B513 Planning Meeting, 18^th May 2001

Present: Tony Cass, Anne Funken, Jukka Lindroos, Dick Minchin, Dave Underhill, Mario Vergari

1. Civil Engineering Work for the Vault

· Tony presented a plan to remodel the Vault. The relevant elements of this plan were

a. Removal of S-407, 408, 409 & 432 (former MG room ventilation rooms), integrating the space into S-034
Here, Jukka commented that these areas house fresh air input and hot air extraction ducts and that he planned to make use of these for smoke extraction (see below). It was concluded that these rooms could be removed, but that the plenum between S-030 and S-034 should probably remain.

b. Removal of S-406

c. Construction of a wall along the line of pillars separating S-412 and S-040. This wall to extend into S-411 (the UPS battery room), cutting this in half.
Anne commented that she had been considering housing a future UPS system in B513, thus reducing the size of any external “energy centre”. Constructing this wall would severely limit the technical space within B513. Given this point, it was agreed that building a wall in line with the existing doors into S-411 could be considered. This would make some 200m² available for use as technical space.

· Anne presented an initial estimate of 161K for

a. Removal and evacuation of existing false floor and partitions,

b. Painting the concrete floor,

c. A false floor of 1200, and

d. Stairs, access ramps and barriers.

2. Smoke Extraction for the Vault

Following discussions with Fabio Corsanego of TIS, it is clear that provision for smoke extraction in the vault is required. Jukka has reviewed the Swiss and French codes for computer centres (the latter is considered as the reference for CERN), and, for a room the size of the vault, provision must be made for an air flow of 10m³/s.

· Jukka considers that these requirements can be met without the need for overhead ducting.

a. Smoke can be extracted via the chimney above the archive store—although this will require the installation of a fan. As noted above, smoke could be routed to the chimney either via the plenum between S-030 and S-034 or via a duct on the S-030 side (smoke is extracted at ceiling level only).

b. Fresh air can be routed to the “far side” of S-034 via ducts under the false floor which connect to the fresh air inlets in S-406 and S-407.

Jukka drew attention to the fact that a fire in winter could lead to an inrush of air at temperatures as low as -12°C. If this could damage equipment then a heat exchange system to warm incoming air would be needed (and could be provided easily given the heating plant already present in B513).
Tony mentioned that he and Dave had discussed some aspects of fire protection during a visit to discuss access control. Marc Vadon had suggested that barriers be installed at ceiling level to limit the spread of smoke. This was felt to be especially important as it is unlikely that computer equipment will ever be halogen free—and so even a small fire could cause extensive smoke damage.
It was felt that such a system of barriers would only make sense if they were installed in parallel rows at right angles to the S-030/S-034 plenum. However, Jukka was concerned about the potential interference with air flow in normal circumstances; he will discuss this issue with Marc Vadon.

3. HV Supply and UPS for B513

Anne had prepared a paper with size and costings for various UPS options. The key points are as follows.

				Space
	Load	Time	Cost	UPS	Batt.	Comments
1	2MW	10’	630K	80m²	60m²	Required to cover brief interruptions and auto-commute between Swiss & French supplies
2	200kW	30’	80K	10m²	18m²	Additional protection for the basic computing infrastructure (but if auto-commute fails, power outages of up to 120’ can be envisaged).
3	200kW	120’	130K	10m²	75m²	Given the number of battery modules required, a static UPS solution may not be advisable for this load/time combination.

In addition to the MTBF concerns for the 3^rd solution (which Anne will be discussing with MGE next week), the air-conditioning implications also need to be understood. There may be a requirement to support air-conditioning in the vault if some 100kW of equipment is to be kept running for 2 hours.
Equally, the air-conditioning load has to be included if the full equipment load has to be supported during power cuts of more than 10 minutes. A standalone static UPS system is unfeasible; either diesel backup or a rotary UPS must be installed.
Housing a combination of solutions 1+3 in B513 does seem possible, especially if the extra technical space is made available. However, arranging an orderly installation and changeover from today’s UPS could be rather difficult.
Planning for the inclusion of B513 in the 18kV loop continues—including costings for an “energy centre” that could house either a static or rotary UPS system. It was noted that issues such as noise and air intake/exhaust provision have to be taken into account whrn housing diesel engines (either alone or as part of a rotary solution). These factors might prevent installing diesel engines in B513 and would increase the cost of an external energy centre.
Finally, Anne was concerned that the upstream electricity supply would not be able to support the additional 500kW load being projected for the vault until B513 is included in the 18kV loop. Tony pointed out that, at least initially, it was more a question of installing equipment in the vault rather than in the machine room and power consumption would not leap by 500kW. On this basis, Anne agreed that there would be no problem installing equipment in the vault, but said that power consumption could not increase by more than 200kW.

4. Low Voltage Distribution in the Vault

· Anne presented an initial proposal for a low-voltage distribution system in B513. In addition to meeting the likely power demands, this solution makes use of outlet boxes pre-connected to busbar trunking (Normabarres) in order to limit the need for specialised electrical expertise when connecting equipment. Those present from IT commented that

a. it was not acceptable for sockets to be provided at floor level; all cabling must run under the false floor, and

b. both single- and 3-phase sockets must be provided.

Tony also commented that Anne’s costings for the distribution system should end at the sockets. The costs of any cabling and equipment required at the rack level are covered elsewhere.

The proposed distribution scheme includes normabarres for physics load and for basic computing infrastructure. Equipment with dual power inputs could be connected to both networks. There is no provision for a “second level”switchboard-style system connected to two normabarres in order to provide a dual-feed supply for equipment with only one power input. However, this could be added.

· The proposed switchboards support a 400A load, trip the output feeders (normabarres) on power loss and support remote switching (on or off) of the outputs. One 400A switchboard would power four 100A normabarres, each of which can support up to seven racks of 40 PCs. Switchboards that would support six 100A could be available from July.

Anne’s costings were worked on the basis of support for a CPU farm for a single LHC experiment. It was agreed that a proposal for equipping the vault with power outlets at a reasonable granularity should be prepared. A very preliminary review suggested that three switchboards would be required to support the physics load with further switchboards for the basic computing infrastructure and for the air conditioning units.

5. AOB

It was noted that the B513 goods lift is scheduled for renovation in September/October.