All, very brief notes from my presentation at the E3D stakeholder
meeting. Slides attached.
/Lars
- Presentation went well; E3D people understood the suggested change
from routed network to point to point, with all traffic going through a
data centre. I believe this was generally accepted as a better way.
- Led to good discussion of picking a data centre, and the important of
doing so early. Craig asked relevant questions on what happens if they
have to move from DC A to DC B
- Craig seem not uninterested in using a data centre on-ring and close
to Tromsø that is becoming available
- Craig appear to be slowing moving away from 90’s style thinking
about computing and network. The NeIC project has been good in this
respect.
- Craig is also considering keeping all data in Skibotn for the first
couple of years. We might like to think about a grow-as-you-go plan
(i.e, from 100G from site to TB from sites, and
Several participants were very interested in the TB/s options and how
that changes long-term computing options.
- I suggested that we stop thinking about the network as a separate
thing, but rather consider it an integral component in the computing -
instrument, site network, site computing, WAN, DC network, DC computing,
storage. The WAN can go anywhere in that picture, really.
- None of this was conclusive. It was a conversation starter. My
immediate guess is that they will go for 2x100G per site now, moving to
much more capacity in the future. Possibly, this will be linked with a
model where they do all computing at Skibotn now, and move to a DC in
2-3 years.
--
Lars Fischer - Strategy & Policy, NORDUnet
<lars(a)nordu.net>, +45 2288 1729, @lpfischer
