With help and a pointer from Peter, we ve set up the DECnet over Multinet link between LEGATO and STUPI to use TCP instead of UDP. It seems to work fine whether it s better than the UDP version only time will tell.
It turns out that none of this is magic it s all pretty well documented by Process Software. Just type HELP MULTINET SET /DECNET and you ll see the /TCP qualifier.
There s also a /FILTER_OUT_OF_ORDER qualifier which forces the driver to drop out of order UDP packets. Of course, that doesn t apply to TCP, but it does remove that objection to the UDP implementation.
Bob
On May 28, 2014, at 9:40 AM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-28 15:33, Paul_Koning at Dell.com wrote:
...
DDCMP is a data link layer that deals with packet loss. It also deals with reordering, within reason, by treating it as packet loss, just as NSP did until Phase V. It should be ok with limited duplication as well though duplication is not one of the supposed characteristics of UDP.
Duplication is definitely a possible event with UDP. It happens from time to time. It's essentially because duplication is possible with IP. UDP is just IP with ports and a checksum (if you are lucky). And IP makes no promises at all. Packets might not arrive, might arrive out of order, become duplicated, delayed, or even corrupted.
Note that I meant the DDCMP protocol, not the DDCMP point to point service. In other words, the UDP packets would carry DDCMP frames, with DDCMP header (including sequence number and all that). I think SIMH has that right now, in V4.0 DMC emulation.
Ok. So if I understand correctly, DDCMP is guaranteeing the delivery of data between the two points. Data will arrive in order and without any loss or other confusion, as seen by the layers above DDCMP?
Yes, just as with TCP. The usual disclaimers apply. For example, dups or reorders are detected only up to the wrap point of the sequence number space. Corruption is caught only up to the capabilities of the CRC being used.
Also, guaranteed delivery is a commonly used term. A more accurate term would be guaranteed delivery or notification of failure . Connection oriented services like DDCMP and TCP and TP4 and NSP will deliver the data stream intact to the other end, OR they will tell you that they could not do so.
Seems like UDP could work then, but maybe TCP would be better?
Not clear. If things get reordered or delayed to the point that the DDCMP sequence number space is no longer sufficient, then TCP would help. It makes things somewhat more complicated (as Rob Jarratt pointed out) because TCP introduces the asymmetry of who connects while DDCMP does not have than and neither does UDP.
paul
On 2014-05-28 15:33, Paul_Koning at Dell.com wrote:
On May 27, 2014, at 8:53 PM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-27 22:09, Paul_Koning at Dell.com wrote:
On May 27, 2014, at 4:04 PM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-27 21:48, Bob Armstrong wrote:
The way to run DECnet over a flaky long distance network is to use point
to point mode with a data link layer that deals with packet loss.
Probably a good idea, but we don't have that option on HECnet.
Well, HECnet is not a static piece of equipment. Anything is possible...
My bridge emulates a simple ethernet segment. Good enough many times, but if we have a link like yours, that sometimes seems to drop packets, then maybe some other alternative should be considered.
Now, the question then becomes, what can we do in this case.
As far as I understand, links using Multinet are more broken, and still use UDP. The same would appear to possibly be the case for Cisco as well?
Do anyone run any links using TCP?
That would work. DDCMP over UDP would work.
Really? UDP can cause packets to arrive in the wrong order, duplicated, or sometimes dropped. I was certain you wrote above "a data link layer that deals with packet loss". Or was that not meant to be read as that the underlaying transport should deal with it?
DDCMP is a data link layer that deals with packet loss. It also deals with reordering, within reason, by treating it as packet loss, just as NSP did until Phase V. It should be ok with limited duplication as well though duplication is not one of the supposed characteristics of UDP.
Duplication is definitely a possible event with UDP. It happens from time to time. It's essentially because duplication is possible with IP. UDP is just IP with ports and a checksum (if you are lucky). And IP makes no promises at all. Packets might not arrive, might arrive out of order, become duplicated, delayed, or even corrupted.
Note that I meant the DDCMP protocol, not the DDCMP point to point service. In other words, the UDP packets would carry DDCMP frames, with DDCMP header (including sequence number and all that). I think SIMH has that right now, in V4.0 DMC emulation.
Ok. So if I understand correctly, DDCMP is guaranteeing the delivery of data between the two points. Data will arrive in order and without any loss or other confusion, as seen by the layers above DDCMP?
Seems like UDP could work then, but maybe TCP would be better?
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
On May 27, 2014, at 8:53 PM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-27 22:09, Paul_Koning at Dell.com wrote:
On May 27, 2014, at 4:04 PM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-27 21:48, Bob Armstrong wrote:
The way to run DECnet over a flaky long distance network is to use point
to point mode with a data link layer that deals with packet loss.
Probably a good idea, but we don't have that option on HECnet.
Well, HECnet is not a static piece of equipment. Anything is possible...
My bridge emulates a simple ethernet segment. Good enough many times, but if we have a link like yours, that sometimes seems to drop packets, then maybe some other alternative should be considered.
Now, the question then becomes, what can we do in this case.
As far as I understand, links using Multinet are more broken, and still use UDP. The same would appear to possibly be the case for Cisco as well?
Do anyone run any links using TCP?
That would work. DDCMP over UDP would work.
Really? UDP can cause packets to arrive in the wrong order, duplicated, or sometimes dropped. I was certain you wrote above "a data link layer that deals with packet loss". Or was that not meant to be read as that the underlaying transport should deal with it?
DDCMP is a data link layer that deals with packet loss. It also deals with reordering, within reason, by treating it as packet loss, just as NSP did until Phase V. It should be ok with limited duplication as well though duplication is not one of the supposed characteristics of UDP.
Note that I meant the DDCMP protocol, not the DDCMP point to point service. In other words, the UDP packets would carry DDCMP frames, with DDCMP header (including sequence number and all that). I think SIMH has that right now, in V4.0 DMC emulation.
paul
Higher up the thread there was mention of using DDCMP, and also of the fact that missing 3 hello messages will cause the adjacency to go down.
Well, the user mode router I have written implements DDCMP now, over TCP. It still has some strange issues that I need to resolve, and it is only passive (ie it won't initiate a connection to a peer). But, it is mostly there otherwise. You can work around the passiveness by using SIMH at the other end and using the emulated DMC11.
Additionally, while I implement the 3 hello limit for Ethernet circuits (BC3TMULT in the routing spec), that could be changed quite easily to be a higher limit, but you would need the router to be configured the same way at both ends of the link for that to work.
Regards
Rob
On 28 May 2014 07:47, Pontus Pihlgren <pontus at update.uu.se> wrote:
On Tue, May 27, 2014 at 11:13:05PM +0200, Peter Lothberg wrote:
>
> (I might even find a switch that does it right and send it to Update)
If you do, we'll try not to procrastinate setting it up for a year :-)
/P
On 2014-05-28 08:52, Cory Smelosky wrote:
Hello all,
So...I have an 11/23+ here that runs RSTS/E and RSX-11M+ however...it
fails to even BOOT XXDP. Which is...interesting to say the least.
The disk image I made booted fine in simh. When I boot it it just drops
me back to ODT with the last address...which was REALLY soon after boot.
Could it be the same RSTS/E disk image size issue I had before? Any
ideas? My bus is contiguous, this part of memory is good, and RSTS/E
and RSX-11M+ work fine.
I would guess on a bad XXDP image, since XXDP requires almost nothing working. If RSRS/E and RSX works then I can't imagine XXDP not working.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
--
DECtec mailing list
http://dectec.info
To unsubscribe from this list see page at: http://dectec.info/mailman/listinfo/dectec_dectec.info
Hello all,
So...I have an 11/23+ here that runs RSTS/E and RSX-11M+ however...it fails to even BOOT XXDP. Which is...interesting to say the least.
The disk image I made booted fine in simh. When I boot it it just drops me back to ODT with the last address...which was REALLY soon after boot.
Could it be the same RSTS/E disk image size issue I had before? Any ideas? My bus is contiguous, this part of memory is good, and RSTS/E and RSX-11M+ work fine.
--
Cory Smelosky
http://gewt.net Personal stuff
http://gimme-sympathy.org Projects
On Tue, May 27, 2014 at 11:13:05PM +0200, Peter Lothberg wrote:
(I might even find a switch that does it right and send it to Update)
If you do, we'll try not to procrastinate setting it up for a year :-)
/P
On 2014-05-27 23:13, Peter Lothberg wrote:
When we where chasing another problem, it was found that packets
"disapear" in the Ethernet fabric at Update, sometimes.
I think you are referring to when we played with transfers and the
performance drops to the floor. That is lost packets because of
interface speed differences. They are not really lost, but there only so
much that can be queued up in the fabric between interfaces with
different speeds.
Maybe Johnny can make a small map of the current topology with all
involved things and the lan-speeds?
That would actually be useful either way, but I'm not sure I can do it
easily.
Johnny
If it was only ONE switch used for all the DECnet speaking things at
Update, it would be simple to make a drawing. And with only one device
per port, I think any resonable switch has enough buffers to make it
work way better.
--P
(I might even find a switch that does it right and send it to Update)
Well, if it was only Update then the obvious other solution would be to just force everything to use 10 Mb/s. I don't think that other than that, changing the switch would help much. If a maching is throwing out DECnet packets on a 100Mb/s interface, and it is received by a machine with a 10Mb/s interface, DECnet will suck because of the way DECnet handles packet loss and so on. Probably made even worse by really slow physical machines that don't even read out the packets from the ethernet interface at any good rates... (Hey, a PDP-11 isn't exactly super fast...)
However, the same problem appears for anyone anywhere when they try to use DECnet locally. If you run it over my bridge, I actually reshape the traffic to make it ok (that is what the throttling is about, Bob).
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
On 2014-05-27 22:09, Paul_Koning at Dell.com wrote:
On May 27, 2014, at 4:04 PM, Johnny Billquist <bqt at softjar.se> wrote:
On 2014-05-27 21:48, Bob Armstrong wrote:
The way to run DECnet over a flaky long distance network is to use point
to point mode with a data link layer that deals with packet loss.
Probably a good idea, but we don't have that option on HECnet.
Well, HECnet is not a static piece of equipment. Anything is possible...
My bridge emulates a simple ethernet segment. Good enough many times, but if we have a link like yours, that sometimes seems to drop packets, then maybe some other alternative should be considered.
Now, the question then becomes, what can we do in this case.
As far as I understand, links using Multinet are more broken, and still use UDP. The same would appear to possibly be the case for Cisco as well?
Do anyone run any links using TCP?
That would work. DDCMP over UDP would work.
Really? UDP can cause packets to arrive in the wrong order, duplicated, or sometimes dropped. I was certain you wrote above "a data link layer that deals with packet loss". Or was that not meant to be read as that the underlaying transport should deal with it?
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
