x
<<   <   Page 2 / 4   >   >>
albertvc 12/4/2012 | 7:29:20 PM
re: MPLS Gets Lukewarm Reviews I agree with you. Connectionless is as old concept as IP hosts talking RIP over a yellow cable.

About the Standards, one should ask "what do we want from the internet?". Personnally, I want internet to carry data, voice and video. This implies predictability, quality, etc. Connectionless has long been tagged as too limited to do all this. The difficult marriage between circuits and datagrams has got to happen. Otherwise, true, we could have a connectionless internet talking RIP, IGRP or OSPF, and carrying our emails. Great?!?
skeptic 12/4/2012 | 7:29:16 PM
re: MPLS Gets Lukewarm Reviews Is not the goal to have MPLS improve the internet experience?
-------------
Not that I'm aware of. There are different
motives at work, but most of them have to do
with operational issues at carriers rather than
anyone's internet experience.

-------------
Is it wrong to say QOS in the same breath as MPLS. Is QOS only Diffserv and RSVP?
-------------

Why not just say diffserv or diffserv aware
TE. MPLS is a transport protocol. By itself,
it can't deliver QOS and it just confuses things
if people say that it can.

And I'm not sure that diffserv is QOS anymore
either. Diffserv standards process at the IETF
is a whole lot more disfunctional than MPLS.





skeptic 12/4/2012 | 7:29:15 PM
re: MPLS Gets Lukewarm Reviews Is the Internet connectionless on an intercity, long-haul basis? I understood that the backbone (Q, T, BRW, WCOM, CWP, etc.) carriers use ATM over SONET connections in between cities, having mapped the IP packets into ATM.
------------------

Few people seem to understand, but there are
two sides to the internet: Routing and transport.
The routing is mostly connectionless datagram
service. The transport side of things is
circuit-based.

The reason why IP and ATM (and now MPLS) co-exist
is that "routing" and "transport" are two
different problems with different optimal
solutions. One can never displace the other
totally because they solve different problems.

[note - routing probably isn't the right term,
but its as good as I could come up with to
describe IP-based connectionless equipment]
wdog 12/4/2012 | 7:29:12 PM
re: MPLS Gets Lukewarm Reviews skeptic is on to something here. Routing and transport are two different problems. Forget MPLS as a solution to QoS. For transport, push goodGÇÖole Sonet TDM circuits as close to the edge of the network as possible with switches and next-gen ATMs, and use IP routers and routing to aggregate traffic and for signaling across the circuited switched network.
albertvc 12/4/2012 | 7:29:11 PM
re: MPLS Gets Lukewarm Reviews Hi

"skeptic is on to something here. Routing and transport are two different problems."

-> Agreed. Nobody got confused I think.

"Forget MPLS as a solution to QoS."

-> So what solution do you suggest for QoS over internet? Those who argue we don't need QoS, would want to limit the internet to best-effort datagram trafic. Well, I hope the internet of the next decade will be able of more.

"For transport, push goodGÇÖole Sonet TDM circuits as close to the edge of the network as possible with switches and next-gen ATMs, and use IP routers and routing to aggregate traffic and for signaling across the circuited switched network."

-> This is +/- what carriers already do or would do. Yet the questions from your say still remain:
- what IP routing? RIP? OSPF? IGRP?... MPLS?
- what traffic do you want to see over internet? data, voice-over-ip, video streams, real-time trafic, etc?

Now the question is: what do we want from internet in 5 years? Same stuff than today? I hope something much much better. Then, coming back to the original article: if the IETF has become like the ITU of 20 years ago, with big networking vendors imposing what is good for them rather than what is good for the internet mass, then maybe we've got a problem there.


nonobvious 12/4/2012 | 7:29:08 PM
re: MPLS Gets Lukewarm Reviews >>There's a reason why the Internet has scaled to what it has, it's connectionless.<<

The Internet is bursting at the seams with only ~110k backbone routes. Its roughly equivalent to the postal service in terms of reliability and profitability. A resounding success for connectionless? Maybe for mass market "3rd class mail" applications. For the important stuff, there's PSTN, Frame Relay, and FedEx.
skeptic 12/4/2012 | 7:29:05 PM
re: MPLS Gets Lukewarm Reviews The Internet is bursting at the seams with only ~110k backbone routes. Its roughly equivalent to the postal service in terms of reliability and profitability. A resounding success for connectionless? Maybe for mass market "3rd class mail" applications. For the important stuff, there's PSTN, Frame Relay, and FedEx.
---------------------

What you miss is that internet is actually
moving the traffic between multiple points
for the end-user. The most reliable FR or
PSTN service in the word still ends up
offering an inflexable services that moves
traffic between two fixed points. And thats
great for some applications, but not others.

And as far as route capacity goes, there are
exactly the same problems in connection-oriented
networks with signaling state.

fgoldstein 12/4/2012 | 7:29:02 PM
re: MPLS Gets Lukewarm Reviews Nonobvious, you're too unkind to the post office!

The Internet runs on brute force. It's not profitable -- look at ISP financials! We all love the service it provides, but inside it's worse than a sausage factory.

Connectionless was a great idea, in 1971. It was necessary to counter the PTT's X.25, which made all connections heavy and slow. But connections don't have to be heavy and slow. Frame Relay and ATM are more efficient, for instance, while MPLS is really a polite euphemism for Frame Relay, minus the ITU standards baggage.

Connectionless works well enough if the application is something for which the ARPAnet was designed, like say email, FTP, or even web browsing. But it's not sensible for streaming. The "IP under alles" crowd seems to think that if the packet loss causes, say, a dropout in streamed music, well then that's Good Enough! I don't think paying customers will all go along.

Any of these "standards" will only be adequate if the committees don't weight them down with too many agendas. It's easy to show failure of QoS projects, but that doesn't change the need for QoS, whether on the Internet proper or parallel to it.
WeNoNeedNoStinkinMPLS 12/4/2012 | 7:28:58 PM
re: MPLS Gets Lukewarm Reviews -> So what solution do you suggest for QoS over internet? Those who argue we don't need QoS, would want to limit the internet to best-effort datagram trafic. Well, I hope the internet of the next decade will be able of more.

No matter how much we want to make the Internet something it's not, there will NEVER be a solution for QOS over the Internet. The political boundaries between autonomous systems and the pure nature of the peer to peer network from its inception will NOT allow any provider to send anything but best effort IP (aka 3rd class mail) over domain boundaries.

We've gone through iterations of bandwidth brokers trying to create an inband SS7 network to doing DiffServ codepoints that do not require ay type of connection oriented circuits and still you see none of it deployed.

Carrier's can only control what stays under their domain and because of the nature of the Internet where control points shift as we cross domain boundaries - you have domains that traffic transits that have no financial responsiblity. Get off the drugs...it's the reality of how the network was built.

No carrier will ever allow another carrier to setup an LSP across domain boundaries. The entire notion of "interdomain" kills just about every technology that gets written about. This is not a technical problem, it's a political problem.


-->Connectionless works well enough if the application is something for which the ARPAnet was designed, like say email, FTP, or even web browsing. But it's not sensible for streaming. The "IP under alles" crowd seems to think that if the packet loss causes, say, a dropout in streamed music, well then that's Good Enough! I don't think paying customers will all go along.

Agreed completely. Some applications work better than others on a connectionless network - your example is clearly one that the Internet is not equipped for. The Internet IS NOT a one size fits all.

Fact remains, the Internet while it may be great for guys like me who want to cheat the phone company doing toll bypass. I will NEVER pay more to a carrier for carrying my traffic at higher priority. Guess what, my VOIP line works great.

Interdomain QOS is an absolute blackhole and is going nowhwere anytime soon.

Interdomain MPLS is an absolute blackhole and is going nowhere anytime soon.

Why haven't carriers deployed QOS capabilities even within their own domains? It's NOT a box level problem, it's a domain level problem and no matter if a single vendor claims that they are the best QOS box in the planet, it's only as good as the weakest link within a single domain.

MPLS is a complete blackhole. From the original inception of doing traffic engineering to what it's become it's simply killing itself.

Do you remember the days of full meshed 1483? Do you remember how difficult it was to manage?

The Internet will never be suited to do all the applications that people want it to be. It was designed for applications such as email, ftp, telnet, not to be a replacement for the PSTN.
rjmcmahon 12/4/2012 | 7:28:57 PM
re: MPLS Gets Lukewarm Reviews The political boundaries between autonomous systems and the pure nature of the peer to peer network from its inception will NOT allow any provider to send anything but best effort IP (aka 3rd class mail) over domain boundaries.
________________________

Commercial boundries which provide profits to the network owners and to the content owners will drive the future of packet networks. The needs of paying customers will drive QoS technologies into these networks.

<<   <   Page 2 / 4   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE