No Jitter’s Matt Brunk: SIP Means Change
Earlier this week Matt Brunk penned a post over at CMP’s No Jitter blog entitled “SIP Means Change.” It’s a short piece detailing the contrast between SIP phones, Asterisk servers and legacy proprietary equipment. In particular it dwells on the boot times of the various items. He highlights how anything that takes longer, even just a little longer, ultimately has a higher cost.
Matt points out that the older, proprietary digital phones were effectively instant-on devices compared to SIP phones. This is a little obtuse in that SIP isn’t really the culprit. It’s just a protocol. Cisco phones running SCCP would have similar boot times to SIP handsets. I presume that Nortel phones running UNISTIM would also have similar boot times.
No, the real issue here is putting intelligence in the end-points vs the traditional approach of all the smarts being in the switch. Modern IP phones are little computers and they act like it in so many ways. Some of their behaviors, like lengthy boot times, are not advantageous.
Looking past the end-points to the core systems the post further points out that the boot time for an Asterisk server can be over seven minutes. That’s certainly much longer than a traditional phone switch. This is absolutely true.
While I’d like to draw this as an issue of dedicated hardware vs open platforms, or embedded systems vs traditional servers…it’s not that simple. Some SIP phones take a considerable time to boot but it’s variable across brands, models, even firmware versions. It’d be nice to see a reference table of all the common phones and their associated boot times. But what a project that would be to undertake!
Even just looking at Asterisk systems you’re going to see a range of results across the different hardware platforms offered. Is the machine in question a traditional server or a diskless appliance? My last Asterisk system was a diskless, fanless little box drawing under 10 watts. It booted pretty quickly all things considered, but not as fast as my old Panasonic KSU.
Sure, old school hardware based systems boot up fast. But remember how that same hardware tied your hands in so many ways? It tried to lock you into one vendor. It cost a king’s ransom to maintain or expand. Then you eventually had to replace it entirely when it simply couldn’t be expanded any further.
If you want the cost and flexibility advantages of commodity hardware then you may end up living with certain disadvantages as well. But none of that is specifically about SIP.
Comments are closed.
Even running an HA cluster for Asterisk (with the extra time of establishing primary), I would say 7 minutes is a bit on the extreme side. A normal * server on a decent server should be up from a cold start in under 5 minutes. Normally I use centos as the base and even include in that time the dependent services, such as ftp for the phone configs, mysql, etc. But, again, being able to choose the hardware can be a plus. If you want a super fast start, use great hardware and tweak the os. 🙂
@Fred
Yep, if you want super-fast startup…don’t use rotating disks. Use an appliance approach, and I happen to like Astlinux for that.
What about that idea of collecting info to build a database of typical boot times for common SIP phones? Handy? Or just a curiosity?
“No, the real issue here is putting intelligence in the end-points vs the traditional approach of all the smarts being in the switch.” Bingo! Wang proved this a long time ago. Their word processing terminals had 64K of RAM and provided instant-on word processing through a host computer located down the hall in a closet or computer room. To the user, it functioned like today’s standalone PCs. In reality, it was little more than a dumb terminal with an extra chunk of RAM to offload the WP software in chunks as needed. It made dumb terminals a thing of the past at least for word processing… until the slow-booting PC came along. Funny how history repeats itself.
Graves,
Thanks for putting in words what I felt when reading the same post.
Tsahi
That’s an awesome idea.