(Jan. 2004)
I have decided (equipment is ordered, just waiting on it now) to go for two * servers running identical hardware:
We're talking CDN$3800-4200 systems. The dual power supplies really adds a lot to the cost, and the TE410Ps aren't cheap, either. These are SuperMicro 4U systems. I looked at Dell and yes with warranty and everything the systems were nice, but I couldn't justify purchasing proprietary PC hardware in the end.
Both these systems are online all the time, but in a hot-failover instead of load balancing configuration. My telco's PRI (one right now) plugs in to a little black box which has three T1 ports on it. PRI goes in one side, and two * servers to the other two ports. The box monitors the T1 for alarms and will physically route the T1 to the other port should any alarms come up. I saw this box on ebay once and have not been able to remember the name of it. If anyone has a name for a manufacturer, I would certainly appreciate it. It's my missing link at this time.
Each power supply plugs in to a UPS which is connected to alternate sides of the incoming 240VAC. (remember electricty 101? North American power comes in as two 120VAC lines and a neutral and are 180 degrees out of phase. If you're lucky enough to get three phase it's three hots 120 degrees apart, but we don't have that at the colo, just at the plant. :-) — that way if we lose a phase (it's happened before) we're covered, and if we lose both phases the UPSes are there.
I do not have redundant internet links but will have two network switches, one on each UPS — I don't care if VOIP calls don't go through, but I need the closest to be able to communicate. :-)
Anyway yes if a * box goes down you lose all your current calls, but that isn't a (big) issue for me. The bigger issue is that when I go to place a call that it gets through.
I will have one POTS line at the office, but it won't be hooked up to a computer. It will be wired to a regular phone with no bells and whistles, and will be for emergencies. It won't be equivalent to any of the DIDs.
The advantage to this setup is that while it breaks the hot-failover, I can definitely test things on one box without causing hassles on the other, and upgrades are trivial.
As far as a PRI going down — it's as likely as your POTS lines going down, IMO. Perhaps a little more so but in the end I have been in telephony (dial-up internet) for close to 6 years now and on anything from single T1 PRIs to multiple DS3s, I have never had their end go down once it was provisioned.
I did have the very unlucky experience of having a double-controller failure on an M13 with no controllers handy once... And to add insult to injury UPS had lost the package from CAC not once, but twice.
Go back to Asterisk failover
Case 1: Twin design
I have decided (equipment is ordered, just waiting on it now) to go for two * servers running identical hardware:
- Dual power supplies
- Linux software RAID1 on hot-swappable SCA drives
- Single-proc P4 (dual proc mobo for expansion)
- TE410P
We're talking CDN$3800-4200 systems. The dual power supplies really adds a lot to the cost, and the TE410Ps aren't cheap, either. These are SuperMicro 4U systems. I looked at Dell and yes with warranty and everything the systems were nice, but I couldn't justify purchasing proprietary PC hardware in the end.
Both these systems are online all the time, but in a hot-failover instead of load balancing configuration. My telco's PRI (one right now) plugs in to a little black box which has three T1 ports on it. PRI goes in one side, and two * servers to the other two ports. The box monitors the T1 for alarms and will physically route the T1 to the other port should any alarms come up. I saw this box on ebay once and have not been able to remember the name of it. If anyone has a name for a manufacturer, I would certainly appreciate it. It's my missing link at this time.
Each power supply plugs in to a UPS which is connected to alternate sides of the incoming 240VAC. (remember electricty 101? North American power comes in as two 120VAC lines and a neutral and are 180 degrees out of phase. If you're lucky enough to get three phase it's three hots 120 degrees apart, but we don't have that at the colo, just at the plant. :-) — that way if we lose a phase (it's happened before) we're covered, and if we lose both phases the UPSes are there.
I do not have redundant internet links but will have two network switches, one on each UPS — I don't care if VOIP calls don't go through, but I need the closest to be able to communicate. :-)
Anyway yes if a * box goes down you lose all your current calls, but that isn't a (big) issue for me. The bigger issue is that when I go to place a call that it gets through.
I will have one POTS line at the office, but it won't be hooked up to a computer. It will be wired to a regular phone with no bells and whistles, and will be for emergencies. It won't be equivalent to any of the DIDs.
The advantage to this setup is that while it breaks the hot-failover, I can definitely test things on one box without causing hassles on the other, and upgrades are trivial.
The telco side
As far as a PRI going down — it's as likely as your POTS lines going down, IMO. Perhaps a little more so but in the end I have been in telephony (dial-up internet) for close to 6 years now and on anything from single T1 PRIs to multiple DS3s, I have never had their end go down once it was provisioned.
I did have the very unlucky experience of having a double-controller failure on an M13 with no controllers handy once... And to add insult to injury UPS had lost the package from CAC not once, but twice.
Hardware to secure fallover
- Junghanns: ISDNguard
- Western Telematic: A/B Fallback Switches
- Cornet Technology: Intelligent Switches
- DataProbe: Line Protection Switches
- Redfone Communications: FoneBridge T1/E1/PRI Bridge
Go back to Asterisk failover