On 25 Dec 2012, at 15:40, sampsa at mac.com wrote:
On 25 Dec 2012, at 22:38, Cory Smelosky <b4 at gewt.net> wrote:
http://www2.openvms.org/kparris/Bootcamp_2010_Using_IP_OpenVMS_Cluster_Inte…
It even mentioned hobbyist clusters over the internet as a practical implementation. ;)
I'd be up for trying a cluster over the internet
Do you have the bandwidth for a high-performance one, or would we end up with a high-latency constantly-exploding one? ;)
sampsa
On 12/25/2012 03:38 PM, Cory Smelosky wrote:
http://www2.openvms.org/kparris/Bootcamp_2010_Using_IP_OpenVMS_Cluster_Inte…
It even mentioned hobbyist clusters over the internet as a practical implementation. ;)
That's downright frightening. :) I may have to try that!
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
On 25 Dec 2012, at 15:47, sampsa at mac.com wrote:
On 25 Dec 2012, at 22:45, Dave McGuire <mcguire at neurotica.com> wrote:
On 12/25/2012 06:22 AM, sampsa at mac.com wrote:
That is VERY cool!
-Dave
Thanks!
Basically, Brian has written some code on CHIMPY that produces output that I graph.
The grapher is more less ready - the end node scanner is still borked, waiting on Brian to fix it but it's Xmas and all :)
But it's very very close to mapping most of HECnet automagically.
One issue, can you use different colours? I have difficulty differentiating certain colours and I find it a little hard to see the differences between the lines.
On 25 Dec 2012, at 22:45, Dave McGuire <mcguire at neurotica.com> wrote:
On 12/25/2012 06:22 AM, sampsa at mac.com wrote:
That is VERY cool!
-Dave
Thanks!
Basically, Brian has written some code on CHIMPY that produces output that I graph.
The grapher is more less ready - the end node scanner is still borked, waiting on Brian to fix it but it's Xmas and all :)
But it's very very close to mapping most of HECnet automagically.
On 12/25/2012 06:22 AM, sampsa at mac.com wrote:
[Brian, please don't kill me for announcing this too early :P ]
Guys,
We're working on a network walker that will eventually produce a graph of HECnet.
We're basically doing a NCP SHOW KNOW CIRC, grabbing each area routing node and then recursively walking those.
Well it's a bit more complex than that, but that's the general idea :)
Once this is done, we do a NCP SHOW ADJ NODES to get the nodes in the area of each area router.
This is what we've come up with so far (the ADJ NODES code is under work, so if your node is missing, don't worry - mainly checking all the area routers are there):
http://sampsa.com/routers.svg
That is VERY cool!
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
On 25 Dec 2012, at 15:27, Cory Smelosky <b4 at gewt.net> wrote:
On 25 Dec 2012, at 15:21, Dave McGuire <mcguire at neurotica.com> wrote:
On 12/25/2012 03:07 PM, Cory Smelosky wrote:
Ah. Is any sort of encryption done for cluster-over-IP?
Oh heck no. That sort of overhead's resultant latency would destroy
the performance of a cluster on all but the very fastest of machines.
Ah, right, I'm forgetting that clustering sends a lot of data.
Yes, and it's all pretty latency-sensitive.
Now I want to set up a cluster and intentionally crash it just to see how exactly it all falls over.
Further, most VMS clusters are built with machines in the same
datacenter. It's not generally a wide-area thing.
Mmmm, true. Clustering is generally used for practical purposes and not my crazy experiments.
Right. And the VMS world isn't generally too worried about cracker
kiddiez and their root kits.
Nope. ;)
Now, if you manage to bring up a wide-area cluster (which I'd love to
see!), of course you could encrypt your tunnels. (I would)
I think I was going to try at some point I have a couple systems running simulated OpenVMS 8.4 and I almost set up an IP-based cluster. ;)
I don't THINK i actually did it but I almost did.
http://www2.openvms.org/kparris/Bootcamp_2010_Using_IP_OpenVMS_Cluster_Inte…
It even mentioned hobbyist clusters over the internet as a practical implementation. ;)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
Seriously thought guys, I think we should set up a get-together party somewhere cool.
I realise Beirut is a bit extreme but Cyprus is very nice and Malta is very very cheap.
Of course I'm not opposed to a state-side party - as long as it's warm :)
Sampsa
Hmmm.... space for what though? I can't see me bringing my MicroVAX II. :-)
-----Original Message-----
From: owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE]
On Behalf Of Cory Smelosky
Sent: 25 December 2012 19:07
To: hecnet at Update.UU.SE
Subject: Re: [HECnet] Happy Holidays HECnet
On 25 Dec 2012, at 14:05, sampsa at mac.com wrote:
Why? It's freezing. Let's go somewhere warm. What about Malta?
How about the sun? I hear that's pretty warm. It would offer plenty of
space for us, too.
Sampsa
On 25 Dec 2012, at 21:04, Gregg Levine <gregg.drwho8 at gmail.com>
wrote:
On Tue, Dec 25, 2012 at 1:54 PM, <sampsa at mac.com> wrote:
On 25 Dec 2012, at 20:51, Gregg Levine <gregg.drwho8 at gmail.com>
wrote:
Hello!
First off, I do hope your kidding Sampsa, you're better off landing
in Israel regardless of one or two little details.
Not kidding. And not setting foot in Israel :)
Hello!
I've got a better idea. One of the channel islands. Or perhaps one of
the islands that's somewhere around the Scottish North Sea coast,
also known as the Hebrides.
That part of the Middle East is still unstable despite the current
government's claim. (I stress the word current.)
-----
Gregg C Levine gregg.drwho8 at gmail.com "This signature fought the
Time
Wars, time and again."
On 25 Dec 2012, at 15:26, Dave McGuire <mcguire at neurotica.com> wrote:
On 12/25/2012 03:21 PM, Cory Smelosky wrote:
Brian (Hechinger) and I once discussed the possibility of
implementing a similar clustering system within NetBSD. I
think just implementing the SCS protocol itself would be a good
start. I wonder if there's enough documentation out there to
do it.
What'd be really interesting is joining that to a VMS cluster.
;)
It would be a huge amount of work, and I'm not sure it'd be
practical. There are a lot of things in VMS that just don't have a
mappable counterpart in the UNIX world. RMS, for instance.
And a versioned filesystem. ;)
Yes, but there's nothing stopping an enterprising soul from
implementing ODS under UNIX. Speaking of which, this is no relation to
ODS, but I had the beginnings of the RT-11 filesystem implemented under
NetBSD many years ago. (I really should finish that, but time )
If you end up finishing that, I have a friend that would probably really love it.
Brian H and I were discussing this (this was, what, eight years
ago Brian?) in the context of building a tightly-coupled cluster of
NetBSD machines, with the same basic capabilities (as adjusted for
the environment) as VMS' clustering...tight shared filesystem,
distributed locking, etc. We even discussed process migration.
Nice! I still don't see why it's taking UNIX/Linux so long to catch
up to features VMS has had since the '80s. ;)
Mostly because the current featureset is most of what the world needs
and expects right now. Sure, one could do a LOT more with the
underlying features of VMS, but nobody left today knows how to do those
things...everything is a flat "stream of bytes" file, etc. We've gained
a lot, but we've also lost a lot.
At least it's not an entire filesystem that's an SQL database. ;)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
