Crispy Branzino on Nostr: If you encounter a linux server with a hardware RAID os partition this is a very old ...
If you encounter a linux server with a hardware RAID os partition this is a very old paradigm.
In the 90s servers were outrageously expensive so vendors built in redundancy at the hardware level. Multiple RAID controllers, multidisk data, redundant ram controllers, pretty wild stuff. HP's Superdome setup even had a commercial where a marksman shot it with a high power rifle and it still ran.
At the time hardware raid was the only game in town so you just did that. Bitrot was even detectable by enterprise controllers.
In 2025 servers are as cheap as $400 on eBay and instead of monolithic sun or SGI hardware everything is a distributed commodity architecture. The entire OS can be redeployed in 15 minutes, backups are either cloud or disk to disk, and multipath raid controller setups are losing ground to stuff like ceph and drbd.
Open source means most of the special sauce stuff you'd spend weeks rehydrating from lto1 with legacy proprietary vendors can just be redeployed from a pipeline.
In the 90s servers were outrageously expensive so vendors built in redundancy at the hardware level. Multiple RAID controllers, multidisk data, redundant ram controllers, pretty wild stuff. HP's Superdome setup even had a commercial where a marksman shot it with a high power rifle and it still ran.
At the time hardware raid was the only game in town so you just did that. Bitrot was even detectable by enterprise controllers.
In 2025 servers are as cheap as $400 on eBay and instead of monolithic sun or SGI hardware everything is a distributed commodity architecture. The entire OS can be redeployed in 15 minutes, backups are either cloud or disk to disk, and multipath raid controller setups are losing ground to stuff like ceph and drbd.
Open source means most of the special sauce stuff you'd spend weeks rehydrating from lto1 with legacy proprietary vendors can just be redeployed from a pipeline.