nanog mailing list archives

Re: Rack Locks


From: "Bob Purdon" <bobp () purdon id au>
Date: Wed, 25 Nov 2015 13:21:49 +1100

So what did we do. I used to use a relay type system in 2007-10 in my
previous data centre life. It¹s pretty good but a bit ³industrial². It¹s
also so 2007 (even 1990) and doesn¹t scale well when you are trying to do
3,000 racks and 6,000 doors per facility.
Part of the scaling issue was the door locks on that system were conventional solenoids, which from memory needed about 
1A @ 12VDC to fire.  If a customer had 30-40 racks (and a couple did in that facility), you'd need to potentially fire 
60-80 doors, or need 60-80 amps available (I have a recollection we used a 12V SLA battery to ride out those peaks).  
Additionally, monitoring lock status would have needed separate wiring and separate inputs.  Cabling was a star 
topology (each rack directly back to the controller).
The TZ locks use a fraction of that power - from memory, only a few amps to do a pod of 30 or so racks.  Firing a lock 
is measured in milliamps, not amps.  The locks are controlled over RS485, so you get lock control and monitoring over a 
single cat-5.  From memory the cable topology is technically hierarchical, but you could loosely consider it to be a 
bus.  Overall, vastly superior to the 'industrial' style system.
I looked at the APC electronic
locking system, but the big issue is that some fool in product decided to
remove radius authentication, allowing a decent independent
command/control capability.
At the time the available version of the product didn't deal with too many racks, which also meant a lot of under-floor 
power outlets to feed the controllers).  I think they were coming up with a denser version, but I didn't see it.


Current thread: