[Osmf-talk] Consultation on fundraising strategy

[Paul Norman]

On 2023-03-17 10:15 a.m., john whelan wrote:
> Perhaps it is time to look at costs.  Ten years ago the cost of 
> electricity for a Windows server running 24 / 7 / 365 a year for 
> three years was greater than the cost of the hardware and I don't 
> believe that has changed much.  Some other vendors say their systems 
> are better than this but when you look at the cost per MIP their cost 
> is higher.  The machines themselves are more expensive.
>
> ARM based servers with their RISC architecture might do a little better.

Our AWS sponsored sponsored server is running on an ARM server. For 
various supply reasons, ARM has not been the best option when we've made 
purchases.

> So probably our greatest cost is electricity either directly or 
> including cooling.  There is a reason why many data farms are located 
> where they are and the cost of electricity is one of the reasons.

The OWG is very aware of our power usage and takes steps to minimize it.

> Servers are difficult to size, in general you buy one with more 
> capacity than you need.

Servers aren't difficult to size if you have a good prediction of future 
hardware requirements, and our predictions have historically been good. 
Growth has historically been very stable.

> Would running OSM in the cloud be cheaper?  I'd probably go with 
> Microsoft and switch to SQL Server but then we all have our personal 
> preferences but mine are based on leading a team of DBAs looking after 
> more than 100 servers running a variety of operating systems and 
> databases.

No, it would be more expensive. We use the cloud where it makes sense, 
but for many of our services, the operational costs of a year in the 
cloud exceed the capital costs associated with five years of that service.

> The second thing I'd look at is the efficiency of getting money from 
> donators to OSM.  PayPal takes a cut on the way and a more hidden cut 
> in the exchange rate.
>
> Bank wires are expensive at both ends.  Within Canada we can do low 
> cost interact transfers, within the UK again payments into a bank 
> account are low cost and very fast.
>
> One thought would be talk nicely to one of the charities that gives us 
> some funds today.  Sometimes these funds can be earmarked for a 
> particular project.  Mapping? Perhaps and let people wishing to donate 
> to OSM donate via. xyz charity.

None of the above really matters to what this thread is about. The 
predicted gap is 500k GBP - efficiencies could reduce this, but it's not 
going to change the need for a different fundraising strategy than we 
have right now.