Could you add a visualization tools, to be able see which blocks of a volume are in cache (L1 & L2), like defragmenter tools block visualize?
And if possible, the files in block.
[2017-05-27] PrimoCache 3.0.0 Beta is available now!
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
@manus, thank you for your suggestion. This feature is already in our task list, however, I'm sorry that it may not come out soon.
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
I decided to give it another go and well it works great but still has a long-standing bug that makes it unusable on the main system I want to use it on. There seems to be an incompatibility with my onboard Killer E2200 NIC after a couple of reboots the NIC drops until you uninstall PrimoCach.
To reproduce simply install PrimoCach reboot, setup the cache, reboot and possibly reboot one or two more times and the NIC won't work. Device Manager will say there is some sort of hardware problem. Uninstalling the NIC and reinstalling it won't fix it (Code 10) trying the minimal INF driver also won't fix it, but uninstalling PrimoCach will.
If I put something like an Intel PCI-E NIC in the system it will work fine but I normally have some other cards in the system and I don't have enough PCI-E lanes on this CPU so it kills my GPU performance.
System specs: http://speccy.piriform.com/results/f2sC ... WBQIXh5oCm
Let me know if you need me to try anything.
Also looks like I need to fix my ram frequency but that's irrelevant...
To reproduce simply install PrimoCach reboot, setup the cache, reboot and possibly reboot one or two more times and the NIC won't work. Device Manager will say there is some sort of hardware problem. Uninstalling the NIC and reinstalling it won't fix it (Code 10) trying the minimal INF driver also won't fix it, but uninstalling PrimoCach will.
If I put something like an Intel PCI-E NIC in the system it will work fine but I normally have some other cards in the system and I don't have enough PCI-E lanes on this CPU so it kills my GPU performance.
System specs: http://speccy.piriform.com/results/f2sC ... WBQIXh5oCm
Let me know if you need me to try anything.
Also looks like I need to fix my ram frequency but that's irrelevant...
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
I have a quick question for the L2 write cache, how often/quick is the write cache flushed to the store the cache is backing?
The reason I ask is, I have been doing some work where I use VMware workstation to build and test system settings, this usually involves snap-shoting a base known config, make changes, test, revert to base snapshot and rinse and repeat.
I've noticed that every now and again that my vm's, especially after performing an os software update operation and reboot, either don't boot, or, files are "corrupted", in this morning's case, it is the yum repository cache, right after a reboot, performing a yum check-update on the vm reports invalid cache content. A seperate example from earlier this week was, performing an update, where sshd software was updated within the vm, then performing a reboot, it appears that the /var/empty/sshd directory goes missing and needs to be recreated. I've just tried the same updates again this morning on a store not backed by primocache and I don't get these types of issues.
For the store that has primocache backing it, I am running version 3.0.0 beta under eval mode, the l2 cache is split into a read/write cache (95% (430GB) read 5% (20GB) write). Defer-Write is disabled, the L1 cache is an 8GB read cache.
Is anyone else seeing something similar, or, is there something else going on that doesn't involve primocache at all.
Thoughts greatly appreciated.
The reason I ask is, I have been doing some work where I use VMware workstation to build and test system settings, this usually involves snap-shoting a base known config, make changes, test, revert to base snapshot and rinse and repeat.
I've noticed that every now and again that my vm's, especially after performing an os software update operation and reboot, either don't boot, or, files are "corrupted", in this morning's case, it is the yum repository cache, right after a reboot, performing a yum check-update on the vm reports invalid cache content. A seperate example from earlier this week was, performing an update, where sshd software was updated within the vm, then performing a reboot, it appears that the /var/empty/sshd directory goes missing and needs to be recreated. I've just tried the same updates again this morning on a store not backed by primocache and I don't get these types of issues.
For the store that has primocache backing it, I am running version 3.0.0 beta under eval mode, the l2 cache is split into a read/write cache (95% (430GB) read 5% (20GB) write). Defer-Write is disabled, the L1 cache is an 8GB read cache.
Is anyone else seeing something similar, or, is there something else going on that doesn't involve primocache at all.
Thoughts greatly appreciated.
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
@RED404, sorry for the late reply!
This problem is introduced since Windows 10 build 14393 (1607). It seems that Microsoft changed the memory allocation method, causing the memory conflict between PrimoCache and E2200 NIC. We made efforts to avoid such conflict and the version 2.7.0 is workable with most network adapter. But for MSI X99 series with Killer E2200 NIC, we don't know why the problem still exists. Even PrimoCache only allocates 128MB memory for L1 cache, it still will fail E2200 NIC. It seems that E2200 driver requires certain memory address which has been occupied by PrimoCache. But all memory allocations are controlled by Windows, and we don't know the real reason that E2200 fails. Perhaps we may need to contact Killer support.
I'll keep you updated if we have some news on this problem. Thanks.
This problem is introduced since Windows 10 build 14393 (1607). It seems that Microsoft changed the memory allocation method, causing the memory conflict between PrimoCache and E2200 NIC. We made efforts to avoid such conflict and the version 2.7.0 is workable with most network adapter. But for MSI X99 series with Killer E2200 NIC, we don't know why the problem still exists. Even PrimoCache only allocates 128MB memory for L1 cache, it still will fail E2200 NIC. It seems that E2200 driver requires certain memory address which has been occupied by PrimoCache. But all memory allocations are controlled by Windows, and we don't know the real reason that E2200 fails. Perhaps we may need to contact Killer support.
I'll keep you updated if we have some news on this problem. Thanks.
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
@Asgaroth, sorry for the late reply!
When defer-write is disabled, all write-data are immediately written to underlying disks. So in your case, I don't think you need L2 write cache. Could you check if the problem still exists when you change the L2 cache to 100% read?
When defer-write is disabled, all write-data are immediately written to underlying disks. So in your case, I don't think you need L2 write cache. Could you check if the problem still exists when you change the L2 cache to 100% read?
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
Hello,
Is PrimoCache 3 close to final release? I am looking forward to trying it out but I would prefer not to use a beta version at this time.
Regards,
Eric
Is PrimoCache 3 close to final release? I am looking forward to trying it out but I would prefer not to use a beta version at this time.
Regards,
Eric
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
I am curious to know too. The 3.0 L1 has been working great for me. How about an update?
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
Patiently waiting for the full 3.0. Looking forward to it.
Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!
Will the server version 3.0 be released at the same time?