Re: [RRG] Are host-stack modifications allowed or disallowed ?

[Brian E Carpenter <brian.e.carpenter@gmail.com>]

On 2008-03-05 04:24, Scott Brim wrote:
> On 2/24/08 1:24 PM, Brian E Carpenter allegedly wrote:
>> I'm thinking that there may be a class of solutions which will
>> work as far as aggregation and scaling are concerned without
>> any host modifications, but would allow host modifications to
>> have substantial impact on traffic engineering for the benefit
>> of sites hosting major server farms. This type of site cares
>> about traffic engineering to improve its own performance and
>> reliability, quite independently of ISP concerns about traffic
>> engineering. I could imagine a server farm dynamically adjusting
>> its loc-id mapping as the work load shifts around between
>> phsyical servers, or as it observes performance changes
>> from one ISP or another. I would expect this would need tight
>> interaction between host virtualisation software and the
>> routing system.
> 
> So ...
> 
>   - a host or server farm can change its loc-id mapping and control
>     which site interfaces it receives traffic on.  This can be
>     per-flow.
> 
>   - on a broader level, a site interconnect point can control how
>     traffic to an including prefix enters the site.
> 
> Since loc-id mappings for whole prefixes will be cached all over the
> Internet, the only way for the host (or server farm) to control this
> dynamically would be to use an ID from a different prefix.  

Or, use the loc-id map to punch a hole in a prefix - exactly as they
might use BGP today. They might also game whatever TTL mechanism the mapping
system supports (as people game the DNS TTL today). For rough-and-ready
load balancing, the method doesn't have to be watertight.

> Would that
> do what you want?  Or do you want to be able to say "Dear edge, I know
> your policy for traffic to prefix X is to bring it in through these two
> points, but I would like it all brought into just this one point"?  If
> so, what's the protocol?

I was imagining that the site would just spit out mapping updates;
specific topology isn't generally known to end sites.

> 
> Also, what do you do about interactions between host requests and
> possible traffic oscillations?  The edge would have to take all requests
> into account and come up with its own synthesis, filtered through
> overall network policy anyway?

Yes, I'd expect something of an arms race to develop.

> 
> It seems to me that anything the server/host could influence at the
> level of loc-id mapping would probably already be taken into account in
> the interconnect's behavior, including whether to have any dynamic
> behavior at all.  For its own purposes, the host can use different
> addresses as needed (and we've been there before).
> 
>> However, the fact that sites use DNS and BGP4 to attract
>> traffic to particular prefixes today strongly suggests that
>> they will use any deployed loc-id mapping system the same
>> way.
> 
> Sure.  That's site level policy.
> 
> I wish I could say we should talk about it in person.

I think the question is whether The Solution should
a) recognize that sites may legitimately influence TE policy and
provide hooks;
b) turn a blind eye to the issue, which will probably cause sites
to influence TE by back door methods as today;
or c) allow ISPs to defeat any attempt by sites to influence TE.

My preference is for a).

     Brian

--
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