As we've mentioned a couple of times over the last few months, we've been working hard behind the scenes on improvements to everything 
So far we've upgraded a LOT of the server hardware on our shared network and increased the bandwidth allowances.
In order to make all that happen we joined INEX and RIPE. We then moved all our equipment and that of our clients onto our new IP space gradually, in order to avoid downtime and disruption.
Our network staff, having carefully planned the new network rollout, are installing new Cisco switches across the entire network.
What this means in plain English is that we're removing any possible bottlenecks on our internal network ie. between your sites / servers and the internet.
To do that we have to switch the network cables over from one set of switches to another. If only there was a magical way to do that on several hundred servers at once!
Our CTO, Paul Kelly, explains it as follows:
Are there up sides and down sides to this work? The answer to that
question is very simple. Both.
Downsides:
We have to re-patch 5 cabinets worth of servers into new
infrastructure.
What does this mean?
Ok well we basically have all the new equipment in
place and configured and linked into the core network.
We have all the cabling ready to go.
All we have to do is pull out your old network cable and plug in the new one.
There may be some convergence times for mac addresses to appear on the
new port in the core of the network. This should be brief but could be
5-10 minutes for our less busy customers.
Upsides:
Well you will now be connected into one of the most modern
networks in the industry in Ireland today. We'd go so far to say it is
one of the best designed networks that any of us have come across. We
say this with pride as we've spent a lot of time, energy, money and a
lot of thinking to get it to this point.
We've enabled a feature called "portfast" on all customer ports so as
the time taken to negotiate the connection speed and detect if there is
a loop takes less than 3 seconds. So for the most part none of you will
notice anything.
We've increased resilience from all our equipment cabinets. What we mean
by this is that we've now got 2 x links into the core network, into
separate core access switches and we're using rapid spanning tree to
fail over from one to the other in the event of a switch failure.
We're also giving more aggregate bandwidth per customer cabinet. We used
to allow 100mbit per pair of switches. We're now going to allow 1Gigabit
from each cabinet into the core network and from there to the Internet!
Ok after reading all that you now need to know when this is happening.
We want to do it next Thursday the 25th of Jan 2007. We hope to start
around 9pm and finish by midnight. We don't want to rush things, we want
to verify as we go that customers are coming back up. For this reason we
are scheduling a 3 hour maintenance window 21:00 - 00:00.
A point to note: Most customers who have gone live in the last 12 months
will already be on the new network infrastructure and this maintenance
window should not affect them. However we're applying this to all customers
just in case we need to make any global changes to the network to
accommodate extra capacity or link failures.
So basically, as they say, "it's all good".
If anyone has any questions or queries please let us know.
