HPE Community read-only access December 15, 2018
This is a maintenance upgrade. You will be able to read articles and posts, but not post or reply.
Hours:
Dec 15, 4:00 am to 10:00 am UTC
Dec 14, 10:00 pm CST to Dec 15, 4:00 am CST
Dec 14, 8:00 pm PST to Dec 15, 2:00 am PST
BladeSystem - General
cancel
Showing results for 
Search instead for 
Did you mean: 

Expected behaviour of Blade chassis during switchover to generator power ?

 
chuckk281
Trusted Contributor

Expected behaviour of Blade chassis during switchover to generator power ?

How much time do I have I to get to standby power? That was a question posed by Stephen's customer:

 

********************************************************

 

Can anyone answer the following two questions - can't find this info in QuickSpecs etc

 

1.How long can a C-class blade chassis "survive" during a switchover of power from one source to another - eg from AC to generator?

 

2.What is the time lag for a PSU to change from its Standby state to fully active when one half of the chassis loses mains feed?

 

 

The background to this is that customer tested the emergency generator of their datacenter (which has always worked fine for rack servers) but, during switchover to generator power, 4 chassis (out of a total of 15) experienced a complete chassis power off/on reboot (very bad) and 6 did not reboot (good) but reported "insufficient power to support all 16 blades (bad).

 

5 chassis behaved as expected (hoped) and did not experience any reboot or insufficient power problems (good).

 

All of the chassis which rebooted had "Power Saving Mode - ON" whereas 4 of the 6 which reported insufficient power had "Power Saving Mode = OFF" (but 2 had it ON).

 

In all cases of failure, the OA reported one or more PSUs as going "unknown".

 

This case in under investigation by HP Support (EMEA case reference 4617757219) but I would be interested to hear if anyone else has experienced similar issues when switching input power supply to blade chassis.

 

**********************************************

 

Monty provided some input:

 

**********************************************

 

The exact answer depends on the load, power supply capacity and how long the power supply can sustain full output when the input power is removed.

 

Most HP server power supplies are only designed to deliver full output for a single cycle of AC input power lost – 16 msec for 60Hz power.  This is much shorter than the time it takes to switchover to generator power (seconds).  Sustaining server power across loss of a single AC feed while the generators are coming online sounds impractical.  This is typically done using UPS to back up a single AC feed.  The UPS can sustain full power while generators are being started and coming up to speed – or just long enough to gracefully power down the servers.

 

If the DL server was only consuming 400W and there were two common slot 1200W power supplies, it could survive 2.5x longer than a DL consuming 1000W with those same power supplies where the computed ‘hold’ time is total load at that moment, divided into the number of watts of total capacity, multiplied by power supply hold time of 16msec.

 

  • Ex: DL using 400W with two 1200W power supplies.  Expected time to brownout when all AC inputs fail: (1200W*2)/400W*0.016 = 96 msec.
  • Ex: DL using 1000W with two 1200W power supplies.  Expected time to brownout when all AC inputs fail: (1200W*2)/1000W*0.016 = 38.4 msec.

 

For the blade enclosures, you have the same math, although Dynamic Power Savings does provide an additional delay on bringing all power supplies online if there is a loss in AC power.

 

The enclosure Dynamic Power Savings mode is designed to maximize power supply efficiency by placing pairs of power supplies in standby mode to increase the load on the remaining power supplies to near 50%.  This design point protects the enclosure from loss of one AC ‘side’, as the active power supplies on the remaining side can easily handle double the instantaneous load, and the enclosure sets all other remaining power supplies on that side to active to handle any dynamic load changes.  The enclosure power micro performs the real-time calculations for Dynamic Power savings many times a second but will likely not be able to set the remaining power supplies active in less than a power cycle.

 

The computed ‘hold’ time for an enclosure losing all AC input power depends upon the total enclosure load at that moment, divided into the number of watts of capacity, multiplied by power supply hold time of 16msec.

  • Ex: c7000 enclosure with six 2500W power supplies, with enclosure Dynamic Power Savings disabled.  Sixteen server blades, interconnects and fans consuming 2250W (OS idle). Expected time to brownout when all AC inputs fail: (2500W*6)/2250W*0.016 = 107 msec.
  • Ex: c7000 enclosure with six 2500W power supplies, with enclosure Dynamic Power Savings enabled.  Sixteen server blades, interconnects and fans consuming 4500W.  DPS would enable only four power supplies – all of which would be delivering 1125W.  Expected time to brownout when all AC inputs fail: (2500W*4)/4500W*0.016 = 35.5 msec.

 

Hope this helps – you must have very quick generators to turn on and provide good power in a fraction of a second.  I would bet on needing UPS for this type of outage.

 

The Enclosure Dynamic Power savings is designed for customers with redundant AC feeds and can sustain loss of a single AC feed – not both feeds.

 

***********************************************

 

Any additional input? Have you designed and tested your power backup for your blades? Can you share your experiences?