Dell quotes Dell. HPE quotes HPE. Who checks both?
Windows Server 2016 leaves Microsoft support on January 12, 2027,¹³ and server CPU and board lead times run near a year.¹⁴ We carry Dell and HPE, with Cisco, Arista, NetApp, Nutanix, Veeam and others, and we design, supply, integrate, deploy and support what you choose. Whoever recommends it delivers it, and if one maker's date slips we can offer another. We earn when you buy through us. One room or a whole campus.
Design: 2 hours to recover¹
Design: at most 15 minutes lost²
Design: one spare server per site³
Test plan: 20 tests with pass marks⁴
Three power feeds. One fire. Airport closed.
- The cases on this page are public record, not our clients. Heathrow airport has three separate power feeds. On the night of March 20, 2025, a fire at a nearby substation cut one. Part of the airport lost power, including critical systems, and Heathrow closed. Its backup plan relied on manual switching, an estimated 10 to 12 hours.⁵
- The grid, generators and building are yours or your facility partner's. We write into your requirement: what each rack draws, from the configured spec sheets, what carries it when the grid goes, how fast it switches over, and what shuts down first.
- Our published design starts with a survey of both sites: rack space, power per rack, cooling, uplinks.⁶
- We check: every option against that line, with a source under every figure.
Two data centers. One heatwave.
The cooling plant is yours or your facility partner's. We give them the load per rack of what we supply.
In a July 2022 heatwave that reached 40°C, these London hospitals' two data centers, under two miles apart and each the other's backup, both failed.⁷ Each site must be able to carry the whole estate.
In a July 2022 heatwave that reached 40°C, these London hospitals' two data centers, under two miles apart and each the other's backup, both failed.⁷ Each site must be able to carry the whole estate. Guy's and St Thomas' hospitals
Twenty minutes to critical heat.
Proton, the encrypted email company, says denser servers cut the time to critical heat from 3 to 4 hours to 20 minutes.⁸ So power and cooling limits per rack go in writing, from the configured spec sheets.
We supply and install the racks, A and B power feeds, top-of-rack and out-of-band switches. Elevations drawn before staging, every rack in the as-built.⁹
AMD or Intel? Dell or HPE?
We supply both. Dell PowerEdge R7725 with AMD EPYC 9005, up to 6 TB of memory. HPE ProLiant DL380 Gen12 with Intel Xeon 6, up to 8 TB.¹⁰ Staged and burned in before install, with the end-of-support date of every item on the quote.¹¹
Proton says a server shortage meant its equipment, if lost, could not be replaced on short timelines.⁸ So one spare server per site.³ Our test plan: a server stops under load, and its machines restart elsewhere within 5 minutes.⁴ Which platform runs on top? Our six-platform comparison.
Ask the architect who wrote it.
Go through our published two-site design and its 20 tests with the architect who wrote them.
Start with one purchase.
Send what you need to buy and by when. You get a reply the same day, then a quote.
Led by Steve Phillips
Steve Phillips, Principal Data Center Architect · CCSP · CCIE · AWS Certified Security – Specialty — credential IDs on request
He wrote the two-site private-cloud design: two sites 40 km apart, each able to carry the whole estate, on hyperconverged servers or VMware Cloud Foundation.
- Two sites, 40 km apart
- Sized for 250 virtual machines, 400 by 2029
- One spare server per site
- Recovery plans tested twice a year
- 20 tests with pass marks
Two data centers. Tested in one.
UK bank TSB planned to test digital banking across both data centers for its 2018 migration, then tested on one. After go-live, configuration errors in the two data centers failed customer sessions. In 2022 its regulator fined TSB, for wider failings in risk management and governance.¹²
We write into your requirement: test both sites, in the setup you will run. Our test plan: a whole site is lost, and everything protected runs at the other, then fails back.⁴
Network.
Cisco, Arista, Juniper and others, supplied and installed. Our design: two top-of-rack switches at 100G per site. Power one off: no impact.⁴
Storage.
NetApp, Dell, HPE, Nutanix and others, supplied and installed. Pull a drive under load: no impact.⁴
Bring one purchase.
Send what you need to buy and by when. You get a reply the same day, then a quote with the end-of-support date of every item.¹¹
