[OSM-dev] [OSM-talk] Tile and dev down

[Nick Hill]

Hello Etienne

If there is a local power failure, given the upstream network 
infrastructure will not be on UPS, we will inevitably experience some 
service outage. If we can ensure the servers come back up in the event 
of power interruption, all should be well.

There are some other aspects of UPS which come into the equation; 
potential data loss, potential UPS failures and increased energy demand 
through UPS systems.

Data loss is the biggest concern. UPS can help, but this is really best 
addressed through backup strategy, as there are so many potential points 
of failure which could lead to data loss, not related to power source 
failure.

Potential UPS failure; If the UPS failed, all systems will be offline 
until someone can bypass the UPS. UPSs also need periodic maintenance 
and replacement of lead-acid accumulators. This brings in disposal 
issues and environmental issues.

Increased energy demand. UPS systems loose some supply energy. 
Currently, this financial cost would be externalised from the OSM 
project as we don't (currently) pay for the direct electricity and 
electricity used for air conditioning cooling. However, OSM servers are 
built with environmental consideration and use recycled materials where 
appropriate. OSM can therefore claim to be an environmentally conscious 
project. I personally try to live with a small ecological footprint, so 
I prefer options which don't involve energy and material consumption.

What's next for OSM?
Hardware-wise, I will be installing a new tile server this week with 
much more on-line storage. We need an intelligent tile invalidation and 
rendering scheduling system. This new server I intend to install this 
week (Athlon 64/400GB) will provide the smooth slippy map rendering we 
crave *once we have the as-yet unwritten invalidation/scheduling 
system*. For more info:
http://lists.openstreetmap.org/pipermail/dev/2006-September/002125.html

Hardware-wise, the next best thing we can do is swap the core of the new 
A64 server to dual-core Perhaps in a month. When 2Gb fast DDR DIMMs 
become affordable, upgrade the memory on DB then move dual-channel 
memory from DB to tile++.





Etienne wrote:
> On 9/24/06, *Nick Hill* <nick at nickhill.co.uk 
> <mailto:nick at nickhill.co.uk>> wrote:
> 
>     A power interruption at 07:09 saturday morning possibly explains dev and
>     tile being down. These machines don't automatically power up after power
>     interruption.
> 
> 
> Nick
> Thank you for the update.  Do you need a UPS for these servers?  We have 
> the money - what do you need?
> 
> Etienne
> 
> 
>  
> 
>     I have the tile++ machine to (re-)install. I hope to install tile++,
>     and
>     restart dev and tile tomorow. I will also take db down for a while, to
>     increase the size of the root hard drive so we can automate a fast,
>     (short read-only time) backup process.
> 
>     I may also perform surgery on tile and dev to bypass the soft power
>     systems, in the hope they will come up automatically in the event of
>     future power failures.
> 
> 
>     System maintenance/service interruption times on Monday or Tuesday TBA.
> 
>     _______________________________________________
>     talk mailing list
>     talk at openstreetmap.org <mailto:talk at openstreetmap.org>
>     http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>     <http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk>
> 
>