Page 2 of 2

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Fri Jan 24, 2014 10:37 pm
by bjoerns
Deferred writes do only work for me if L1 not equals 0, but only for (L1/BlockSize) Blocks. With L1==8MB and Block Size==8kB, DeferredBlocks never exceeds 1024 (but free cache L1 stays always at 7.99MB).

The defered wites to L2 (Changelog #1) seems not to be completely broken, there is some used cache on my system but it does not increase as i expect (urgent writes does not increase while deferred blocks is at max)... L2Storage Write is 180Megs after 1h idletime and ~14GB writes.

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Fri Jan 24, 2014 11:09 pm
by piquadrat
bjoerns wrote:Deferred writes do only work for me if L1 not equals 0, but only for (L1/BlockSize) Blocks. With L1==8MB and Block Size==8kB, DeferredBlocks never exceeds 1024 (but free cache L1 stays always at 7.99MB).

The defered wites to L2 (Changelog #1) seems not to be completely broken, there is some used cache on my system but it does not increase as i expect (urgent writes does not increase while deferred blocks is at max)... L2Storage Write is 180Megs after 1h idletime and ~14GB writes.
This seems to be consistent with my experience. Bear in mind that I'm not using L1 at all. Mostly to isolate the issue. There are some writes to L2 but Hit Rate never exceeds 1%. In theory if some part of data (files/sectors which belong to the installed program on the volume) is cached in L2, every run of this program should increase Hit Rate. Obviously that is not the case. I can run the program 40 times in a raw or put system restarting between to clear windows caches and Hit Rate stays the same.

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Sat Jan 25, 2014 6:17 pm
by n6666661
I am running Windows Server Datacenter 2012 and I can't install v0.9.8 it says my system isn't supported.
I can't find on the site a link to download "server edition".
So if you have a valid url to download server ed...

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Sat Jan 25, 2014 7:44 pm
by minhgi
piquadrat wrote:
bjoerns wrote:Deferred writes do only work for me if L1 not equals 0, but only for (L1/BlockSize) Blocks. With L1==8MB and Block Size==8kB, DeferredBlocks never exceeds 1024 (but free cache L1 stays always at 7.99MB).

The defered wites to L2 (Changelog #1) seems not to be completely broken, there is some used cache on my system but it does not increase as i expect (urgent writes does not increase while deferred blocks is at max)... L2Storage Write is 180Megs after 1h idletime and ~14GB writes.
This seems to be consistent with my experience. Bear in mind that I'm not using L1 at all. Mostly to isolate the issue. There are some writes to L2 but Hit Rate never exceeds 1%. In theory if some part of data (files/sectors which belong to the installed program on the volume) is cached in L2, every run of this program should increase Hit Rate. Obviously that is not the case. I can run the program 40 times in a raw or put system restarting between to clear windows caches and Hit Rate stays the same.
If you use the infinity delay write, it will only write a small amount of data. Use the default 10 seconds should be sufficient.

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Sat Jan 25, 2014 10:37 pm
by InquiringMind
n6666661 wrote:I can't find on the site a link to download "server edition".
So if you have a valid url to download server ed...
The PrimoCache download page has four items, "PrimoCache (Beta)" and "PrimoCache Server Edition (Beta)" being the first two.

If you aren't seeing the second link for some reason, this is its download URL.

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Sun Jan 26, 2014 1:54 pm
by manus
I deployed 0.9.8 last week on (they all got PrimoCache/FancyCache since months):
- 4x Windows 8.1 (3 with SSD cache)
- 5x Windows Server 2012 R2 (1 with SSD cache)
- 30x Windows Server 2008 R2

None of them have bug.
All computer with SSD cache have a much better hit ratio (more than 80%) with the same use and computer are very responsive.
All others with only RAM cache works well as 0.9.2, so depending on use they can get 20% to 90% hit ratio.

Do you have a date for commercial product? this year?
Do you have a price for workstation and for server or a range?

Re: [2014-01-20] PrimoCache beta 0.9.8 released!

Posted: Tue Jan 28, 2014 6:36 am
by piquadrat
minhgi wrote:
piquadrat wrote:
bjoerns wrote:Deferred writes do only work for me if L1 not equals 0, but only for (L1/BlockSize) Blocks. With L1==8MB and Block Size==8kB, DeferredBlocks never exceeds 1024 (but free cache L1 stays always at 7.99MB).

The defered wites to L2 (Changelog #1) seems not to be completely broken, there is some used cache on my system but it does not increase as i expect (urgent writes does not increase while deferred blocks is at max)... L2Storage Write is 180Megs after 1h idletime and ~14GB writes.
This seems to be consistent with my experience. Bear in mind that I'm not using L1 at all. Mostly to isolate the issue. There are some writes to L2 but Hit Rate never exceeds 1%. In theory if some part of data (files/sectors which belong to the installed program on the volume) is cached in L2, every run of this program should increase Hit Rate. Obviously that is not the case. I can run the program 40 times in a raw or put system restarting between to clear windows caches and Hit Rate stays the same.
If you use the infinity delay write, it will only write a small amount of data. Use the default 10 seconds should be sufficient.
I'm not using delay write at all. After picking up "Read (SSD Cache Only)" profile during task creation this option is not set. I suspect that "Enable Defer-Write" is not applicable to L2 at all as L2 is only for read acceleration in this version of PC. L2 does not contribute to writing to the volume, so no data in L2 supposed to be delayed.
I can't get this thing to work. Bad luck I suppose as 0.9.2 worked for me without any trouble. Don't know what causes that.
Majority of opinions about 0.9.8 come from people using L1/L2 on a system drive.
I'm using it not on a system drive and without L1 but even while in L1/L2 mode can't get L2Storage read go distinctly above 0. Hit Rate below 1%. Far from 60-90% people report.
As nobody seems to report this problem I'm alone with it. Bad luck.