The Cruft Of My Brain

Could it be?  I just read this over at GigaOm and I think Allan’s missing the boat.  Now, I’m sure he’s tapped into a lot more resources than I am, but it just doesn’t seem right.  I could see Cisco wanting to see how they can help in the data center to make VMWare a better platform than it already is.  The VMotion technology already works very well, but what’s to stop Cisco from putting some performance hooks in their secret sauce to make it that much better than Juniper.  Also, considering their push into SAN technologies it’s really not a stretch.

I have a different thought.

Cisco’s been shipping what are essentially Linux appliances on NetMods for quite a while now.  The NM-CUE, the IPS module and the WAAS modules are examples, but there are others.  I believe these are x86 based appliances.  These appliances are part of the less frequently used “Branch in a Box” term that Cisco targets for their 2800 and 3800 series routers.  The idea being a single router, which has to be there no matter what, can also bundle in an Ethernet switch with POE, Voice interfaces to service local calls, SRST for redundancy on the voice, a wireless AP on an HWIC, local Unity Express perhaps, local IPS for distributed security, onboard IOS firewall, WAAS module…and I’m probably missing something.

Here’s the current problem.  You can only put a couple of these in at a time!  There’s no way to put all of this in one box (nevermind the wisdom of doing such a thing).

Regarding Allen’s assertion that developers would have to develop for something other than x86 because Cisco doesn’t use that, sorry Allen, you’d be wrong about that.  The NM-CUE and similar modules use a P3 and even the AIM-CUE which is a tiny motherboard attached daughtercard uses a Celeron!

My prediction, look for some beefier NM type appliances.  Maybe even something with Dual Cores.  This will be a “VMWare ESX” appliance in the NM form factor.  It’ll probably even be a couple of them at different sizes to fit the double wide slots.  You’ll choose based on how many VM’s you want to run at a time.  Of course, the IPS, Unity Express, WAAS and other images will all be available, for a license fee.

The big question is, will they also open it to allow you to run your own VM’s on it?  Maybe but maybe not.  Cisco has a history of not allowing that kind of thing.  But here’s an idea, how about running your small(ish) Windows 2003/8 server in VM?  Maybe even something with the WAAS secret sauce built right in.  If it’s a new enough processor you might even be looking a x64.  How about installing Exchange 2007 in that VM as well!

Now we really are talking about Branch in a Box!

I’m not sure how it happened but it would seem FreePBX didn’t install right on the RaqBuntu.  I’ve bene trying to figure out why things won’t register and calls won’t go through.  Turns out the include statements didn’t get populated into sip.conf, just as an example, so it doesn’t seem able to ever find the “additional” stuff FreePBX creates.  Now I’m wondering what else is broken with FreePBX.

So, do I try to fix it?  Maybe I can just reinstall FreePBX and see if that does it.

Or, do I start over on RaqBuntu with a clean slate?

I really don’t feel like reloading that whole thing.  I’m tired…off to bed.

We’ve been working on becoming a Voice Ready Partner with Microsoft for a couple of months now.  We’re finally fully certified (www.ftdata.com) and I’ve learned some things about how Microsoft is approaching the market and from my perspective it looks like the Enterprise UC market is going to be very interesting in a year or so.

Cisco took the approach in the beginning of aiming squarely at PBX replacement.  They developed the CallManager platform into something pretty solid and broad in scope.  In the last couple of years they’ve been working from the middle, towards the edge with app development.  What I mean by that is they’ve been growing apps like Presence and Mobility now that the PBX side of things is pretty much wrapped up.

Microsoft is taking the opposite approach.  Microsoft has been growing their UC product from the outside in.  First with LCS and now with OCS they have a very stable and broad product that does a great job of supporting IM with voice and video.  The PBX side of things needs some development though.

Microsoft is making it clear that they intend on owning that space, primarily by trying to change the way people do business with voice.  I think they have a long way to go but their strategy will get them there.

Consider this.  I was at a customer some months ago and saw them running the Office Communicator client.  I asked what motivated them to do that.  The answer was simple.  It was basically free since they already had an Enterprise Agreement.  Sure, those costs might be kind of hidden but the customer impression is that they are already paying for it so they might as well use it.  Oh, and a $50 eyeball camera and they had video with another office on the other side of the country.  Microsoft will make inroads for this reason alone.  When they finally strengthen some of the other pieces like PSTN access and call center functionality they’ll already have a huge installed base of Communicator clients waiting for those added features.  It’ll be harder and harder to pick up that Cisco phone sitting on the desk.

I’ll be at the Microsoft UC Partner Symposium in Houston on Monday.  I don’t have a lot of free time but if you read this blog and are interested in meeting up to talk about Voice, feel free to drop me a line.

I’ve been working on getting Asterisk running on my RaqAst (that’s what I’m calling it) Cobalt RAQ3.  For various reasons mostly relating to getting chan_sccp to work properly I’ve gone from 1.2 to 1.4 to 1.2 to 1.4.  Yeah, it’s been frustrating.  Anyway, I had a comment left on this post detailing a method of using Asterisk just for voicemail with CME doing the phone management duties.  You might remember I had already managed to do that, again, referenced on that same post, but had failed to get MWI working.  MWI’s important for the WAF.  So, that solution got tossed and I went back to chan_sccp on Asterisk.

Well, the comment directed me to http://uc-b.blogspot.com/ where Andre has apparently gotten MWI working.  So, I’m going to give that a shot and I’ll report back how it works.

In the meantime I need to get Asterisk 1.4 running on RaqAst so I attempted to remove all vestiges of Asterisk 1.2 and install 1.4.  Ok, that seemed to work but every time I’d try to start Asterisk it would look like it was working but it wasn’t.  I’d get no indication it wasn’t starting but a “ps -aef | grep ast” showed nothing.  Also, there was nothing I could find in any logs.

Finally, I figured it out.  I hadn’t removed the 1.2 version of /etc/init.d/asterisk and the 1.4 install wasn’t overwriting it.  Apparently the 1.2 version won’t start 1.4 code.  Go figure.  Anyway, after cp’ing /usr/src/{ast-1.4}/contrib/init.d/rc.debian.asterisk over to /etc/init.d/asterisk everything was happy, as far as I can tell.

I haven’t started the configuration yet but that’s next, now that I have a running Asterisk.

I picked up two Cobalt RAQs. One’s a 3 and the other is a 3i. My intention has been to use the 3 for Asterisk and the 3i for my main firewall. After all the round about with getting a usable image on one of them I wanted to see if I could just clone it to the other drive.

I used Ultimate Boot CD and the g4u tool.  I made a backup of the working drive and that seemed to work.  However, copying the backup back to the second drive kept failing on the download.  So, I just did a disk to disk copy and that worked fine.

One of my concerns is that I did the initial install of Ubuntu on a 20gb drive with a 14gb partition.  The second drive is 15gb.  The second drive boots although there’s a weird character that gets inserted during the kernel load.  Not sure it matters but time will tell on that one.  Hopefully I don’t need to rebuild it fresh.

Lesson learned, it’s easier to grow a drive than shrink it.  Start with the smaller drive!

After trying to be smart about it and figure out my own way to get Ubuntu 8.04 on the RAQ3 I decided to just suck it up and follow others leads.  So, I followed Jim Tuttle’s site about getting Ubuntu 6.06 on his Qube.  The Qube and Raq are very similar as far as I can tell so I thought I would have some success with this.  I combined this with Tim Wiley’s instructions for building a fresher kernel.  Success!  At this point I had a working Ubuntu 6.06 install.

So, I crossed my fingers and used upgrade-manager to bring it to 8.04.  The first time I ran it it timed out on one of the mirrors.  That was a little scary but it looked like it probably recovered ok from it so I tried the same thing again.  Success!  Almost seemed too easy.  After some time it asked if I wanted to remove obsolete apps so I did and then asked to reboot, which I did.  Came right back up and showed 8.04 and the uname -r still showed my custom kernel.

Sweet!

I haven’t really put any time on the box yet so I don’t know how it’ll perform but it looks pretty stable right now.  I’m going to try to clone the drive to the RAQ3i drive so I don’t have to go through the rather lengthy process again.  It all works but it’s hours long.  Kernel compiling on the RAQ isn’t what I’d call a snappy process.

More later once I get Asterisk up and running.

I’ve already been finding it useful.  You might too!

CiscoBlog Forum

Maybe not.  After poking around to find a way to recompile the AsteriskNow kernel I gave up.  rPath has a rather convoluted way (to the neophyte) of managing kernels.  So, I figured, let’s try Ubuntu following the instructions for getting Debian 3.1 to work.  I’m able to get it to Boot From Network and I can see it mounting the NFSRoot.  However, it always fails immediately with:

FATAL: kernel toKernel panic: Attempted to kill init!
o old

I’ve been looking around and haven’t found an answer to that yet.  Something would seem to be buggered in the server side nfsroot.  I did find a french page that had the exact same problem and the guy punted and installed Debian 3.1 instead.  I might fiddle with this some more but that’s probably what I’ll do too.

I knew this would be a bit of a challenge but I wasn’t expecting it to be this hard.

I was working on the AstRaq (yes, that’s what I’m calling it) last night and I’m struggling with my limited knowledge of the guts of Linux. For a long time I’ve used TrixBox and it’s worked fine for me. I’ve been considering switching to AsteriskNow though, partly because it’ll serve my needs just as well and partly because it’s put out by Digium.

The problem I’m running into is that I need to compile a new kernel for the RAQ support. I’ve compiled kernels before, although it’s been quite a while. It’s just that it’s way outside my comfort zone. So, I started looking into getting source and the process of compiling a kernel and what I think I’ve found is that you want to be consistent with the distro you’re using. In other words, it’s possible to just do a straight kernel upgrade but it’s better to use the tools provided with the distro. If someone can tell me that I’m crazy and overthinking this I’d appreciate it.

More digging and I discover that AsteriskNow is built off of rPath. Again, lack of familiarity. I’ve been seeing rPath in a lot of appliance style distro’s lately and I get the impression it has some really nice tools for building that kind of package. I found that the package management for rPath is Conary and that there are some ways to build a custom kernel for an appliance. I imagine this works for the appliance after it’s been installed but that’s not clear to me. Also, this tutorial I found makes reference to things I imagine rPath developers implicitly understand such as “cooking” in the changes. More research I’ll have to do.

But this has left me wondering…do I take the time to understand rPath and it’s oddities? Or do I do a basic Debian/Ubuntu install and load Asterisk from apt-get? That won’t get me the gui tools but I probably don’t need them. Might be nice for someone else though. And then, will I ever be able to run the updating tools in AsteriskNow if I’m running a custom kernel? Do I need to worry about AsteriskNow wiping out my customer kernel?

So many questions and not enough time for answers. If anyone has some insight I’d sure be happy to hear it.

I’m now the proud owner of a Cobalt RAQ 3 and a 3i. The 3 will be Asterisk and the 3i will be my firewall.

So far I’ve updated the flash in the 3 using these instructions and no problems. Now I’m working on building an AsteriskNow image in vmware that’ll play nice on the RAQ. It looks like it’s not that hard as long as you follow the necessary steps regarding partitions and some kernel drivers. I’ll report back once it’s in action.