[RRG] Comparison table - LISP/APT/Ivip/TRRP

[Robin Whittle <rw@firstpr.com.au>]

Hi Brian,

I don't have time at present to respond to all the discussions on
this list or on the RAM list which I understand should now move here.

Below is an updated table, including what I currently understand
about TRRP - which may be wrong.

I have not added a Traffic Engineering column, because the various
systems have different approaches to this.  For instance Ivip (and
maybe TRRP) has no explicit TE functions, but could be used for some
TE purposes by splitting traffic up per IP address so one or more IP
addresses is mapped to one ETR with other IP addresses mapped to
other ETRs.

I have added a column for "No delay ITRs".  As per your message:

http://www1.ietf.org/mail-archive/web/ram/current/msg01818.html

I think that any system where packets have to wait while a query
traverses some global system will not be acceptable.  This makes
LISP-CONS global CDR network approach:

   ITR -> CAR -> CDR*n ->
                           CAR (authoritative)
   ITR <- CAR <- CDR*n <-

and TRRP's global DNS approach look really unattractive.

In LISP-NERD every ITR has a copy of the database.  In eFIT-APT, the
ITR can get a response from a local Default Mapper very quickly.
Ivip has some ITRs with a full database and others which cache - the
caching ITRs get responses very quickly from local Query Servers.


You wrote:

> IMHO address "portability" is a bogus criterion, since a
> solution that doesn't require portability shouldn't be measured
> against it.

I have retained Portability because as far as I can see, a
significant number of end-users with substantial networks want and
need to keep the address range of their networks when they change ISP.

  http://www1.ietf.org/mail-archive/web/ram/current/msg01783.html

I think it is mainly - or only - people who are trying to run the
Internet who maintain some hope that end-users will one day not care
about having their entire network renumbered.  I can't imagine why
anyone running a substantial end-user network today would consider
this a possibility.

Perhaps in the distant future when a totally new system beyond IPv4
or IPv6 arrives, maybe they won't care about their entire network
being renumbered - for multihoming service restoration and/or for
mobility and/or for choosing another ISP.

For now, for end-users with substantial IPv4 and IPv6 networks -
those who currently want to, or will want to, get PI space - need
portability.  So I am adamant this is not a bogus criteria for
selecting an architectural upgrade which existing networks are
supposed to be compatible with - in the 0 to 15 year timeframe.


I have added a column for the locus of control of multihoming
service restoration.

  B = Built into protocol = monolithic system.
  E = External MH monitoring system required - the proposed
      architectural addition is a modular component.

  H = Host-level control.
  I = ITRs work independently.
  R = Router level control (Six/One, the CE router).



             SHIM6  Six/  Mobile  LISP-   LISP-   eFIT-  Ivip   TRRP
                    One   IPv6    NERD    CONS    APT

No delay     na     na    na      Y               Y      Y
ITRs

Address
portability                       Y       Y       Y      Y      Y

Multihoming   Y     Y             Y       Y       Y      Y      Y

MH Service
Restoration   BH    BHR           BI      BI      BI     E      E

Provides (1)
Mobility                  Y                              Y*

IPv4 too                          Y       Y       Y      Y      Y

No host                           Y       Y       Y      Y*     Y
changes


(1) Actually provides mobility, rather than being in some way
    compatible with it.

 *  Mobile IPv4 or IPv6 hosts making use of Ivip will need new host
    software.

There are many other things which could be compared, but they don't
fit in a table.

My comparison:

  http://www.firstpr.com.au/ip/ivip/comp/

remains somewhat relevant, but doesn't cover TRRP or the
developments on the RAM and RRG lists since then.  For instance, my
interpretation of:

  http://psg.com/lists/rrg/2007/msg00260.html
  http://psg.com/lists/rrg/2007/msg00264.html

is that a long-term arrangement for LISP-NERD being compatible with
non-upgraded networks is similar or identical to the "anycast ITRs"
approach of Ivip.

All these proposals - LISP, eFIT-APT, Ivip and TRRP - are, in my
view, ugly kludges.  They all have similar problems with tunneling
overhead, fragmentation, breaking traceroute and Path MTU Discovery.
 All but Ivip, in my view, have a problem with the ETRs finding it
extremely onerous to ensure that local source addresses are not
spoofed in the packets they decapsulate:

http://www.firstpr.com.au/ip/ivip/draft-whittle-ivip-arch-00.html#anchor67

(I need to respond to Bill Herrin's message about this.)

Assuming we on this list are right in our general consensus that
something must be done in the next few years to stop the growth in
the number of advertised prefixes, while still allowing for general
Internet growth and in particular the growth in the number of
multihomed end-user networks, then (along with modest improvements
to BGP) maybe one of these kludges is a lesser evil than any other
alternative.  But someone would need to do a cost-benefit analysis
of simply building bigger routers compared to introducing one of
these ITR-ETR architectural schemes.

I am not convinced everyone shares this consensus.  Those who don't
are probably not active on this list.  I recently corresponded with
a knowledgeable and communicative fellow on another list who figured
that the current growth in multihomed sites is just the middle of an
S curve (it is difficult in ASCII art):

                                         .....................
                          ...............
                  ........
              ....
            ..
        ....
........

and that it will flatten out before too long, with current and
near-future router capabilities coping with the remaining rate of
growth without any dramatic architectural changes as we are
contemplating.

I don't believe we are near the end of the growth phase of an
S-curve, and I am excited about the ability of Ivip (and potentially
the others) to use IPv4 address space more efficiently.  I am also
excited about Ivip's ability (assuming the database replication can
be done fast enough) to support a dramatically improved form of
mobility for IPv4 and IPv6.

I am unsure about when the rising costs of routers, if that is the
outcome of doing nothing, will be such a problem as to convince a
sufficiently large majority of  people (ISPs mainly, I guess) that
something dramatic needs to be done.


  - Robin

--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg