80% Rule with all Flash for Surveillance Video

[Fratopolis]

Question?

Ok I am using three Mini XL+ outfitted with 10 5210 ION 7.68TB drives each for surveillance. So far I am following the 80% rule.

I am trying to test filling to 95%. So I created another vol with the space that hits the 95% mark and filled the vol with junk data. and going to let the system run for couple months. That is about how long it will fit video for 80%.

If I see degradation I can simply delete the Vol using 15% for this test. Will there be any repercussions to testing this. Obviously fragmentation of current data but that will all be replaced over time. Or should this just be all fine as my use case is 2 month old data is deleted and replaced with new data consistently.

also if I am not mistaken these drives have reserved space not shown to the OS for this issue alone.

 

[jgreco]

also if I am not mistaken these drives have reserved space not shown to the OS for this issue alone.

You're referring to the free page pool. I recommend a read-over of the 2013-era Ars Technica article on how SSD's work if you're not clear on the mechanics here. If you have a steady ingest rate within the ability of the drive to reorganize and keep up, then you might be okay-ish. The 80% rule is actually part of a performance loss mitigation strategy that is aimed more at HDD's, but also has some validity for SSD's. The dynamics are complicated and very dependent on case-specific factors.

I am trying to test filling to 95%. So I created another vol with the space that hits the 95% mark and filled the vol with junk data. and going to let the system run for couple months. That is about how long it will fit video for 80%.

If you run this for a year or three, you will actually end up with the answer to your question about how well this works. In the end, it is very difficult to make accurate predictions or to try to artificially "benchmark" these things.

 

[Fratopolis]

If you run this for a year or three, you will actually end up with the answer to your question about how well this works. In the end, it is very difficult to make accurate predictions or to try to artificially "benchmark" these things.

First off thanks for the response. And I have the https://arstechnica.com/information...-revolution-how-solid-state-disks-really-work bookmarked for a read. I think I may have read this before but another look never hurts.

Would this be and artificial benchmark though. I will actually have 95% filled and letting the system run through a complete cycle of video replacement. Or maybe I should let it run through 2-3 cycles. 4-6months

I have three of these systems and I am running the test only on one. So I will be able to see the differences from one to another.
I guess I can keep this thread updated with my results as I test iops/load/response time/etc...

 

[jgreco]

Or maybe I should let it run through 2-3 cycles. 4-6months

You probably need at least a dozen cycles, probably a good bit more. I wasn't kidding when I said a year or three. There will be a gradual degradation as cycles increase and fragmentation increases. This isn't a "one or two" cycles type of issue. It is the insidious slow degradation of performance that gets people.

 

[Fratopolis]

Thanks for the info. I'll bookmark this page and write back my findings. We are only required to keep 1 month so if it becomes bad ill move the cameras to other systems wipe and try 90% lol.