Let’s get physical
We started by rethinking the physical infrastructure as a whole. What if we designed all the hardware needed to build a cloud – regardless of whether it was bringing compute, storage or networking to the party to share a bunch of common design elements? One platform on which everything could efficiently be built. That would hugely simplify spares handling and lifecycle management of upgrades and the like.
But then, what if we could design an architecture that made these hardware components consumable? Not in a throwaway, single use kind of way, but in the ability to add more of them to the fabric, change them out, or remove them entirely without the rest of the system caring, impacting the end users, and without the engineer installing them ever needing to configure any of them, or not even much care what they were?
Want to add compute or storage resources? Simply plug them in and power them on. Within a few minutes, the resources are available to the entire cluster. Need to decommission a node? A single command migrates all resources from that node and powers it off for physical removal.
A true hardware fleet of everything needed to build a cloud in a data centre – not a selection box of flavours.
That would be transformational, so that’s what we did.
Networking is dumb
In the old world, data centre networking was a tax. Largely unchanged in decades, except for how fast it can shift packets, yet with an ever increasing number of crucial but complex and temperamental parameters to configure, the plumbing of the data centre has become the single biggest headache for IT teams. Ask them. Nine times out of ten, it will be the thing that screws up, adds, moves and changes and impacts service availability and delays deployments.
HyperCloud handles networking configuration changes in a unified way; by integrating the networking into our Interconnects, changes to the virtual network layer automatically propagate to the physical networking layer. As changes are made, whether by the API, CLI, or GUI, those changes instantly take effect where needed physically. No more trunking VLANs to compute nodes and upstream switch ports.
HyperCloud even has IP Address Management functionality built-in, allowing IP address assignment to guest workloads automatically without requiring any guest VM or container configuration.
The network becomes the heart of the fabric.