Firefox Creating a profile off of my SSD

[HaveaMint]

I never knew this existed. If you are the only one using your PC it is worth while to check this out. Firefox writes so much to the drive that it cannot be good for the longevity of an SSD.
https://support.mozilla.org/en-US/kb/pr ... x-profiles

[AZgl1800]

Modern SSDs are almost as good as a HDD, their lifetimes are measured in millions of cycles now.
I personally am not concerned about this.

I have known for years, that FF profiles are stored somewhere besides in the Mozilla directories, even in windows that is true. You can completely deleted Firefox, all vestiges of it, and a reinstall will come up with your profile just like it was before. I just never knew how to find that folder.

I have a crontab job that runs fstrim daily, I don't think there is anything to be worried about on this.

[HaveaMint]

If you use cloud or whatever FF calls it then yes a delete/reinstall will come back as it was. I don't use that so if I delete the .mozilla all is gone. I also keep a backup. What that link does is show you how to create a profile that you can set to a separate partition or drive and all that is stored on the home partition is a few files. SSD's are far better than they used to be but any prolonging of life isn't a bad thing. Saying your not worried about it then why waste 10-15% creating dead space to preserve the life of a SSD? Platter drives last a long time, I still have a 420mb drive and for the life of me don't know why I still have it.

[AZgl1800]

I use Firefox Sync, so I doubt that Profile exists on my PCs.

[HaveaMint]

I use Firefox Sync, so I doubt that Profile exists on my PCs.

It does, all sync does is keep the backup you have. Check your home directory .Mozilla
https://www.servethehome.com/firefox-is ... to-fix-it/
I prefer to go one step further and have it do its thing on a platter drive.

[HaveaMint]

I just created a new profile on another drive a couple of hours ago and this is the cache directory size now.

Selection_003.jpg

[Portreve]

Hey HaveaMint and AZgl1500...

I never knew this existed. If you are the only one using your PC it is worth while to check this out. Firefox writes so much to the drive that it cannot be good for the longevity of an SSD.

Modern SSDs are almost as good as a HDD, their lifetimes are measured in millions of cycles now.
I personally am not concerned about this.

Have either of you looked at this bit from this guide?

[catweazel]

Have either of you looked at this bit from this guide?

There are better ways of limiting the writes without kludging FF and potentially wrecking people's browsing experience. A persistent RAMdisk is one.

That's a comment about Pjotr's page, not about your suggestion to go there ;>

[HaveaMint]

Hey HaveaMint and AZgl1500...

I never knew this existed. If you are the only one using your PC it is worth while to check this out. Firefox writes so much to the drive that it cannot be good for the longevity of an SSD.

Modern SSDs are almost as good as a HDD, their lifetimes are measured in millions of cycles now.
I personally am not concerned about this.

Have either of you looked at this bit from this guide?

I considered limiting cache until I ran across moving the profile. I figured limiting cache would be a small sacrifice in speed of the browser and so would moving the profile from the ssd to a spinner, so it equals out kinda there but the plus of getting it off of the ssd is worth it to me.

[HaveaMint]

It can't hurt.

[catweazel]

I considered limiting cache until I ran across moving the profile. I figured limiting cache would be a small sacrifice in speed of the browser and so would moving the profile from the ssd to a spinner, so it equals out kinda there but the plus of getting it off of the ssd is worth it to me.

Moving the profile won't affect the cache one bit.

[HaveaMint]

I considered limiting cache until I ran across moving the profile. I figured limiting cache would be a small sacrifice in speed of the browser and so would moving the profile from the ssd to a spinner, so it equals out kinda there but the plus of getting it off of the ssd is worth it to me.

Moving the profile won't affect the cache one bit.

Correct, but now the cache is on a spinner and not the ssd.

[rui no onna]

Honestly, I don't understand the logic of buying a pricey SSD in order to get a performance boost and then putting a lot of the stuff that would benefit from lower latency back on a spinner.

Mind, I do have disk cache disabled - RAM is still faster than SSD.

[catweazel]

Honestly, I don't understand the logic of buying a pricey SSD in order to get a performance boost and then putting a lot of the stuff that would benefit from lower latency back on a spinner.

HDDs are optimised for serial read. Reading a cache is a serial read. If people want to move the cache onto a HDD to save their SSD life then that's their prerogative.

[catweazel]

RAM is still faster than SSD.

Not in all cases. I run dual 3.2GB/s NVMe drives in RAID 0 for 6.4GB/s throughput. When you do that, even 3GHz RAM can become a bottleneck.

[rui no onna]

HDDs are optimised for serial read. Reading a cache is a serial read.

With small javascript, jpegs, etc, wouldn't something like a browser cache be predominantly small block random rather than sequential?

Not in all cases. I run dual 3.2GB/s NVMe drives in RAID 0 for 6.4GB/s throughput. When you do that, even 3GHz RAM can become a bottleneck.

Shouldn't 3GHz DDR4 give you around 25.6GB/s of bandwidth theoretically?

[catweazel]

HDDs are optimised for serial read. Reading a cache is a serial read.

With small javascript, jpegs, etc, wouldn't something like a browser cache be predominantly small block random rather than sequential?

Code: Select all

$ ls /mnt/ramdisk/cache/mozilla/firefox/5k84tgp3.default/cache2/entries/
00068CB1E68CB5EE3009C8806EE65A56FBE8CC08
000A63126345050D2C00849EFB45B17743CA7DFF
000C3DAD5A5E6E3DFA97D61B7B90223EDB7428B3
001686FC87FBCC7941C55519FF2569015FCAB204
001907B92F131CDD1C46405EE40CDC07E5F165AE
...
FFEE7AA1F2B7C98EC8688B7304A9E179A56BDDB2
FFEF019ABB7AC4BBB5C20749342BA83226E3CBFC
FFF853C4A0FE6394EA98983034915696AE1814AD
FFFB29AA663D1ECB6E4F07FD59FAE331F373C4AE

No. The 'random' you refer to is the location of files on disk, which are stored in the directory table. Each file is read sequentially. Secondly, the latency in an SSD for the lookup operation in the directory table then moving to the file is in the order of a handful of microseconds, depending on the SSD. Moving from the end of one file, back to the directory table then on to the start of another file is not classed as random reading.

https://flashdba.com/2013/04/15/underst ... equential/

Not in all cases. I run dual 3.2GB/s NVMe drives in RAID 0 for 6.4GB/s throughput. When you do that, even 3GHz RAM can become a bottleneck.

Shouldn't 3GHz DDR4 give you around 25.6GB/s of bandwidth theoretically?

Yes. My error. I misplaced the decimal point in my noggin.

[rui no onna]

No. The 'random' you refer to is the location of files on disk, which are stored in the directory table. Each file is read sequentially. Secondly, the latency in an SSD for the lookup operation in the directory table then moving to the file is in the order of a handful of microseconds, depending on the SSD. Moving from the end of one file, back to the directory table then on to the start of another file is not classed as random reading.

"Random" being just as fast as sequential is just a characteristic of SSDs, no? If you put the files on HDD, you can't expect to have the same access times. Particularly not if you're running multiple programs that are simultaneously hitting the cache drive hard.

An SSD is more likely to die from random failure than P/E cycle wear out. Heck, an HDD is more likely to die from mechanical wear than an SSD from media wear out.

Mind, I just checked one of my Windows 7 SSD (Samsung 840 500GB, planar TLC). At 1233 days of power-on hours, it's used up 55 of its 1000 rated P/E cycles. At that rate, it'll take 50+ years to use up the P/E cycles. I'm sure it'll die of other causes way before then. Only optimization is ~15GB over-provisioning (for consistent performance rather than to improve SSD life).

[catweazel]

No. The 'random' you refer to is the location of files on disk, which are stored in the directory table. Each file is read sequentially. Secondly, the latency in an SSD for the lookup operation in the directory table then moving to the file is in the order of a handful of microseconds, depending on the SSD. Moving from the end of one file, back to the directory table then on to the start of another file is not classed as random reading.

"Random" being just as fast as sequential is just a characteristic of SSDs, no?

No. SSDs still have latency.

[AZgl1800]

Hey HaveaMint and AZgl1500...

I never knew this existed. If you are the only one using your PC it is worth while to check this out. Firefox writes so much to the drive that it cannot be good for the longevity of an SSD.

Modern SSDs are almost as good as a HDD, their lifetimes are measured in millions of cycles now.
I personally am not concerned about this.

Have either of you looked at this bit from this guide?

yup, I left 39.2 gBytes for that purpose.

SSDunallocatedSpace.png