Information Technology Reference
In-Depth Information
Chapter 12
Using Existing Products
\Next item," said Marcus. They were back in the old drawing oce area,
and had spent some hours working their way through a technical audit re-
port. This audit had been one of the results of the ongoing negotiations with
Factotum, and most of its concerns were about resilience.
\Recommendation 27b identies problems of consistency between the in-
stances of repair orders that are duplicated in dierent service centres,"
Eleanor read from the pink sheet in front of her. \They are right; there is
potentially a problem here, but it means adding transactional controls to all
the order update steps. It will be a huge job, particularly when taken to-
gether with the improvements to archiving and disaster recovery we talked
about before coee."
\What about replication transparency? That worked well for the logistics
server, as I remember," said Nigel, hopefully. \Maybe," Eleanor looked out,
for a moment, over the cubicles on the oor below. \But I don't think so; we
would still need to factor in the channels to archival, and linking that whole
subsystem with replication would be too damaging to performance." \Just ex-
plain that to me again will you," broke in Marcus, looking puzzled. \I thought
transparencies hid all that sort of thing." \They hide complexity from the
programmer, but the problem here is the real runtime costs of coordinating
a larger number of objects. In any case, what we want from archival is re-
silience, which means keeping the boundary to it as clean as possible." There
was another pause.
\We could try a dierent approach," said Alex, quietly. \If we don't want
the complexity of all these archive-based resilience mechanisms, can't we avoid
them completely?" \But we spent half the morning identifying reasons why
we must have them," snapped Claire. \Not quite. We have shown that we
need consistency and resilience, not that we need to implement all those horrid
mechanisms ourselves." \So what's the alternative?" \Don't build the subsys-
tem; buy what you want as a service. There have been a lot of developments
recently in cloud computing. You could buy resilient, consistent storage from
one of the global suppliers and use it as a secure mailbox between the local
centres. That way the external supplier has to worry about the coordination
and the disaster recovery. You agree on a suitable SLA with them, and they al-
ready have all the resources needed to provide it. All you have to worry about
167
