Since the topic exploded, what are your opinions on encapsulation of leased line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated. Thanks shawn -----Original Message----- From: Iljitsch van Beijnum [mailto:iljitsch@muada.com] Sent: Wednesday, February 20, 2002 4:28 PM To: Jon Mansey Cc: nanog@merit.edu Subject: Re: Cisco PPP DS-3 limitations - 42.9Mbpbs? On Wed, 20 Feb 2002, Jon Mansey wrote:
OMG! Arent we missing the point here? What about never running links above 60% or so to allow for bursts against the 5 min average, and <shudder> upgrading or adding capacity when we get too little headroom.
And here we are, nickel and diming over a few MBps near to 45M on a DS3...
And why not? Obviously there is a reason why they're not upgrading, because there is plenty of traffic to fill up a second or faster circuit if packets are being dropped because of congestion. (Which has not been confirmed so far.) There shouldn't be any problems pushing a DS3 well beyond 99% utilization, by the way. With an average packet size of 500 bytes and 98 packets in the output queue on average, 99% only introduces a 9 ms delay. The extra RTT will also slow TCP down, but not in such a brutal way as significant numbers of lost packets will. Just use a queue size of 500 or so, and enable (W)RED to throttle back TCP when there are large bursts.
There's no reason to use frame-relay encapsulation unless you're actually going through a frame network. For point-to-point circuits, from Cisco to Cisco, HDLC is the best choice, but it's proprietary (although Juniper has a Cisco HDLC mode). For anything else, I'd recommend PPP. -C On Thu, Feb 21, 2002 at 01:46:44PM -0500, Gyorfy, Shawn wrote:
Since the topic exploded, what are your opinions on encapsulation of leased line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated.
Thanks
shawn
-----Original Message----- From: Iljitsch van Beijnum [mailto:iljitsch@muada.com] Sent: Wednesday, February 20, 2002 4:28 PM To: Jon Mansey Cc: nanog@merit.edu Subject: Re: Cisco PPP DS-3 limitations - 42.9Mbpbs?
On Wed, 20 Feb 2002, Jon Mansey wrote:
OMG! Arent we missing the point here? What about never running links above 60% or so to allow for bursts against the 5 min average, and <shudder> upgrading or adding capacity when we get too little headroom.
And here we are, nickel and diming over a few MBps near to 45M on a DS3...
And why not? Obviously there is a reason why they're not upgrading, because there is plenty of traffic to fill up a second or faster circuit if packets are being dropped because of congestion. (Which has not been confirmed so far.)
There shouldn't be any problems pushing a DS3 well beyond 99% utilization, by the way. With an average packet size of 500 bytes and 98 packets in the output queue on average, 99% only introduces a 9 ms delay. The extra RTT will also slow TCP down, but not in such a brutal way as significant numbers of lost packets will. Just use a queue size of 500 or so, and enable (W)RED to throttle back TCP when there are large bursts.
-- --------------------------- Christopher A. Woodfield rekoil@semihuman.com PGP Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xB887618B
On Thu, 21 Feb 2002, Christopher A. Woodfield wrote:
There's no reason to use frame-relay encapsulation unless you're actually going through a frame network.
For point-to-point circuits, from Cisco to Cisco, HDLC is the best choice, but it's proprietary (although Juniper has a Cisco HDLC mode). For anything else, I'd recommend PPP.
From what I recall, once LCP is established, PPP and HDLC have about the same amount of overhead. I could be wrong. PPP's finite-state machine is substantially more complex than HDLC's to account for PPP multilink, encryption, etc, but that would mostly be in the part of the machine that gets executed prior to or as a part of establishing LCP.
The PPP vs HDLC topic flares up every once in awhile on comp.dcom.sys.cisco. A Google search should turn up lots of reading on the subject. jms
On Thu, Feb 21, 2002 at 01:46:44PM -0500, Gyorfy, Shawn wrote:
Since the topic exploded, what are your opinions on encapsulation of leased line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated.
If you are running frame-relay now, unless you have an asymmetrical configuration (which cisco CAN do) where one end is configured to look like a f/r switch, you probably have your carrier's f/r switch in the middle. Whoever configured it that way in the router did it because that is the king of circuit that was ordered. It may well be that there was a much better price especially if there was a lot of mileage and the cir was set low. f/r can look good sometimes. If it is a real point to point t3, don't use f/r ----- Original Message ----- From: "Gyorfy, Shawn" <sgyorfy@elinkny.com> To: <nanog@merit.edu>; <cisco-nsp@puck.nether.net> Sent: Thursday, February 21, 2002 1:46 PM Subject: Cisco DS3 Questions..
Since the topic exploded, what are your opinions on encapsulation of
line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated.
Thanks
shawn
-----Original Message----- From: Iljitsch van Beijnum [mailto:iljitsch@muada.com] Sent: Wednesday, February 20, 2002 4:28 PM To: Jon Mansey Cc: nanog@merit.edu Subject: Re: Cisco PPP DS-3 limitations - 42.9Mbpbs?
On Wed, 20 Feb 2002, Jon Mansey wrote:
OMG! Arent we missing the point here? What about never running links above 60% or so to allow for bursts against the 5 min average, and <shudder> upgrading or adding capacity when we get too little headroom.
And here we are, nickel and diming over a few MBps near to 45M on a DS3...
And why not? Obviously there is a reason why they're not upgrading, because there is plenty of traffic to fill up a second or faster circuit if packets are being dropped because of congestion. (Which has not been confirmed so far.)
There shouldn't be any problems pushing a DS3 well beyond 99% utilization, by the way. With an average packet size of 500 bytes and 98 packets in
leased the
output queue on average, 99% only introduces a 9 ms delay. The extra RTT will also slow TCP down, but not in such a brutal way as significant numbers of lost packets will. Just use a queue size of 500 or so, and enable (W)RED to throttle back TCP when there are large bursts.
Thus spake "Gyorfy, Shawn" <sgyorfy@elinkny.com>
Since the topic exploded, what are your opinions on encapsulation of leased line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated.
As you're finding out, this is largely a religious issue. There are no significant differences in overhead between HDLC, PPP, and FR. Any performance difference can be more easily attributed to vendor implementation than to protocol efficiency. In practice, HDLC is the dominant encapsulation, primarily since it's Cisco's default. If for no other reason, you should use HDLC because almost everyone expects you to be using it. PPP is obviously present in non-Cisco shops, and anywhere MLPPP or LFI is needed. FR is only used as a p-t-p encapsulation in certain cases that require it; almost nobody uses it without a good reason. S
> In practice, HDLC is the dominant encapsulation, primarily since it's > Cisco's default. If for no other reason, you should use HDLC because almost > everyone expects you to be using it. PPP is obviously present in non-Cisco > shops, and anywhere MLPPP or LFI is needed. FR is only used as a p-t-p > encapsulation in certain cases that require it; almost nobody uses it > without a good reason. I'd second that reasoning. FWIW, MLPPP is one of the main reasons that PPP seems to get used now, the ability to bond multiple DS1s together, or even multiple DS3s, potentially, although there's little market force to do that. The main reason for running frame relay on a single point-to-point would probably be preservation of VLANs across a bridged link in a controlled way. Not saying that's a particularly good use, just that that's the most common reason I've seen it deployed on otherwise simple point-to-points. The idea behind frame relay is that it gives you a multipoint layer-2 protocol to use on a more complicated network of circuits. -Bill
[Replies redirected to cisco-nsp, where this is slightly more on-topic] * ssprunk@cisco.com (Stephen Sprunk) [Fri 22 Feb 2002, 18:49 CET]:
In practice, HDLC is the dominant encapsulation, primarily since it's Cisco's default. If for no other reason, you should use HDLC because almost everyone expects you to be using it. PPP is obviously present in non-Cisco shops, and anywhere MLPPP or LFI is needed. FR is only used as a p-t-p encapsulation in certain cases that require it; almost nobody uses it without a good reason.
Also, PPP supports authentication and you have to be careful with specifying that it's not needed for serial links towards customers, otherwise you suddenly find yourself or your customer locked out; with HDLC the risks of this are lower. Of course, you only run into this sort of trouble if your provisioning system isn't too well-thought-out, IMNSHO, or if you decide to move from basic plain password authentication to something TACACS+-based... Regards, -- Niels.
On Fri, Feb 22, 2002 at 11:27:16AM -0600, Stephen Sprunk wrote:
Thus spake "Gyorfy, Shawn" <sgyorfy@elinkny.com>
Since the topic exploded, what are your opinions on encapsulation of leased line DS3s. We currently use Frame Relay for out Point to Point DS3 connections. Personally, I don't know why we use FR as our encapsulation, and so the question to all. If you are running Cisco to Cisco, would it be wise to run HDLC or PPP? Our DS3s' here are hardly maxed out, 15% or so, so I'm not complaining about the few extra bits I can squeeze out them but maybe that 15% can shrink to 10% with less overhead. Opinions or examples of life appreciated.
As you're finding out, this is largely a religious issue. There are no significant differences in overhead between HDLC, PPP, and FR. Any performance difference can be more easily attributed to vendor implementation than to protocol efficiency.
In practice, HDLC is the dominant encapsulation, primarily since it's Cisco's default. If for no other reason, you should use HDLC because almost everyone expects you to be using it. PPP is obviously present in non-Cisco shops, and anywhere MLPPP or LFI is needed. FR is only used as a p-t-p encapsulation in certain cases that require it; almost nobody uses it without a good reason.
We allways use PPP, the primary reasons being: - The line protocol goes down when the line is looped. - It's easier to debug than HDLC /Jesper -- Jesper Skriver, jesper(at)skriver(dot)dk - CCIE #5456 Work: Network manager @ AS3292 (Tele Danmark DataNetworks) Private: FreeBSD committer @ AS2109 (A much smaller network ;-) One Unix to rule them all, One Resolver to find them, One IP to bring them all and in the zone to bind them.
On Fri, 22 Feb 2002, Jesper Skriver wrote:
We allways use PPP, the primary reasons being:
- The line protocol goes down when the line is looped.
int Serial0 encap hdlc down-when-looped
- It's easier to debug than HDLC
/Jesper
/-----------------------+-----------------------\ | Mike Joseph | Netaxs, Inc. | | Network Engineering | ~~~~~~~~~~~~ | | ~~~~~~~~~~~~~~~~~~~ | (610)825-9800 | | mjoseph@netaxs.com | www.netaxs.com | \-----------------------+-----------------------/
On Fri, Feb 22, 2002 at 03:19:28PM -0500, Mike Joseph wrote:
On Fri, 22 Feb 2002, Jesper Skriver wrote:
We allways use PPP, the primary reasons being:
- The line protocol goes down when the line is looped.
int Serial0 encap hdlc down-when-looped
I know, but that needs to be configued in both ends, and as it's not the default, it's likely not to be configured on the CPE device ... And when using static routing, it's very nice that the line protocol goes down.
- It's easier to debug than HDLC
/Jesper
/Jesper -- Jesper Skriver, jesper(at)skriver(dot)dk - CCIE #5456 Work: Network manager @ AS3292 (Tele Danmark DataNetworks) Private: FreeBSD committer @ AS2109 (A much smaller network ;-) One Unix to rule them all, One Resolver to find them, One IP to bring them all and in the zone to bind them.
At 09:24 PM 2/22/2002 +0100, Jesper Skriver wrote:
int Serial0 encap hdlc down-when-looped
I know, but that needs to be configued in both ends, and as it's not the default, it's likely not to be configured on the CPE device ...
Neither is PPP.
On Mon, Feb 25, 2002 at 07:36:11AM -0800, Eliot Lear wrote:
At 09:24 PM 2/22/2002 +0100, Jesper Skriver wrote:
int Serial0 encap hdlc down-when-looped
I know, but that needs to be configued in both ends, and as it's not the default, it's likely not to be configured on the CPE device ...
Neither is PPP.
The customer will never get the link up if he/she doesn't configure PPP, he/she will get link up if the link is configured for HDLC but they havn't configured 'down-when-looped' /Jesper -- Jesper Skriver, jesper(at)skriver(dot)dk - CCIE #5456 Work: Network manager @ AS3292 (Tele Danmark DataNetworks) Private: FreeBSD committer @ AS2109 (A much smaller network ;-) One Unix to rule them all, One Resolver to find them, One IP to bring them all and in the zone to bind them.
participants (10)
-
Barton F Bruce
-
Bill Woodcock
-
Christopher A. Woodfield
-
Eliot Lear
-
Gyorfy, Shawn
-
Jesper Skriver
-
Justin Streiner
-
Mike Joseph
-
Niels Bakker
-
Stephen Sprunk