>>   > >   foresee a user
>>   > >   >simply defining throughput, latency and cost requirements and
>>   > >   "leaving the
>>   > >   >driving to us".
>>   > >   >
>>   > >   >But I don't think we can ignore the scenarios when high
>>   > >   bandwidth simply is
>>   > >   >not available when and where it is wanted.
>>   > >   >
>>   > >   >Please don't forget the other critical resource when mobile -
>>   > >   battery life.
>>   > >   >(Other than in a fossil-fueled automobile, of course.)  The
>>   > >   faster bit-rates
>>   > >   >use up much more power, and the farther you are from a cell
>>   > >   site, the more
>>   > >   >power required to maintain your bit rate.  For example, my
>>   > cellphone's
>>   > >   >battery dies relatively quickly when it is forced to analog mode.
>>   > >   >
>>   > >   >So the user must now deal with two dimensions of cost: money
>>   > and power.
>>   > >   >(Why does everything always come to money and power?)  (o;)
>>   > >   >
>>   > >   >I don't think we can design services applications for the best
>>   > >   any longer.
>>   > >   >Consider this analogy:  Apple Computer designed for the best,
>>   > >   and dictated
>>   > >   >the entire computer design.  Microsoft was stuck with a
>>   > computer design
>>   > >   >implemented by the lowest bidder.  Microsoft had to start
>>   > >   abstracting the
>>   > >   >design immediately, but in so doing, became excellent
>>   > >   architects (business
>>   > >   >practices aside).
>>   > >   >
>>   > >   >Eric K. Crane
>>   > >   >Motorola Global Telecom Solutions Sector
>>   > >   >Advanced Technology
>>   > >   >1501 W. Shure Dr., Arlington Heights, IL, 60004, USA
>>   > >   >847-632-2855
>>   > >   >e.crane@motorola.com
>>   > >   >Pager: 1338881@skytel.com
>>   > >   >
>>   > >   >
>>   > >   >-----Original Message-----
>>   > >   >From: James Kempf [mailto:James.Kempf@Sun.COM]
>>   > >   >Sent: Monday, August 20, 2001 10:40 AM
>>   > >   >To: tjc@lacunanet.net; dblair@cisco.com; E.Crane;
>>   > >   more@psg.com; Chris.Burke
>>   > >   >Subject: RE: bandwidth mobility
>>   > >   >
>>   > >   >Chris,
>>   > >   >
>>   > >   >I think this really depends on the wireless radio
>>   protocol. People
>>   > >   >are talking about 50 Mbps for 802.11 and 100 Mbps for 802.16
>>   > >   (Hyperlan).
>>   > >   >There are tradeoffs, viz speed of movement and number of mobile
>>   > >   >nodes per cell, but I don't think it necessarily follows
>>   > that wireless
>>   > >   >bandwidth will be limited all the time. One could
>>   imagine a smooth
>>   > >   >intertechnology handover mechanism that allowed more bandwidth
>>   > >   >on additional radio protocols as the user slowed down or fewer
>>   > >   >mobile nodes were in the cell. Combined with a software defined
>>   > >   >radio, so that the intertechnology handover did not
>>   require separate
>>   > >   >physical interfaces in the mobile device, the result would scale
>>   > >   >bandwidth to availability and user movement.
>>   > >   >
>>   > >   >		jak
>>   > >   >
>>   > >   >>From: Burke Chris-CCB007 <Chris.Burke@motorola.com>
>>   > >   >>To: "'tim clifford'" <tjc@lacunanet.net>, "Dana L. Blair"
>>   > >   ><dblair@cisco.com>,
>>   > >   >Crane Eric-ATCM39 <E.Crane@motorola.com>, more@psg.com
>>   > >   >>Subject: RE: bandwidth mobility
>>   > >   >>Date: Mon, 20 Aug 2001 00:13:15 -0500
>>   > >   >>
>>   > >   >>My understanding of the camp stove analogy is that
>>   > available wireless
>>   > >   >bandwidth
>>   > >   >per user will always be RELATIVELY scarce (by about two orders
>>   > >   of magnitude)
>>   > >   >
>>   > >   >compared to available wired bandwidth per user. Therefore,
>>   > >   whatever happens
>>   > >   >to
>>   > >   >bandwidth in the future, applications designed to operate near
>>   > >   the limits of
>>   > >   >the
>>   > >   >"wired" internet will ALWAYS be disadvantaged when running
>>   > >   over wireless
>>   > >   >links.
>>   > >   >>
>>   > >   >>However, for some applications, good is good enough. You can
>>   > >   do a lot in
>>   > >   >128kbps - 2Mbps of personal bandwidth. Just not as much as you
>>   > >   can do in
>>   > >   >2Mbps -
>>   > >   >100Mbps of personal bandwidth.
>>   > >   >>
>>   > >   >>Chris Burke
>>   > >   >>Motorola Internet Software and Content Group
>>   > >   >>
>>   > >   >>-----Original Message-----
>>   > >   >>From: tim clifford [mailto:tjc@lacunanet.net]
>>   > >   >>Sent: Friday, August 17, 2001 12:20 PM
>>   > >   >>To: Dana L. Blair; Crane Eric-ATCM39; more@psg.com
>>   > >   >>Subject: RE: bandwidth mobility
>>   > >   >>
>>   > >   >>
>>   > >   >>i've been wanting to address the turkey on a camp stove
>>   > >   thing, and now the
>>   > >   >>rsvp comments
>>   > >   >>
>>   > >   >>first i'm not sure what you mean by a low bw appliance, sure
>>   > >   a mobile phone
>>   > >   >>with a few lines of b&w text is a low bandwidth appliance but
>>   > >   that seems to
>>   > >   >>be only part of the story.  a phone may never be a great high
>>   > >   bw appliance
>>   > >   >>but a few thoughts change that notion
>>   > >   >>- the phone may act as the equivalent of a modem to a high bw
>>   > >   appliance
>>   > >   >>- the ability to deliver high bw to mobile devices will
>>   > >   change the phone,
>>   > >   >or
>>   > >   >>the appliance
>>   > >   >>- the applications will emerge to demand high bw over
>>   camp stoves.
>>   > >   >>
>>   > >   >>i really don't want to start dreaming about applications but
>>   > >   i don't think
>>   > >   >>its hard to start dreaming up applications that would use a
>>   > >   small screen
>>   > >   >>(i.e., mobile phone) and more bw. i'd love to be able to get
>>   > >   live streaming
>>   > >   >>traffic cameras while i'm driving and i honestly believe the
>>   > >   millions of
>>   > >   >>people dreaming up applications will come up with better stuff.
>>   > >   >>
>>   > >   >>so i don't understand the camp stove analogy, unless you're
>>   > >   trying to say
>>   > >   >>that today's mobile phones don't need high bw, to which i of
>>   > >   course agree.
>>   > >   >>
>>   > >   >>as for rsvp, a question.  conventional knowledge that i
>>   > >   understand says
>>   > >   >that
>>   > >   >>its difficult to manage rsvp to scale.  do you have any ideas
>>   > >   about how
>>   > >   >well
>>   > >   >>an rsvp-based mechanism would scale to the populations
>>   anticipated?
>>   > >   >>
>>   > >   >>tim
>>   > >   >>
>>   > >   >>> -----Original Message-----
>>   > >   >>> From: owner-more@ops.ietf.org
>>   > >   [mailto:owner-more@ops.ietf.org]On Behalf
>>   > >   >>> Of Dana L. Blair
>>   > >   >>> Sent: Monday, August 13, 2001 11:30 AM
>>   > >   >>> To: Crane Eric-ATCM39; more@psg.com
>>   > >   >>> Subject: RE: bandwidth mobility
>>   > >   >>>
>>   > >   >>>
>>   > >   >>> >   -----Original Message-----
>>   > >   >>> >   From: owner-more@ops.ietf.org
>>   > >   >>> [mailto:owner-more@ops.ietf.org]On Behalf
>>   > >   >>> >   Of Crane Eric-ATCM39
>>   > >   >>> >   Sent: Wednesday, August 01, 2001 2:18 PM
>>   > >   >>> >   To: more@psg.com
>>   > >   >>> >   Subject: RE: bandwidth mobility
>>   > >   >>> >
>>   > >   >>> >
>>   > >   >>> >   You cannot cook a turkey on a camp stove (at least not
>>   > >   very easily.)
>>   > >   >>> >   Similarly, you cannot run high bandwidth
>>   applications on low
>>   > >   >bandwidth
>>   > >   >>> >   appliances.  So mobility of high bandwidth
>>   applications across
>>   > >   >varying
>>   > >   >>> >   bandwidth capabilities doesn't really make sense in
>>   > >   this context.
>>   > >   >>>
>>   > >   >>> Most "high" bandwidth applications used today (video
>>   for example)
>>   > >   >>> can accomodate low or high bandwidth connections.  Most video
>>   > >   >>> websites today allow the user to tell the application via
>>   > >   >>> a mouse click what kind of connection they have.
>>   > >   >>>
>>   > >   >>> So, even without wireless, Video applications run across
>>   > >   >>> varying bandwidth capabilities.
>>   > >   >>>
>>   > >   >>> The problem that exists for wired and wireless is how can
>>   > >   >>> an application programmatically discover the
>>   bandwidth capability
>>   > >   >>> of a network.  The RSVP protocol standardized by the IETF
>>   > >   >>> can accomplish this since it allows an application to request
>>   > >   >>> a certain bandwidth, if it is denied then it can request
>>   > >   >>> a lower bandwidth.  No more Modem, DSL, Ethernet widgets
>>   > >   >>> on the web page for the user to click.
>>   > >   >>>
>>   > >   >>> >
>>   > >   >>> >   However, I can boil water on a camp stove.  So I think
>>   > >   the practical
>>   > >   >>> >   mobility problem is how to be able to boil water
>>   on any ol'
>>   > >   >>> >   stove.  Things
>>   > >   >>> >   like pages, notifications, chat groups, stock
>>   tickers, etc.
>>   > >   >>> >   should be able
>>   > >   >>> >   to flow over any ol' bandwidth.
>>   > >   >>>
>>   > >   >>> Mobile IP allows this today.
>>   > >   >>>
>>   > >   >>> >
>>   > >   >>> >   One more thing about portable devices, just like
>>   camp stoves,
>>   > >   >>> they have
>>   > >   >>> >   limited fuel.  You can boil water, but not much.  So some
>>   > >   >>> things you may
>>   > >   >>> >   choose to eat cold to save fuel.  A key capability of a
>>   > >   >>> >   portable device may
>>   > >   >>> >   be to define what you want to eat HOT.
>>   > >   >>>
>>   > >   >>> RSVP allows an application to select hot or cold.
>>   > >   >>>
>>   > >   >>> Dana
>>   > >   >>>
>>   > >   >>> >
>>   > >   >>> >
>>   > >   >>> >   Eric K. Crane
>>   > >   >>> >   Motorola Global Telecom Solutions Sector
>>   > >   >>> >   Advanced Technology
>>   > >   >>> >   1501 W. Shure Dr., Arlington Heights, IL, 60004, USA
>>   > >   >>> >   847-632-2855
>>   > >   >>> >   e.crane@motorola.com
>>   > >   >>> >   Pager: 1338881@skytel.com
>>   > >   >>> >
>>   > >   >>> >
>>   > >   >>> >   -----Original Message-----
>>   > >   >>> >   From: tim clifford [mailto:tjc@lacunanet.net]
>>   > >   >>> >   Sent: Wednesday, August 01, 2001 11:07 AM
>>   > >   >>> >   To: Baba S; more@psg.com
>>   > >   >>> >   Subject: RE:
>>   > >   >>> >
>>   > >   >>> >   actually i think you're right on point.
>>   > >   >>> >
>>   > >   >>> >   i suspect that the operator's don't yet know how
>>   > >   they'll deal with a
>>   > >   >>> >   possibly growing community of independent isp's,
>>   or possibly a
>>   > >   >smaller
>>   > >   >>> >   community of clearinghouses/service providers like
>>   > >   wayport, hereuare,
>>   > >   >>> >   megabeam, mobilestar, gric and ipass (the latter two
>>   > >   when they start
>>   > >   >>> >   offering wireless lan access).  its a great
>>   opportunity but
>>   > >   >>> >   also possibly
>>   > >   >>> >   the worst nightmare.
>>   > >   >>> >
>>   > >   >>> >   so i'm not sure we'll get a requirement but we
>>   might try to
>>   > >   >generalize
>>   > >   >>> >   mobility between domains and see where we go.
>>   > >   >>> >
>>   > >   >>> >   a random thought is that it would seem we'll have a
>>   > more severe
>>   > >   >>> >   transition
>>   > >   >>> >   issue when one moves from a 11 Mbps lan/probably
>>   connected to
>>   > >   >>> >   DSL or cable
>>   > >   >>> >   or high bw facilities and then move to some lower speed
>>   > >   >>> mobile operator
>>   > >   >>> >   provided service.  it would seem that lots of applications
>>   > >   >>> would want to
>>   > >   >>> >   adjust.  effectively this is a migration from
>>   fixed to mobile,
>>   > >   >>> >   not just one
>>   > >   >>> >   mobile to another.
>>   > >   >>> >
>>   > >   >>> >   tim
>>   > >   >>> >
>>   > >   >>> >   > -----Original Message-----
>>   > >   >>> >   > From: owner-more@ops.ietf.org
>>   > >   >>> >   [mailto:owner-more@ops.ietf.org]On Behalf
>>   > >   >>> >   > Of Baba S
>>   > >   >>> >   > Sent: Wednesday, August 01, 2001 11:56 AM
>>   > >   >>> >   > To: more@psg.com
>>   > >   >>> >   > Subject: Re:
>>   > >   >>> >   >
>>   > >   >>> >   >
>>   > >   >>> >   > Tim,
>>   > >   >>> >   >
>>   > >   >>> >   > Though I may be apart from your original points,
>>   > >   >>> >   > I'm also interested in Operator's view about for
>>   > user to cross
>>   > >   >>> >   > (or handoff) between administrative domain in more
>>   > >   general sense,
>>   > >   >>> >   > that is even in case of using the same terminal.
>>   > >   >>> >   >
>>   > >   >>> >   > I guess it can be OK if the movement occurs among
>>   > >   operators which
>>   > >   >>> >   > have a contract with the same "broker" or
>>   clearing house.
>>   > >   >>> >   > And I think this scenario could be happen soon, since
>>   > >   >>> >   > an operator/WISP which has only one 802.11AP inside
>>   > of his/her
>>   > >   >>> >   > coffee shop and needs a help of cellular
>>   service provider to
>>   > >   >>> >   > support roaming PDA is appearing.
>>   > >   >>> >   >
>>   > >   >>> >   > Operator's requirement on this may still be "quick
>>   > and secure
>>   > >   >>> >   > handoff"?
>>   > >   >>> >   >
>>   > >   >>> >   > Regards,
>>   > >   >>> >   >
>>   > >   >>> >   > baba//
>>   > >   >>> >   >
>>   > >   >>> >   >
>>   > >   >>> >   > From: "tim clifford" <tjc@lacunanet.net>
>>   > >   >>> >   > Subject: RE:
>>   > >   >>> >   > Date: Wed, 1 Aug 2001 10:56:26 -0400
>>   > >   >>> >   >
>>   > >   >>> >   > > agreed, actually i think you're on to an interesting
>>   > >   >>> >   requirement about
>>   > >   >>> >   > > awareness of network status, current
>>   connection and user
>>   > >   >>> >   authorization.
>>   > >   >>> >   > > however its something i think we need to
>>   tease out a bit
>>   > >   >>> >   > > - its not clear this is an operator requirement or
>>   > >   at least not
>>   > >   >>> >   > exclusively.
>>   > >   >>> >   > > - it seems one would almost certainly be
>>   jumping between
>>   > >   >>> >   administrative
>>   > >   >>> >   > > domains, it would be interesting to hear about
>>   > how operators
>>   > >   >>> >   > would see such
>>   > >   >>> >   > > a need as it carries interesting competitive issues
>>   > >   >>> >   > > - i'm always a bit quesy about videophone
>>   > requirements, its
>>   > >   >>> >   just never
>>   > >   >>> >   > > panned out as anything more than a novelty.
>>   > >   >>> >   > > - it would be interesting to restate this
>>   scenario in a
>>   > >   >>> >   generalized way,
>>   > >   >>> >   > > i.e., not audio to video but low bit rate, low
>>   > >   packet loss...,
>>   > >   >>> >   > to high bit
>>   > >   >>> >   > > rate, higher packet loss, etc. and see if we
>>   can create a
>>   > >   >>> >   > broader model for
>>   > >   >>> >   > > this "mobility across network and administrative
>>   > boundaries"
>>   > >   >>> >   > requirement.
>>   > >   >>> >   > > has it already been defined?  seems to me MWIF has
>>   > >   an implicit
>>   > >   >>> >   > assumption
>>   > >   >>> >   > > for such a scenario but i'm not sure we've ever
>>   > >   gone throught
>>   > >   >>> >   > the mechanics
>>   > >   >>> >   > >
>>   > >   >>> >   > > tim
>>   > >   >>> >   > >
>>   > >   >>> >   > >
>>   > >   >>> >
>>   > >   >>> >
>>   > >   >>>
>>   > >   >>
>>   > >   >>
>>   > >
>>   > >
>>   >
>>
>>
>