[En-Nut-Discussion] Power Strip Program
Dan Morphis
dan.morphis at gmail.com
Sat Mar 4 17:56:09 CET 2006
Mark,
I believe APC already has a product which does this on the market.
-dan
On 3/3/06, Mark J Elkins <mje at posix.co.za> wrote:
>
> I'm going to send out my idea anyway. I'm interested as to whether
> anyone has done what I'm proposing. I'm intending to make this an Open
> Source project.
>
> The Feedback I'm looking from from this list is:-
>
> Anyone done anything similar? (share?)
>
> Will the EtherNut be suitable - or should I look for a different
> platform?
>
> I'd like to think that ISP's and companies who house multiple machines
> in multiple cabinets would be interested in such a project
>
> Any other reactions?
>
> =========================================================================
>
> IPS - Intelligent Power Strip
> -----------------------------
>
> Scenario: The Dark Office at an ISP (an unmanned office)
>
> A room full of cabinets - each cabinet holding up to twelve Internet
> servers. Each cabinet is self contained in that there is a UPS and a
> Switch. The cabinet thus has two connections to the outside world, an
> UTP Ethernet connection to a central concentrator and a Power Cable
> from the UPS to its own floor socket - leading to its own circuit
> breaker mounted in a traditional electrical distribution box. The
> circuit breakers are rated at 20Amps. There is no Earth Leakage (bad
> idea with UPS's and Computers). The Electrical Distribution box runs
> to a 30KVA diesel generator kit and then on to Grid power.
>
> The cabinets contain machines. The primary requirement of any machine
> is that it can fit in the 19 inch wide cabinet. Usually these
> machines are successfully managed remotely by their owners thus visits
> to the dark room are few. The Machines get their power from a power
> strip running from the top to the bottom of the cabinet which is in
> turn plugged into the UPS.
>
> The primary reason for physical intervention are:-
>
> 1. The machine software goes faulty and needs to be hard rebooted
> (usually the machine is running Microsoft windows)
>
> 2. The machines power supply goes faulty - and shorts. The UPS will
> remove its protection and the trip switch will trip. This results
> in the whole cabinet loosing power.
>
> 3. The machines fans stop turning - things overheat and eventually the
> smoke comes out.
>
> The reactions normally consist of:-
>
> 1. The owner first checks reachability of his machine. He then calls
> the support desk to see if anyone else is in the dark room.
> Ultimately, there is a car journey to simply toggle the power off
> and on again.
>
> 2. Up to twelve clients jump on the phone to the Support Desk at the
> same time. Support Staff jump into a car and drive through to the
> dark room - locate and disconnect the faulty machine, restart all
> the other machine then contact the dead machines owner..
>
> 3. Usually the smell of smoke is detected by a neighbour. The building
> security calls the Support Desk and everyone runs (some to the
> problem - some away)
>
>
>
> Solution (How)
>
> Introduce the Intelligent Power Strip. (IPS)
>
> Socket Control: The IPS can selectively remove and re-apply power to
> any one of its twelve sockets. It can be controlled both locally
> and remotely. Each socket has a (normally closed) relay to
> interrupt supply to that socket. Remote control could be by a
> SNMP (Simple Network Management Protocol) command to the IPS's
> control unit, initiated via a Web Page interface. Local control
> could be by some simple push buttons and a character based LCD
> screen. It would be useful if the power status of each socket
> could be indicated. This could be as simple as a Red LED across
> the coil of the relay (i.e. it would light up in a POWER OFF
> state)
>
> Load Detection: The IPS monitors the current load on each socket.
> Each socket has a "Usual" value associated with it. For unused
> sockets, this is set to a reasonably high default. Once the
> equipment has settled, the "Usual" value can be automatically set.
> Consequently, if the usual value is breached by more than an
> additional 50%, the IPS can remove power to that socket. The
> "Soft Fuse" may allow for start-up loads greater than 50% - for a
> specific time interval - i.e. when switching on a server, the
> initial spinning up of disk drives consumes more power. Thought
> needs to be given to machines with multiple supplies from multiple
> locations - which may not be under the control of the local IPS.
>
> UPS Management: The IPS can monitor and partially control the UPS.
> Most UPS's have Serial based (RS232) management ports. These
> ports can provide information such as the current load status
> (Mains ON, Mains OFF), remaining battery time left, Current Load
> (Amps) and various voltages (i.e. Mains condition, Battery
> condition). Its possible to initiate commands - such as "Switch
> Off" - but experience shows this is a bad thing to do - except in
> an emergency (fire?). The UPS Management system needs to
> understand multiple types of UPS and perhaps have an option of
> providing custom commands for unknown models of UPS's. Up to two
> UPS's can be managed.
>
> Environmental Conditions: The IPS should also monitor a few other
> conditions.
> These include:-
> * The total current consumption from the UPS(s) (compare with
> what the UPS(s) states).
> * The Temperature of the air in the cabinet (top & bottom).
> * Smoke (fire) Detection.
> If either smoke is detected or the temperature rises above a
> certain value, the IPS can switch off all machines and perhaps
> also the UPS(s).
>
>
> The CONTROL BOX would have a "local interface" attached to its front.
> The box would be mounted on the inside of the cabinet in such a way as
> to not consume usable space for servers - ie it should not be rack
> mounted. The case size should not need to be larger than about 100mm
> square and 50mm deep. Optionally - it could be a 1U, full width unit
> if mounted in front of the Ethernet Switch (the switch facing the rear
> of the cabinet). Connection to the power strip would be via a
> suitable multi-core cable and connector. There will be a separate
> cables for the RS232 interfaces between the Control Box and the
> UPS(s). The AVR controller has a second RS232 port - which could be
> wired for a second (redundant) UPS. There will be a standard UTP
> socket for connecting the Control Box to the Ethernet switch mounted
> in the cabinet.
>
> The POWER STRIP will have at least twelve sockets with an additional
> Euro Socket at the top for powering the cabinet fans. The Sockets
> will be those used in South Africa (3 round pins). The sockets and
> mains voltage could be different for different markets. Typically,
> there will be a Euro Plug at the bottom of the strip for a connection
> to the UPS. There will be a "normally closed" relay per machine
> socket. There will be a Red LED per relay showing that the relay is
> now Open (the "unusual" state). If any relay is open, the Red LED on
> the front panel display should also come on. Each Power and Euro
> socket will have a current measuring device to enable the current for
> all devices to be calculated (total input current from the UPS, each
> machine socket and the fan socket). This could be as simple as a very
> low valued resistor in series with the supply, with a "sensor" wire
> from the load side and a single common wire from the supply side -
> although this would result in high voltages being present in the
> control box. **** Of everything in this document - this is the only
> area in which I am personally not sure how to proceed.****
>
> The Power strip should have a (selectable 120/240 voltage) power
> supply (transformer) to provide power to the control box and for the
> relay coils. The power supply would need to be strong enough to
> operate all relays at the same time. The current consumption of the
> Control Box will probably be very low. There may be more than twelve
> power sockets in order to power the Ethernet Switch and other similar
> devices that do not need to be monitored and controlled. The mains
> voltage (and frequency?) should also be measured and monitored.
>
> It is possible for other variations on the Power Strip, i.e. to have
> two separate supplies from two different UPS and to have alternate
> power sockets supplied alternatively from the two power supplies.
> This would allow for machines with multiple supplies to have redundant
> power sources from more than one UPS. In this case, the power supply
> for the Control Box should be able to take power from either or both
> sources - and report the failure of a UPS.
>
> Temperature and Smoke detectors could either plug into the Control Box
> to allow the sensors to be suitably located (near the top of the
> cabinet) or be fitted to the outside of the Control Box case. A
> second Temperature sensor could be mounted near the floor of the
> cabinet. As Hot air rises - this would provide a difference between
> the cooling air entering at the bottom against warm air at the top of
> the cabinet. It would be interesting to monitor the air warming
> factor against the power used in the cabinet. Local Interface
>
> The local interface (front panel display) could consist of:
> * A character based 4 rows by 20 character LCD display with back-light
> (Hitachi HD44780)
> * Green "power" LED (Flashing = No network connection)
> * Red "Alarm" LED (ON = user initiated, Flashing + buzzer = self
> initiated) Buzzer
> * 4 button input
>
>
> The Red LED and buzzer is to attract attention to a problem, the
> buzzer on entry to the cabinets vicinity and the LED to the cabinet in
> question (imagine a typical computer room with rows of cabinets). The
> buzzer can be silenced by acknowledging the fault. The Green LED
> gives a simple and quick status view that the unit is functioning
> correctly and connected to the Internet (it has obtained a valid date
> and time). The four buttons could be used to initialise/install the
> unit (i.e. Entering an IP address for the unit) as well as general
> status and control purposes. The LCD display can provide feedback and
> status information - i.e. the current IP address of the unit, the
> temperature and/or date and time.
>
>
> "Remote" Interfacing
>
> Via the Ethernet interface, numerous methods of control become possible.
>
> Device set-up could be as simple as a fixed, non-routable IP address
> (i.e. 192.168.0.1) which could be altered via a simple web interface.
> A more clever approach would be for the Control Box to run DHCP and
> actively fetch an IP address. Again, a simple web interface should be
> provided in order to select a final IP address.
>
> Interrogation and control of the Control Box could be solely done by
> SNMP (Simple Network Management Protocol) including fixing IP
> addresses but a web interface is almost certainly easier for people to
> administer with. A web interface need not be "pretty" (images are not
> desired!) but must be functional and work under any web browser (even
> text based). An alternative way of configuring the Control Box could
> be by up/down-loading configuration files using TFTP. Being able to
> fetch the current date and time via SNTP makes good sense (and to be
> able to display the time on the LCD display as well as for e-mail
> messages). In order to notify people when something has gone wrong,
> the Control Box should be able to send out e-mail messages using the
> SMTP protocol. There is no need to receive e-mail. Simple DNS
> services may be needed.
>
> -oOo-
>
>
> --
> . . ___. .__ Posix Systems - Sth Africa. e.164 VOIP ready
> /| /| / /__ mje at posix.co.za - Mark J Elkins, Cisco CCIE
> / |/ |ARK \_/ /__ LKINS Tel: +27 12 807 0590 Cell: +27 82 601 0496
> _______________________________________________
> En-Nut-Discussion mailing list
> En-Nut-Discussion at egnite.de
> http://www.egnite.de/mailman/listinfo.cgi/en-nut-discussion
>
More information about the En-Nut-Discussion
mailing list