[Server-devel] location and muscle of the school server

[Sameer Verma]

Bryan Berry wrote:
> Sameer ji,
>
> You make a good point. We have only have 14 hours per day of electricity
> in Nepal so the issue of power is very acute for us.
>
>   
Bryan ji,

I was thinking of a scenario with two hours of power - something my 
family in India is familiar with. Interestingly, come election season, 
the profile inverses. Then they get 22 hours of power :-)

> A school server that caches websites, serves up Moodle pages, serves as
> a mesh portal for the mesh network, and does some kind of content
> management will definitely be a power hog compared to the XO.
> Additionally, it will require a lot more maintenance than the XO's.  I
> am confident that kids and locals will be able to repair the XO's but I
> doubt many communities (at least in Nepal) will have local people w/ the
> Linux skills to manage the School Server.
>
>   
Local skill set is key for long-term sustainability of any project. If 
the locals do not have such skills, the project will fall apart once you 
(or your team) leave. Ideally, the skill set should be minimal so that 
maintenance and repair aren't showstoppers. I am very impressed with the 
design of the XO in terms of how easily the machine comes apart and goes 
back together. On-site repair becomes so much more feasible.
> We are considering centralizing many aspects school server for our fall
> deployment. Our Networking guy, Mahabir Pun, has had a lot of success
> connecting rural villages to the Internet w/ long haul point-to-point
> wireless links that connect to the nearest city. He places a caching,
> e-mail, and voip server in the city. This works because the wireless
> links are high-speed and quite stable. 
>
> We intend to put the servers on site for our April pilot but we may
> centralize them later.
>
> Centralizing servers won't work well if bandwidth to the schools is
> limited and/or expensive. 
>
>   
Indeed, but I wasn't implying that every document has to be downloaded 
for each XO in the school. There of course needs to be some caching 
mechanism that is low in power consumption, but does not necessarily 
host the entire system. It appears that we have multiple scenarios for 
server use. Size of the XO pool, on-site power availability and Internet 
access will determine the most suitable model. I am not convinced that a 
"one-size-fits-all" approach will necessarily work.
> As a side note, in Nepal we will try to avoid satellite Internet
> connections wherever possible. The monthly bandwidth costs are just too
> expensive over time. Comparatively, long-haul wireless links w/ cheap
> equipment from Motorola, Microtek, or Ubiquiti are much more
> cost-effective. 
>
>
>   

Sameer

-- 
Dr. Sameer Verma, Ph.D.
Associate Professor of Information Systems
San Francisco State University
San Francisco CA 94132 USA
http://verma.sfsu.edu/
http://opensource.sfsu.edu/