Thursday, September 16, 2010

Aastra 53i showing SIP 400 "Bad Request"

An amusing gotcha that's taken me two days to pin down...

If you have an Aastra 53i up to firmware version connecting to an Asterisk 1.4.x box do _not_ use the "fromuser=" directive in the SIP.CONF file associated with that phone.  The phone won't accept calls from Asterisk any more, returning error code 400 "Bad Request".

<--- SIP read from --->

SIP/2.0 400 Bad Request

Via: SIP/2.0/UDP;branch=z9hG4bK7b016ea7;rport

From: "Unknown Name" <sip:John Smith@>;tag=as269699d7
To: <;transport=udp>;tag=3007026722

Call-ID: 7e1d46dc3d5168b03720874818000e88@

CSeq: 102 INVITE
Server: Aastra 53i/
Content-Type: text/plain
Content-Length: 23

Invalid header(s): From

... as soon as I removed the "fromuser=" lines from the SIP.CONF file and did a SIP RELOAD at the console, the phones were immediately able to take calls as expected.

Shout-out to Aastra T2 tech support who put me on the right path.

Thursday, August 12, 2010

If the vendor doesn't fit, change him + CentOS 5.5 changing from DHCP to static IP addressing

If the vendor doesn't fit, change him
Fodder for another post is a nightmare I've had with Xorcom USA with one of my newest accounts.  If there was a way to blow the order and the VAR-partner relationship, including lying to the VAR, they've done it with this account.  No fun.

Eventually, we were obligated to drop Xorcom for this project and get a different appliance vendor.  We opted for Rhino Equipment Corp .  24h after the order was in, I got this from the customer:

I got the device. It's great. Perfectly packaged. It's plugged in and I've SSH'd in
First try, and the customer is happy.  Which makes me happy.  After six weeks of delays because of hardware, hopefully we'll get back on track.

CentOS 5.5 changing from DHCP to static IP addressing
The other thing the customer needed was a "how-to" on setting the Ceros-1U to a static IP address.  Here is the info I sent him.  Hopefully it will help some other Netizen out there:

Here is what you do:
cd /etc/sysconfig/network-scripts/
nano (or vi) ifcfg-eth0
... you'll see something like:
... change it to look like this:
...  ensure the IPADDR, GATEWAY and NETMASK lines make sense for your network.  Now, save and exit.  Then type:
service network restart
... your SSH session will likely die at this point, since the machine is now on a different IP address.  PING or/ and SSH to your new IP and you're in business.

Thursday, August 5, 2010

How Much Hardware Do You Need

A question I've seen a lot lately is about how much hardware at the network and in the rack you need to run an Asterisk VOIP environment for the enterprise.  The real answer is "not much".

The Network Matters
On the network side, you're looking for Quality of Service (QOS) and the ability to put your voice traffic on a Virtual LAN (VLAN) so that it is "invisible" to data traffic.  Of the two, QOS is more important and generally easier to manage.  However, if you VLAN your VOIP infrastructure, and you are wired with CAT 5e end to end as a minimum, then the lack of QOS shouldn't bite you.

If you aren't going to VLAN your VOIP infrastructure, depending on what other type of network traffic you have and how many stations, then you might have issues.  Particularly if you are not wired gigabit to the desktop and then plan on using "HD" voice codecs.  "Bursty" traffic like streaming media and large file transfers can saturate a wire resulting in lost voice traffic, which is sent over "expendable" UDP.

Big Irons
As far as which server you get, the biggest issue is not horsepower.  Generally speaking, 1CPU @ 1Ghz + 1Gb RAM + 10Gb disk = 100 conncurent calls (PRI <=> SIP) based on my experience.  Transcoding, recording, conferencing, virtual FAX, or database services running on the PBX will each cost you about 10% of that number.

The real issue in buying a server is reliability.  This is your enterprise telephony stack...  your communications line to your customers.  Hardware RAID-1, dual power supplies, hot-swappable drives and PCI cards, and "can't kill it" features are excellent.  Either that, or buy two cheaper servers and rig them in a Highly Available fail-over cluster.

My House
My home Asterisk server is an IBM PC-700 with 512Mb of RAM and 40Gb of disk.  It supports a Fax server, a Ruby/Rails configuration GUI, and three phones.  There is a VOIP trunk to an ITSP, and a DAHDI PCI card for hook to the PSTN.  The machine is asleep, even while receiving a fax and with my wife on the phone on an LD call.  I'm wired internally with 100M/bit every where in the house, except my wireless at 54M/bit.

JKL5's Cloudphones PBX, which is our growing multi-tenant service, currently has 15 end points for 5 tenants, plus six VoIP trunks split between IAX2 and SIP.  There are calls in progress at any point of of the day.  It is an Intel(R) Xeon(R) CPU @ 3.00GHz, 500Mb of RAM (100mb free) and 28Gb of disk.  Its in a data center with dual OC3 connectivity, so the only bandwidth issues are with the end-stations, usually home XDSL or Cable Modem.  Current load average at 9am on a business day is 0.09.

Wrapping Up
I personally suggest to clients be more worried about having a good network environment and a good T1 card than worry about needing a massive server.  The only thing you need to worry about with the hardware is what your plan is if there is a component failure.

Wednesday, July 28, 2010

Getting started

The whole "hedgehog concept" has pushed me to setting up the JKL5 Blog over here at We'd had one that we were running on our own servers using Typo but maintaining and managing it is just another "thing to do" that isn't our core business.

Thanks for following us and we're looking forward to sharing some of ideas and thoughts about the Free Open Source Software (FOSS) Voice Over IP (VoIP) industry. Like Social Media, FOSS VoIP is still in its infancy, and its having a profound impact on the business world, whether the Big Boys want to talk about us or not. Where it is going, I do not think anyone knows for sure. However, I think everyone who takes it seriously does agree that Small and Medium sized Businesses (SMBs) need to be on the bus to give them the ability to service their customers in a way that would have been out of their weight class a decade ago.

Thanks for reading along.