-4
There are huge amount (thousand of thousands) of storages in big data centers. So eveday some number of them are wear out and even sometimes burn out. Are there any software or intellectual systems (neural networks?) that can help to predict and prevent such situations to replace "future-failed" storages in advance?
Kenenbek Arzymatov
Posted 2017-11-04T15:37:24.283
Reputation: 103
1
If you're refering to Linux, then smartctl. However SSD disks may be connected with HW RAID controllers, then you have to use utilities which can work with such a controller (usually provided by the HW vendor).
It's also common to use spare disks, so in case some disk goes down, it's replaced with one of spare disks.
The SNMP alarm generation is usually also possible.
Jaroslav Kucera
Posted 2017-11-04T15:37:24.283
Reputation: 1 352
This is a comment—not an answer—to a question that is really low quality. – JakeGould – 2017-11-04T16:16:58.333
2All data center storage is RAIDed so one drive failure does not mean the whole cluster fails. Additionally, SMART alerts can be setup on these systems to monitor drive health and let one know when a drive might be coming close to failure so systems engineers can step in, swap one out and replace it. – JakeGould – 2017-11-04T16:16:36.350
@JakeGould , question about prevention single drive failure and intellectual system for identifying such problems in advance. Do you have links for SMART alerts, how do they work? – Kenenbek Arzymatov – 2017-11-04T16:29:23.437
1“Do you have links for SMART alerts, how do they work?” You really need to do a basic Google search for SMART functionality and RAID setup. Pretty much any RAID device out of the box will have some kind of system to alert you when drives fail. This is not deep knowledge. This is quiet common. – JakeGould – 2017-11-04T16:44:58.807