B513 Meeting, 6th June 2001
Present:
Tony Cass, Fabio Corsanego, Anne Funken, Jukka Lindroos, Dick
Minchin, Dave Underhill, Tim Whibley
1.
Previous Minutes
- Some
comments from Anne regarding the electrical supply.
2.
Vault Air Conditioning & Smoke extraction
- Fabio Corsanego confirmed that a smoke extraction
capability of 1m3 per 100m2 of surface area is
required. The fresh air supply must be capable of providing enough
replacement air during smoke extraction.
- Jukka commented that it would be preferable to install
ducting on the S-030 side of the vault wall to met smoke extraction needs
rather than using the existing plenum.
- Various options for the installation of smoke curtains
were discussed. One concern is that these would interfere with the
circulation of air for cooling. Although there are smoke curtain systems
that descend automatically when smoke is detected, these are more
complicated and the cable trays present obstacles. It was agreed that
installing two lines of smoke curtains that divided the vault into three
volumes would be a reasonable solution. Each volume would have at least
one cooling unit so the impact on air flow would be minimal.
- Fabio commented that the most effective way of limiting
smoke damage would be through the division of the space into separate compartments.
This would created difficulties, however, and would certainly require the
need for ducting to meet smoke extraction requirements.
- Jukka agreed that there was now sufficient information to
cost an air conditioning and smoke extraction solution for the vault.
3.
Fire Protection issues
- Fabio agreed that the emergency exit into the plant room
on the west side of the vault could be suppressed provided a new exit is
created on the south wall. The new exit would be in line with the door
from the old MG room into the south corridor.
- It was agreed that a VESDA smoke detection system would
be installed with a granularity corresponding to either the smoke curtain
sections or to individual rows of racks. The VESDA system(s) would be able
to power off equipment in the affected area. There was some discussion
about the level of sensitivity of VESDA detectors and the possibility of
false alarms. However, the sensitivity levels are adjustable and
experience at CERN indicates that most “false” alarms are in fact caused
by smokers.
- On the question of fire extinction, Fabio commented that
although it would be technically feasible to install a gas based
extinction system in the vault, such a system would be very expensive and
rather impractical. The most likely incident would lead to a release of
smoke possibly followed by a small fire which could be extinguished by the
on-site Fire Brigade. He felt that the risk of a rapid conflagration is
rather low.
- A “Hi-Fog” water based system could be considered,
especially if distilled water is used to reduce the impact on equipment.
Dick Minchin commented that tests some years ago had shown that although
sprinklers extinguished fires beneath the heads, the downdraught fanned
the flames of fires that were between sprinkler rows. Fabio said in
response that the downdraught from Hi-Fog style systems was much lower.
Additionally, any sprinkler system in the vault would have the heads
aligned above the equipment rows.
- Hi-Fog systems will be demonstrated at CERN in the near
future and IT representatives will be invited. If possible, we should also
try testing the effect of such systems on CPU servers—especially as tests
on obsolete systems would be adequate.
- Finally, Fabio asked that particular attention be paid to
the cables installed in cable trays or underfloor. These should be fire
retardant and halogen free in accordance with the provisions of IS 23.
4.
LV distribution in the vault
- Anne has had a meeting with MGE to discussUPS solution in more depth. There is still no
clear answer concerning the practicality of a UPS to support a 200kW load
for 2hours, but MGE have confirmed that the height of battery racks is
limited to 2m. Anne suggested that the power needs of the basic computing
infrastructure should be reconsidered—it is unlikely that such equipment
has a power consumption of 200kW today.
- The cost for a vault-wide distribution, as opposed to “per
farm” installation, won’t be known until the next meeting. However, one
point that has been noted is that the vault wide distribution system is
able to support a higher power load than could be supported. Tony
commented that this is anyway clear given that air conditioning capacity
for the vault is only 500W/m2 whilst equipment power loads can
be up to 4kW/m2. To avoid any problems, some form of current
monitoring system will be needed. Tony asked that a display be made
available in the operator control room for any system that is installed.
- Anne is concerned about the level of harmonic distortion
caused by PCs. Initial tests on her own desktop PC suggest the need for
harmonic compensation within the PDUs. Tony will arrange for tests to be
carried out on full racks of PCs.
- In terms of planning, Anne said that once go ahead is
given, there would be a delay of 6 weeks to have the components delivered
to CERN with installation requiring a further 6 weeks.
5.
18kV supply to B513
- The study to include a new substation for B513 in the
sitewide 18kV loop is well advanced and could be ready by the end of
September. As nearly all of the necessary equipment is covered by existing
contracts the actual work could be done in 3-4 months with a further 2
months for commissioning and reception. However, all of this assumes that
we know where the equipment is to be installed.
- An initial review of the building plans suggests that it
would be difficult to install all of the equipment in B513. Although the
area available is probably sufficient, access is difficult and the 18kV
switchgear that would be required needs significant clearance.
Additionally, safety considerations probably mean that the transformers
have to be installed outside.
- Anne will investigate these issues further before the
next meeting, whilst Tony will discuss the feasibility of an external
building within IT.
6.
AOB
- Next meeting on Tuesday 26th June at 15:00.