[Dxspider-support] multiple logins + real time user info
Martin Davies G0HDB
g0hdb at amdavies.demon.co.uk
Sun Jan 3 14:28:03 GMT 2016
On 2 Jan 2016 at 20:13, djk wrote:
> [Snipped]
> I am working (very slowly) on a newer version of DXSpider and I would
> like to discuss the whole area of user "management". I really do wonder
> whether - in the modern cluster - with 6000+ concurrent users we
> continue to require "real time" user connectivity information in its
> present form.
>
> Please discuss.
Hello Dirk (and all,) thanks for an interesting and thought-provoking message. Here's my
current understanding of and thinking on the topic:
1. Currently, when a user connects to or disconnects from a node, the details of the
connection or disconnection event are disseminated in 'near-real-time' to (all?) the
other nodes in the cluster so that every node has, in theory, a reasonably accurate
view of which users are connected to which nodes in the entire cluster, although as
you've pointed out the view might not always be 100% accurate. As you've also
explained, this automatic forwarding of user connection/disconnection event details
results in lots of inter-node PC16/17/92 traffic that can consume a fair amount of
bandwidth on the inter-node links.
2. It seems to me that this 'all-informed' view of the cluster only becomes useful when a
user on a node does a 'sh/co' command to enable them to see which users are
connected to the other nodes in the cluster - for much, or perhaps even most, of the
time the status of users' connections to nodes isn't serving any useful purpose. I'm
not aware of the user status information being used for anything other than 'sh/co'.
This leads me on to thinking...
3. The system currently operates on a 'push' basis so that a user connection or
disconnection event on a node is automatically passed using PC16/17/92 messages
from the node to its adjacent nodes and thence into the wider cluster. Why not turn
things round so that the status of users' connections to a node is done on a 'pull'
basis by a node on which a user has done a sh/co' command?
4. In this 'on-request' model each node would maintain its own list of user connections
and disconnections but wouldn't automatically forward the user state information into
the rest of the cluster until requested by an adjacent node, ie. when a user on a
distant node has done a 'sh/co'. This would, I believe, get rid of a lot of the
'near-real-time' user-related inter-node traffic but would still enable a user anywhere
to get an up-to-date picture of which users are connected to which nodes although
there might be a slight delay in retrieving the user information from all the nodes in
the cluster.
5. The suggested 'on-request' model might also result in a more accurate picture of the
configuration of the cluster being available to users via the 'sh/co' command because
each node would only return its current user status when requested by another node -
there shouldn't be any out-of-date informatiion floating around the cluster.
As for compatibility with AK1A, as far as I'm aware my GB7DXC node is the last remaining
AK1A-based node anywhere and we're probably going to switch it off in the not-too-distant
future - I've persuaded our fairly small user community to use the DXSpider-based
GB7DXC-5 node instead and the legacy 'DXC node hasn't had anyone connect to it for ages!
On this basis I don't think you should worry too much about trying to make any new "user
management" capability backward-compatible with AK1A.
As for users connecting to multiple nodes, I can't personally see any need for anyone to do
this but if someone wants to do it then, in the suggested 'on-request' model, I don't see it
causing a problem. The presence of a user's callsign on multiple nodes would only be
reported when someone did a 'sh/co' or a 'sh/sta' command. I've just done a 'sh/sta gw0ogi'
command on 'DXC-5 and it has reported that GW0OGI is 'at' GB7MBC, ZL2ARN-10 and
W1DX - I wonder what benefit he's getting from the connections to those distant nodes... :-)
All comments welcome... :-)
--
73, Martin G0HDB
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
More information about the Dxspider-support
mailing list