From factory switchrooms to fourteen data halls

Mandate

Run the estate within its redundancy and energy commitments while adding capacity in Jiaxing and reducing dependence on the largest customer.

Outcomes

• Opened Jiaxing Building 2 in March 2024; its 2025 PUE was 1.24 at 74 per cent of design load
• Held hall redundancy through the 2022 festival peak; the largest customer's revenue share fell from 31 to 24 per cent by 2025
• Reduced the estate's annual PUE from 1.41 in 2021 to 1.33 in 2025
• Maintained multi-level protection scheme assessments for every hosted platform

Decisions in context

When I became COO, the company had four buildings, a construction programme for a fifth and a customer list in which one e-commerce operator mattered more than any operating policy. My work since has been to make the estate's limits explicit and then sell within them. I own facilities operations, construction and service delivery, and I present the operating report and construction programme at every board meeting. The festival decision in 2022 and the Jiaxing redesign were the most consequential calls of this period, and both cost revenue in the short term. I also attend to the less visible parts of the role: shift rotas, spare parts, generator fuel contracts and metering by hall. I hold no directorship at Haolinjiang. The founders and the board decide strategy; my job is to make sure their decisions rest on what the buildings can actually do.

Reason for the move

Current appointment; seeking one compatible non-executive role with employer consent.

Mandate

Bring every campus under one operating standard and open the first Jiaxing building.

Outcomes

• Moved planned electrical switching into staffed daytime windows after the April 2019 incident
• Brought electrical maintenance in house with eighteen authorised electricians
• Opened Jiaxing Building 1 in August 2020 after full integrated testing despite travel restrictions
• Installed 212 automatic transfer switches for single-corded customer equipment

Decisions in context

This was the period in which I moved from engineering standards to accountability for every operating result. The April 2019 switching incident tested that change quickly. I had written many of the procedures the contractor followed, so I could not treat the error as someone else's failure. The reforms that followed were mine to defend with customers, with the night teams and with the chief executive, who had to accept a permanent cost increase. In 2020 we completed Jiaxing Building 1 while travel restrictions kept some vendor engineers away. We tested with remote vendor support and delayed opening by six weeks rather than accept partial testing. I also began monthly near-miss reviews with duty managers. By the end of the period I spent less time on engineering detail and more on the commercial consequences of operating rules, which prepared me for the wider role.

Reason for the move

Appointed COO when the founders separated the chief executive's commercial role from operations.

Mandate

Write engineering standards for power, cooling, fire suppression and maintenance across a growing estate.

Outcomes

• Introduced integrated systems testing with load banks before any new hall opened
• Opened Suzhou Building 2 in 2016 with five halls
• Wrote a capacity statement for every hall stating its load with redundancy intact
• Replaced the UPS batteries in Suzhou Building 1 without a customer interruption

Decisions in context

I joined from the customer side, and that shaped the first thing I changed. Haolinjiang's halls were sold on design figures, and nobody could tell a customer exactly what a hall carried with one cooling unit failed. I wrote capacity statements for each hall and found two that had been sold beyond what their redundancy supported; we moved three customers, at our own cost, to restore it. I also introduced integrated systems testing before handover, which the construction contractor for Suzhou Building 2 resisted because it lengthened the programme by three weeks. The test found a generator synchronisation fault that would otherwise have appeared during a real grid loss. My sixty-four engineers came from installation, hosting and building-services backgrounds, and much of my time went on making their standards consistent across two campuses.

Reason for the move

Promoted when operations and construction were combined under one executive.

Mandate

Keep the systems behind production planning, traceability and customer orders running across two factories.

Outcomes

• Reduced production stoppages caused by server-room faults from eleven in 2008 to one in 2012
• Built a recovery room in the second factory and tested failover twice a year
• Moved non-production systems to an external colocation provider in 2012

Decisions in context

Ruitaigang made circuit-board assemblies for industrial and automotive customers, and every board carried a traceability record held on servers I was responsible for. When those servers stopped, the lines stopped, and the customers' quality teams asked questions for weeks afterwards. I ran two server rooms and, from 2010, a recovery room in the second factory. Most of the early faults were ordinary: overloaded circuits, air-conditioning units without maintenance, batteries past their life. Fixing them taught me how much of reliability is routine. In 2012 we moved non-production systems to Haolinjiang, then a young colocation provider, and I wrote the service requirements. Seeing a provider from the customer's side, including how it reported incidents, made me want to work on the other side of the contract, which is how the founders came to know me.

Reason for the move

Joined Haolinjiang, by then a supplier to the company, to work on hosting at greater scale.

Mandate

Keep power and cooling running for about 1,100 hosted cabinets on a twelve-hour shift rota.

Outcomes

• Led a shift through two grid losses without a customer interruption
• Introduced a customer call list for alarms affecting hosted equipment
• Recorded the summer 2006 cooling shortfall that halted new sales in one room

Decisions in context

Suyuanxin hosted servers for Suzhou businesses on floors it leased in telecommunications buildings. The work was physical and largely nocturnal: battery checks, air-conditioning filters, generator test runs and long nights watching alarm panels. From 2005 I led a shift of five. The company sold cabinets faster than its cooling could support, and in summer 2006 we moved portable fans between rows to keep temperatures within limits. I recorded the temperatures every hour for six weeks and gave the record to the facilities manager, who used it to stop new sales in one room until additional units were installed. That was the first time I saw an engineering record change a commercial decision. I left because I wanted to understand what our failures meant for the businesses whose servers we hosted.

Reason for the move

Moved to a manufacturer to understand what hosting failures cost a customer.

Mandate

Install, test and document factory substations and distribution systems.

Outcomes

• Tested protection settings on twenty-three substation projects
• Prepared as-built records accepted by factory owners at handover
• Took part in the 1999 commissioning test that found a relay set for the wrong transformer rating

Decisions in context

My first employer installed substations and distribution boards for textile, chemical and machinery factories around Wuxi and Changzhou. I began as an assistant engineer drawing circuits and preparing test sheets, and from 1999 I ran testing on my own projects. The commissioning of a dyeing plant's switchroom that year, when my supervisor refused to energise early and the test found a mis-set relay, taught me the value of a test that a client is impatient to skip. I also learnt earthing, cable sizing and the discipline of an as-built record, which most clients never read until something failed. By 2002 several of our factory customers were installing server rooms, and the people running them knew little about power. I could see a small hosting trade forming in Suzhou that needed engineers who understood electricity better than computers.

Reason for the move

Joined a hosting company as Suzhou's first commercial server rooms began to need power engineers.