- Production
PhotonSpark infrastructure
Compute, shared storage and network operations across a 15+ server fleet
I designed and operate the shared infrastructure behind PhotonSpark’s services: two Proxmox VE clusters backed by Ceph, a fleet of more than 15 servers, and services running on the company’s own IP space. I support AS216249 alongside the company’s network engineer.
- Proxmox VE
- Ceph
- Linux
- eBGP
- GRE
- Firewalls
- 25
- 256 TiB
- 2
- 15+
Components
Shared storage
- Three-way replication; 256 TiB is raw capacity, before replication overhead.
- RBD for virtual-machine disks and CephFS for shared files.
Compute fleet
- Dell R720 and HPE DL360 Gen8/Gen10 servers.
- i9-13900KS GPU nodes for inference workloads.
Network responsibilities
- Support operation of AS216249 alongside the company’s network engineer.
- A company-owned /24 IPv4 block announced over eBGP to upstream transit.
- Standby DDoS mitigation sessions and scripted GRE tunnels for resilience.