Mailing List Archive

packet flow
Hi All,

Just read the packet flow section in Sybex JNCIA book and curious about why
the Inbound DBM ASIC have to send the packet's J-cells to all FPCs in the
router on a round-robin basis? After done that, are the J-cells in the
shared memory pool deleted?
Thanks.
-Janto
packet flow [ In reply to ]
Hi Janto,

The J-cells are sent to the FPCs for packet buffering prior to
retransmission out the egress port. Once the packets are ready for
retransmission they are reassembled into their original form and content
with the exception of the link layer header, which is added after the
packet is reassembled. The packet is then sent out the egress port.

As you note, packets are sent or "sprayed" across all FPCs in a
round-robin fashion. Once the ASIC has finished one round it begins all
over again, overwriting any previous packets stored in memory. However,
these previous stored packets have already "left the building" so it's
not a problem overwriting them.

HTH,
Steve

> -----Original Message-----
> From: juniper-nsp-bounces@puck.nether.net
> [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
> Sent: Wednesday, March 03, 2004 8:08 PM
> To: juniper-nsp@puck.nether.net
> Subject: [j-nsp] packet flow
>
>
> Hi All,
>
> Just read the packet flow section in Sybex JNCIA book and
> curious about why the Inbound DBM ASIC have to send the
> packet's J-cells to all FPCs in the router on a round-robin
> basis? After done that, are the J-cells in the shared memory
> pool deleted? Thanks. -Janto
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/junipe> r-nsp
>
packet flow [ In reply to ]
Hi Steve,
Is that mean after packets are sent across all FPCs then
the router gather all "sprayed" J-cells from all FPCs to reassembly the
packet again?
Thanks,
Janto
----- Original Message -----
From: "Steve Holman" <sholman@juniper.net>
To: "Janto Cin" <jantocin@datacomm.co.id>; <juniper-nsp@puck.nether.net>
Sent: Thursday, March 04, 2004 12:02 PM
Subject: RE: [j-nsp] packet flow


Hi Janto,

The J-cells are sent to the FPCs for packet buffering prior to
retransmission out the egress port. Once the packets are ready for
retransmission they are reassembled into their original form and content
with the exception of the link layer header, which is added after the
packet is reassembled. The packet is then sent out the egress port.

As you note, packets are sent or "sprayed" across all FPCs in a
round-robin fashion. Once the ASIC has finished one round it begins all
over again, overwriting any previous packets stored in memory. However,
these previous stored packets have already "left the building" so it's
not a problem overwriting them.

HTH,
Steve

> -----Original Message-----
> From: juniper-nsp-bounces@puck.nether.net
> [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
> Sent: Wednesday, March 03, 2004 8:08 PM
> To: juniper-nsp@puck.nether.net
> Subject: [j-nsp] packet flow
>
>
> Hi All,
>
> Just read the packet flow section in Sybex JNCIA book and
> curious about why the Inbound DBM ASIC have to send the
> packet's J-cells to all FPCs in the router on a round-robin
> basis? After done that, are the J-cells in the shared memory
> pool deleted? Thanks. -Janto
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/junipe> r-nsp
>
packet flow [ In reply to ]
Yes. The packets are buffered in the memory on the FPCs while route
lookup determines the nexthop. Once the nexthop for a given packet is
determined, the ASIC collected the J-cells from memory, reassembles the
packet and forwards it to the egress port.

Make sense?

Cheers,
Steve

> -----Original Message-----
> From: Janto Cin [mailto:jantocin@datacomm.co.id]
> Sent: Wednesday, March 03, 2004 9:12 PM
> To: Steve Holman; juniper-nsp@puck.nether.net
> Subject: Re: [j-nsp] packet flow
> Importance: High
>
>
> Hi Steve,
> Is that mean after packets are sent across all FPCs then
> the router gather all "sprayed" J-cells from all FPCs to
> reassembly the packet again? Thanks, Janto
> ----- Original Message -----
> From: "Steve Holman" <sholman@juniper.net>
> To: "Janto Cin" <jantocin@datacomm.co.id>;
> <juniper-nsp@puck.nether.net>
> Sent: Thursday, March 04, 2004 12:02 PM
> Subject: RE: [j-nsp] packet flow
>
>
> Hi Janto,
>
> The J-cells are sent to the FPCs for packet buffering prior
> to retransmission out the egress port. Once the packets are
> ready for retransmission they are reassembled into their
> original form and content with the exception of the link
> layer header, which is added after the packet is reassembled.
> The packet is then sent out the egress port.
>
> As you note, packets are sent or "sprayed" across all FPCs in
> a round-robin fashion. Once the ASIC has finished one round
> it begins all over again, overwriting any previous packets
> stored in memory. However, these previous stored packets
> have already "left the building" so it's not a problem
> overwriting them.
>
> HTH,
> Steve
>
> > -----Original Message-----
> > From: juniper-nsp-bounces@puck.nether.net
> > [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
> > Sent: Wednesday, March 03, 2004 8:08 PM
> > To: juniper-nsp@puck.nether.net
> > Subject: [j-nsp] packet flow
> >
> >
> > Hi All,
> >
> > Just read the packet flow section in Sybex JNCIA book and curious
> > about why the Inbound DBM ASIC have to send the packet's J-cells to
> > all FPCs in the router on a round-robin basis? After done that, are
> > the J-cells in the shared memory pool deleted? Thanks. -Janto
> >
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp@puck.nether.net
> > http://puck.nether.net/mailman/listinfo/junipe> r-nsp
> >
>
>
packet flow [ In reply to ]
Is the notification and result cell only consist of header?
And the J-cell packet consist of payload?
TIA,
Janto
----- Original Message -----
From: "Steve Holman" <sholman@juniper.net>
To: "Janto Cin" <jantocin@datacomm.co.id>; <juniper-nsp@puck.nether.net>
Sent: Thursday, March 04, 2004 12:33 PM
Subject: RE: [j-nsp] packet flow


Yes. The packets are buffered in the memory on the FPCs while route
lookup determines the nexthop. Once the nexthop for a given packet is
determined, the ASIC collected the J-cells from memory, reassembles the
packet and forwards it to the egress port.

Make sense?

Cheers,
Steve

> -----Original Message-----
> From: Janto Cin [mailto:jantocin@datacomm.co.id]
> Sent: Wednesday, March 03, 2004 9:12 PM
> To: Steve Holman; juniper-nsp@puck.nether.net
> Subject: Re: [j-nsp] packet flow
> Importance: High
>
>
> Hi Steve,
> Is that mean after packets are sent across all FPCs then
> the router gather all "sprayed" J-cells from all FPCs to
> reassembly the packet again? Thanks, Janto
> ----- Original Message -----
> From: "Steve Holman" <sholman@juniper.net>
> To: "Janto Cin" <jantocin@datacomm.co.id>;
> <juniper-nsp@puck.nether.net>
> Sent: Thursday, March 04, 2004 12:02 PM
> Subject: RE: [j-nsp] packet flow
>
>
> Hi Janto,
>
> The J-cells are sent to the FPCs for packet buffering prior
> to retransmission out the egress port. Once the packets are
> ready for retransmission they are reassembled into their
> original form and content with the exception of the link
> layer header, which is added after the packet is reassembled.
> The packet is then sent out the egress port.
>
> As you note, packets are sent or "sprayed" across all FPCs in
> a round-robin fashion. Once the ASIC has finished one round
> it begins all over again, overwriting any previous packets
> stored in memory. However, these previous stored packets
> have already "left the building" so it's not a problem
> overwriting them.
>
> HTH,
> Steve
>
> > -----Original Message-----
> > From: juniper-nsp-bounces@puck.nether.net
> > [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
> > Sent: Wednesday, March 03, 2004 8:08 PM
> > To: juniper-nsp@puck.nether.net
> > Subject: [j-nsp] packet flow
> >
> >
> > Hi All,
> >
> > Just read the packet flow section in Sybex JNCIA book and curious
> > about why the Inbound DBM ASIC have to send the packet's J-cells to
> > all FPCs in the router on a round-robin basis? After done that, are
> > the J-cells in the shared memory pool deleted? Thanks. -Janto
> >
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp@puck.nether.net
> > http://puck.nether.net/mailman/listinfo/junipe> r-nsp
> >
>
>
packet flow [ In reply to ]
This is not entirely easy to say. But in general yes. J-Cells in fpc
memory are data cells. Encaps and headers are built by the system
depending on user configuration etc. For example TOX bits in the
header will be rewritten depending on COS config. Is this what you are
asking?
Gary

On Mar 5, 2004, at 1:39 AM, Janto Cin wrote:

> Is the notification and result cell only consist of header?
> And the J-cell packet consist of payload?
> TIA,
> Janto
> ----- Original Message -----
> From: "Steve Holman" <sholman@juniper.net>
> To: "Janto Cin" <jantocin@datacomm.co.id>;
> <juniper-nsp@puck.nether.net>
> Sent: Thursday, March 04, 2004 12:33 PM
> Subject: RE: [j-nsp] packet flow
>
>
> Yes. The packets are buffered in the memory on the FPCs while route
> lookup determines the nexthop. Once the nexthop for a given packet is
> determined, the ASIC collected the J-cells from memory, reassembles the
> packet and forwards it to the egress port.
>
> Make sense?
>
> Cheers,
> Steve
>
>> -----Original Message-----
>> From: Janto Cin [mailto:jantocin@datacomm.co.id]
>> Sent: Wednesday, March 03, 2004 9:12 PM
>> To: Steve Holman; juniper-nsp@puck.nether.net
>> Subject: Re: [j-nsp] packet flow
>> Importance: High
>>
>>
>> Hi Steve,
>> Is that mean after packets are sent across all FPCs then
>> the router gather all "sprayed" J-cells from all FPCs to
>> reassembly the packet again? Thanks, Janto
>> ----- Original Message -----
>> From: "Steve Holman" <sholman@juniper.net>
>> To: "Janto Cin" <jantocin@datacomm.co.id>;
>> <juniper-nsp@puck.nether.net>
>> Sent: Thursday, March 04, 2004 12:02 PM
>> Subject: RE: [j-nsp] packet flow
>>
>>
>> Hi Janto,
>>
>> The J-cells are sent to the FPCs for packet buffering prior
>> to retransmission out the egress port. Once the packets are
>> ready for retransmission they are reassembled into their
>> original form and content with the exception of the link
>> layer header, which is added after the packet is reassembled.
>> The packet is then sent out the egress port.
>>
>> As you note, packets are sent or "sprayed" across all FPCs in
>> a round-robin fashion. Once the ASIC has finished one round
>> it begins all over again, overwriting any previous packets
>> stored in memory. However, these previous stored packets
>> have already "left the building" so it's not a problem
>> overwriting them.
>>
>> HTH,
>> Steve
>>
>>> -----Original Message-----
>>> From: juniper-nsp-bounces@puck.nether.net
>>> [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
>>> Sent: Wednesday, March 03, 2004 8:08 PM
>>> To: juniper-nsp@puck.nether.net
>>> Subject: [j-nsp] packet flow
>>>
>>>
>>> Hi All,
>>>
>>> Just read the packet flow section in Sybex JNCIA book and curious
>>> about why the Inbound DBM ASIC have to send the packet's J-cells to
>>> all FPCs in the router on a round-robin basis? After done that, are
>>> the J-cells in the shared memory pool deleted? Thanks. -Janto
>>>
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp@puck.nether.net
>>> http://puck.nether.net/mailman/listinfo/junipe> r-nsp
>>>
>>
>>
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/juniper-nsp
>
packet flow [ In reply to ]
Yes, thank you.

----- Original Message -----
From: "Gary Tate" <gtate@juniper.net>
To: "Janto Cin" <jantocin@datacomm.co.id>
Cc: <juniper-nsp@puck.nether.net>; "Gary Tate" <gtate@juniper.net>; "Steve
Holman" <sholman@juniper.net>
Sent: Friday, March 05, 2004 5:43 PM
Subject: Re: [j-nsp] packet flow


> This is not entirely easy to say. But in general yes. J-Cells in fpc
> memory are data cells. Encaps and headers are built by the system
> depending on user configuration etc. For example TOX bits in the
> header will be rewritten depending on COS config. Is this what you are
> asking?
> Gary
>
> On Mar 5, 2004, at 1:39 AM, Janto Cin wrote:
>
> > Is the notification and result cell only consist of header?
> > And the J-cell packet consist of payload?
> > TIA,
> > Janto
> > ----- Original Message -----
> > From: "Steve Holman" <sholman@juniper.net>
> > To: "Janto Cin" <jantocin@datacomm.co.id>;
> > <juniper-nsp@puck.nether.net>
> > Sent: Thursday, March 04, 2004 12:33 PM
> > Subject: RE: [j-nsp] packet flow
> >
> >
> > Yes. The packets are buffered in the memory on the FPCs while route
> > lookup determines the nexthop. Once the nexthop for a given packet is
> > determined, the ASIC collected the J-cells from memory, reassembles the
> > packet and forwards it to the egress port.
> >
> > Make sense?
> >
> > Cheers,
> > Steve
> >
> >> -----Original Message-----
> >> From: Janto Cin [mailto:jantocin@datacomm.co.id]
> >> Sent: Wednesday, March 03, 2004 9:12 PM
> >> To: Steve Holman; juniper-nsp@puck.nether.net
> >> Subject: Re: [j-nsp] packet flow
> >> Importance: High
> >>
> >>
> >> Hi Steve,
> >> Is that mean after packets are sent across all FPCs then
> >> the router gather all "sprayed" J-cells from all FPCs to
> >> reassembly the packet again? Thanks, Janto
> >> ----- Original Message -----
> >> From: "Steve Holman" <sholman@juniper.net>
> >> To: "Janto Cin" <jantocin@datacomm.co.id>;
> >> <juniper-nsp@puck.nether.net>
> >> Sent: Thursday, March 04, 2004 12:02 PM
> >> Subject: RE: [j-nsp] packet flow
> >>
> >>
> >> Hi Janto,
> >>
> >> The J-cells are sent to the FPCs for packet buffering prior
> >> to retransmission out the egress port. Once the packets are
> >> ready for retransmission they are reassembled into their
> >> original form and content with the exception of the link
> >> layer header, which is added after the packet is reassembled.
> >> The packet is then sent out the egress port.
> >>
> >> As you note, packets are sent or "sprayed" across all FPCs in
> >> a round-robin fashion. Once the ASIC has finished one round
> >> it begins all over again, overwriting any previous packets
> >> stored in memory. However, these previous stored packets
> >> have already "left the building" so it's not a problem
> >> overwriting them.
> >>
> >> HTH,
> >> Steve
> >>
> >>> -----Original Message-----
> >>> From: juniper-nsp-bounces@puck.nether.net
> >>> [mailto:juniper-nsp-bounces@puck.nether.net] On Behalf Of Janto Cin
> >>> Sent: Wednesday, March 03, 2004 8:08 PM
> >>> To: juniper-nsp@puck.nether.net
> >>> Subject: [j-nsp] packet flow
> >>>
> >>>
> >>> Hi All,
> >>>
> >>> Just read the packet flow section in Sybex JNCIA book and curious
> >>> about why the Inbound DBM ASIC have to send the packet's J-cells to
> >>> all FPCs in the router on a round-robin basis? After done that, are
> >>> the J-cells in the shared memory pool deleted? Thanks. -Janto
> >>>
> >>> _______________________________________________
> >>> juniper-nsp mailing list juniper-nsp@puck.nether.net
> >>> http://puck.nether.net/mailman/listinfo/junipe> r-nsp
> >>>
> >>
> >>
> >
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp@puck.nether.net
> > http://puck.nether.net/mailman/listinfo/juniper-nsp
> >
>