Next Version? Topic is solved

FAQ, getting help, user experience about FancyCache
User avatar
Support
Support Team
Support Team
Posts: 3623
Joined: Sun Dec 21, 2008 2:42 am

Re: Next Version?

Post by Support »

@twu2,

We have checked the dump file and found a very strange error that caused the bluescreen. May I have your disk/volume information, memory installed, and cache configurations?

According to the dump file, this strange error happened with Defer-Write. Could you try cache without Defer-Write enabled?

Thank you.
twu2 wrote:here is my minidump related to PrimoCache.
It happen when I create a cache pool for 15G size.
User avatar
Support
Support Team
Support Team
Posts: 3623
Joined: Sun Dec 21, 2008 2:42 am

Re: Next Version?

Post by Support »

@manus,

L2 won't support Defer-Write. So it is not very useful to split the strategy between L1 and L2.
manus wrote:@support do you think that you can split the strategy between L1 and L2?
To be able have Read (or Read/Write but no deferred between L1 and L2) on L1 and Read/Write on L2 with deferred.
manus
Level 4
Level 4
Posts: 28
Joined: Fri Nov 18, 2011 6:03 pm

Re: Next Version?

Post by manus »

So all write between L1, L2 and HDD are sync...
Anyway to implement defer on L2?
Because if you do it, we can have secure READ&WRITE in L1 synced with L2 and defer on HDD, so if we get a power lost L2 have the last data. And with that we can have better write performance, because we use L2 full speed.
minhgi
Level 10
Level 10
Posts: 255
Joined: Tue May 17, 2011 3:52 pm

Re: Next Version?

Post by minhgi »

support wrote:@minhgi,

Program will check the volume size to see if it is supported or not. Actually program supports up to 16TB volume size with 4KB cache block size. With larger block size, program can support a volume larger than 16TB.

We have analyzed your dump file and known what part of code caused the crash. But the root cause hasn't been known yet. We'll keep you posted as soon as we have any news.

BTW, what block size did you configure? And is the whole raid5 disk partitoned into one volume?
minhgi wrote:This now just came to my mind. What is the space limit that Primocache can handle? My Raid 5 is 5.7TB. That 4 2TB hard drive in Raid5. Would that be issue?
The Raid5 use 4K formatted NTFS and use a 64K block sector for the raid controller ( which is default and optimize for my file types use).
minhgi
Level 10
Level 10
Posts: 255
Joined: Tue May 17, 2011 3:52 pm

Re: Next Version?

Post by minhgi »

support wrote:@minhgi,

Were there any other prorams possibly dismounting, mounting, or formatting this raid5 volume, for eg, backup software? We find that when the crash happened, Windows was mounting this volume. This seems a little wierd...
The only programs that could have mounted is Diskeeper 2012 with auto defrag. But it works fine with other caching software. I also do have Acronis True Image 2013 install as a backup software.

On my Lenovo Thinkpad, I have these programs install Diskeeper, Acronis, and Primocache . Both L1 and L2 enable without any problem.

I think I have found out way my computer is crashing on my Raid 5. Primocache driver seem like it can not handle large overhead memory consumption. When I lower the cluster size to 8k, the overhead memory requirement is at a staggering 2.7GB of physical memory. Primocache did not crash. If I lower the cluster size to 16k, the overhead is 1.4GB. My guess is when I set the cluster to 4k, it should be at 5.4GB for the overhead and my computer would crash.

If I set my partition to 4TB, the 4k overhead is 3.7GB and not crash. Primocache should have been able to handle such a large overhead without any problem.
Attachments
2013-06-26_154327.png
2013-06-26_154327.png (58.15 KiB) Viewed 75280 times
Last edited by minhgi on Wed Jun 26, 2013 8:58 pm, edited 2 times in total.
Cachemaniak
Level 2
Level 2
Posts: 6
Joined: Wed Jun 26, 2013 5:06 pm

Re: Next Version?

Post by Cachemaniak »

I just discovered this software and replaced Express Cache with it. Express Cache was allways wiped out anytime my OS wasn't shut down properly (i.e. 4s power down button).
I've got some questions about Primocache:
1. How can I check how much disk space is used for the cache out of my whole cache partition ?
2. Is the persistent cache cleared with dirty OS shutdown (i.e. blue screen) ?
3. Can I use Primocache to shut down slow hdd to conserve battery and only sync the cache every 4 hours ? This way I could spin down my hdd an work on battery more time.
Incriminated
Level 6
Level 6
Posts: 65
Joined: Fri May 31, 2013 3:03 pm

Re: Next Version?

Post by Incriminated »

support wrote:
Incriminated wrote:And i have another little question: When PrimoCache starts it searches for updates. Is this actually working in the beta-stadium or is it just for later use. Will I get i.e. 0.9.1 automatically, or do I have to scan the forum-threads?
Automatic update check works in the beta version.
Incriminated wrote:I think the option "Release After Write" and "Overcome HDD C1 Issue" did make a serious difference in performance in fancycache versions depending on the scenario.
Well, "Overcome HDD C1 Issue" is not designed to improve the performance. Regarding with "Release After Write", did you really see the great performance boost with this option?
Let's say i have 2GiB R/W-cache. Then I write 1GiB on that disk. With Release after Write after that this amount is beeing freed instantly. Now lets say i want to write a second file that also has 1GiB. This would mean when release after write is not enabled that it keeps the first 1GiB in cache and free's the data that is stored in the read-cache to make place for the second 1GiB write. So when i next try to open a file that was cached before, it needs to be read from disk once again.

This is hard to measure, but i definetly think what I assume by that is correct and that "Release after write" can have a serious impact dpenending on strategy and cache-size. Correct me when Im wrong.

And please tell me why you not want me - as an advanced user - to be free to choose to enable/disable such things by my own... i already heard that you want to simplify, i have no prblem with an optional simple/advanced settings-structure. But one other way would be to simply provide a meaningful manual and not some wish-wash like here: http://www.romexsoftware.com/en-us/fanc ... /help.html <- Nothing here is of help, not a single option is properly explained so I get a meaning of what to choose. i.e LFU/LRU ... it is of practically interest what to choose in which scenerio for benefit, not what it technically does and force the costumer to study IT to get an idea of what he wants in his particular use-case! An easy explanation is more worth than a simplification without a note.

Counterwise I would like to ask you, as the support/developers: Did you NOT see any great performance boost from this feature.

In the end, i don't know if the files I write now get released after write or if they will be kept until all read-cache is flushed and cache is full.
Last thing is totally not improving my performance because i nearly NEVER access the files i just write right after.

I want to use lowest possible amount of the L1 to be used for write... so NOT to flush out any read-data if avoidable.
I simply do want to tell the program to "Release (the data) after write".

Whats wrong with my demand?

PS: Im sry that i have to correct you with your own "help-file". Overcome HDD C1 Issue: C1 value on some mechanical hard disks, especially laptop disks, may increase quickly when Defer-Write is enabled. This option is designed to avoid such issue. SSD users can disable this option for better performance. -> So obviously YES this option is described/designed to improve performance on SSD, so Im asking why it is removed and how PrimoCache now make sure SSD doesn't run into performance problems while this option is gone without a note.

You should try to inform your costumers about the functions of your programm instead of start simplifying at all cost.

I personally think even more possibilities for advanced settings are welcome. This programm beeing complex is natural, and I very much respect it and would buy it especially because it is highly configurable. Still there's no problem with a simplyified "cache-setup-assistent" or other way around "advanced-options-tab". Making it suitable for both - simple and advanced users - is not - I repeat: IS NOT - in opposition to each other.
turkina
Level 3
Level 3
Posts: 12
Joined: Thu Sep 27, 2012 9:56 pm

Re: Next Version?

Post by turkina »

minhgi wrote:I think I have found out way my computer is crashing on my Raid 5. Primocache driver seem like it can not handle large overhead memory consumption. When I lower the cluster size to 8k, the overhead memory requirement is at a staggering 2.7GB of physical memory. Primocache did not crash. If I lower the cluster size to 16k, the overhead is 1.4GB. My guess is when I set the cluster to 4k, it should be at 5.4GB for the overhead and my computer would crash.

If I set my partition to 4TB, the 4k overhead is 3.7GB and not crash. Primocache should have been able to handle such a large overhead without any problem.
My Primocache currently has 7.81GB of overhead and hasn't crashed since I set it up on release day.
minhgi
Level 10
Level 10
Posts: 255
Joined: Tue May 17, 2011 3:52 pm

Re: Next Version?

Post by minhgi »

turkina wrote:
minhgi wrote:I think I have found out way my computer is crashing on my Raid 5. Primocache driver seem like it can not handle large overhead memory consumption. When I lower the cluster size to 8k, the overhead memory requirement is at a staggering 2.7GB of physical memory. Primocache did not crash. If I lower the cluster size to 16k, the overhead is 1.4GB. My guess is when I set the cluster to 4k, it should be at 5.4GB for the overhead and my computer would crash.

If I set my partition to 4TB, the 4k overhead is 3.7GB and not crash. Primocache should have been able to handle such a large overhead without any problem.
My Primocache currently has 7.81GB of overhead and hasn't crashed since I set it up on release day.
Am not sure why it crash, but if I set my Raid5 to 4k it crash. All other cluster size works for me. Why is your overhead so large?
User avatar
Support
Support Team
Support Team
Posts: 3623
Joined: Sun Dec 21, 2008 2:42 am

Re: Next Version?

Post by Support »

@minhgi,
Thank you very much! You're right. There is an overflow that will cause the crash if the partition to be cached is too big. I'm sorry that I made a mistake on the supported maximum partition size. PrimoCache aslo forgot to do a full parameter check before starting cache. This issue will be fixed in the next version. Here is the list of supported maximum partition size.

Cache Block Size Maximum Partition Size
4KB 4TB
8KB 8TB
.... ...
256KB 256TB

@turkina,
Do multiple partitions share one cache in your scenario? And each partition is less than 4TB? If there's a partition larger than 4TB, it is better to change the block size to a larger value.
Post Reply