High availability topic session objectives

1
High availability topic session objectives

• List of issues/questions/topics
• Make a diagram of the whole (hardware) system
  (num i/o crates, etc, etc, etc). Include remote
  operations centers (GAN)
• Need a list of functionality and high level apps.
• Need to pull out feedback-specific,
  security-specific, etc requirements
• Make a table of front-end items and the level of
  HA each needs.
• Failure modes analysis
• Do we count VHDL code in all this?
• Almost everything now appears in the middleware
• Do we include multiple control rooms (GAN) in
  cost model?
• Every control system subsystem needs a failure
  modes/effects analysis.
• Tom will publish availability formulae.
• Goal make unavailability go to zero.
• Goal make no unsafe failures for the machine.
• Goal cannot damage the machine or equipment from
  the control room or high level apps, or middleware

2

• How do we include high availability into the
  controls baseline architecture for the reference
  design? What is our baseline for costing?
• We are going to do an HA control system!
• Front-end hardware
• Switch from VME to ATCA in baseline for RDR.
• Need to know which things need hardware
  redundancy at front-end level.
• Servers etc
• Need to budget for HA relational databases,
  software, central computing, data archivers, etc
• Central computing archiving data backups, etc
  is part of controls domain. MIS computing is not.
• Control system
• We will provide an adequate infrastructure for
  implementing HA functionality.
• We have to make the middleware reliable
  (especially).
• Cost will be in implementing the recovery
  policies.
• To what level do we go with HA software in the
  various layers?

3

• RD Roadmap (starting point)
• Claudes proposal
• ATCA EPICS iocCore shelf manager ICE
  middleware
• Claude to set up discussion with interested
  parties.
• Develop an AMC module to plug in / integrate with
  Claudes ATCA crate.
• How do we utilize the test stands (ILCTA, ATF,
  TTF)? Feb 9/10 Fermilab mtg

4
High availability questions/topics

• HA for central computing and data storage?