Hi, I am new to PrimoCache and don't understand the statistics screen.
As you can see on my attachement it said 'Total read' is lower than read from L2storage.
Only explanation I can imagine, there is such an big overhead, because blocksize between cache and harddisk differ.
Now my second question. How do I find the best trade off between my blocksize and the memory requirement.
My second screenshot shows how much more ram(?) memory I would need to reduce the blocksize from 64k to 4k.
Need statistics Explanation Topic is solved
Need statistics Explanation
- Attachments
-
- PC_statitics_weirdRead.jpg (74.28 KiB) Viewed 7417 times
-
- PC_settings_overhead.jpg (146.8 KiB) Viewed 7417 times
Re: Need statistics Explanation
Did you reset statistics data before?FalkH wrote:As you can see on my attachement it said 'Total read' is lower than read from L2storage.
The memory amount needed is that Level-1 Cache size plus Memory Overhead.FalkH wrote:My second screenshot shows how much more ram(?) memory I would need to reduce the blocksize from 64k to 4k.
Re: Need statistics Explanation
1. No, and it's not the first time I noted this. This time I just made a screenshot.support wrote:Did you reset statistics data before?
The memory amount needed is that Level-1 Cache size plus Memory Overhead.
2. Really? It shows an overhead of 882.20MB while my 1st level has 4GB, another sum result I don't get than.
Just changing the 2nd level cluster size makes an increase of the overhead by 12GB, why would this contain the 1st level suddenly?
Please check my screens again and maybe take a coffee first

My point is, how should I calculate the best settings for my cache?
I have 32GB RAM, an 480GB SSD to cache an 4TB HDD, cluster size 4K.
Since the content of the 2nd level cache seems to grow really slow and an huge 2nd level brings the disadvantage of this huge memory footprint, should I reduce my 2nd level cache size?
Maybe it's an better idea to change the cluster size of my hdd?
Re: Need statistics Explanation
1. Regarding the problem of "Total Read" < "L2Storage Read", we'll check this problem.
2. It also needs lots of additional memory to support level-2 cache. Bigger L2 cache size, more additional (overhead) memory. Also smaller block size, more additional memory. It is possible that you see 12GB memory overhead with 4KB block size. In the future 3.0 version, we have improved the design to reduce memory overhead.
When you see memory overhead is too big, just increase block size.
3. If your drive is very busy and you see the L2 gathering data very slow, you may open cache configuration dialog and set the "Gather Interval" to "FASTEST" in the "Advanced Level-2 Cache options".
2. It also needs lots of additional memory to support level-2 cache. Bigger L2 cache size, more additional (overhead) memory. Also smaller block size, more additional memory. It is possible that you see 12GB memory overhead with 4KB block size. In the future 3.0 version, we have improved the design to reduce memory overhead.
When you see memory overhead is too big, just increase block size.
3. If your drive is very busy and you see the L2 gathering data very slow, you may open cache configuration dialog and set the "Gather Interval" to "FASTEST" in the "Advanced Level-2 Cache options".
Re: Need statistics Explanation
1. Looking forward to hear about your findings
2. I am still in the trial period, if I buy a license for 2.5 will it cover an upgrade to 3.x? Will a license grant lifetime upgrades?
3. Works, but I experienced another weird behavior (explanation below).
I used the win10 taskmanager to monitor how primocache works when I tried your advice 'how to fill the cache'.
1. create 10GB of data with h2testw on the cached harddrive
- first the L1 get filled, after this it starts to write to the hdd, no L2 is used now
- h2testw shows the average transfer speed since start writting (248MB/s, this includes fast L1 transfer + slower hdd)
- taskman shows current writing speed to hdd (155 MB/s, Datenträger 2), L2 (Datenträger 1) is almost not utilized now
2. once all writing is done, h2testw starts reading the data
- taskman and also h2testw show a low read speed, the taskman also shows there is still write access to the harddisk
(probably the L1 still has to be flushed to disk)
- this combined read/write access on hdd slows down everything, the common sequential transfer speed of this is hdd is 100-180MB/s
3. all h2testw activities on hdd are done, now primocache starts to fill the L2 cache
- I put focus on L2 (write 32,9MB/s, Datenträger 1), this is really slow for an sequential write on ssd
- the reported response time for L2 (73405ms, Durchschnittliche Antwortzeit) is also slow, should be less than 0,5ms
Any idea how to optimize this? I guess I will try to reconnect the drives.
Now, 1st controller connects system ssd, 2nd controller connects L2 ssd and hdd.
I wanna try 1st controller connects system ssd and hdd, 2nd controller connects L2.
Do you think this might help?
P.S. I know my test is not a real world test. I just try to understand how the cache should be set to get best performance with my hardware.
2. I am still in the trial period, if I buy a license for 2.5 will it cover an upgrade to 3.x? Will a license grant lifetime upgrades?
3. Works, but I experienced another weird behavior (explanation below).
I used the win10 taskmanager to monitor how primocache works when I tried your advice 'how to fill the cache'.
1. create 10GB of data with h2testw on the cached harddrive
- first the L1 get filled, after this it starts to write to the hdd, no L2 is used now
- h2testw shows the average transfer speed since start writting (248MB/s, this includes fast L1 transfer + slower hdd)
- taskman shows current writing speed to hdd (155 MB/s, Datenträger 2), L2 (Datenträger 1) is almost not utilized now
2. once all writing is done, h2testw starts reading the data
- taskman and also h2testw show a low read speed, the taskman also shows there is still write access to the harddisk
(probably the L1 still has to be flushed to disk)
- this combined read/write access on hdd slows down everything, the common sequential transfer speed of this is hdd is 100-180MB/s
3. all h2testw activities on hdd are done, now primocache starts to fill the L2 cache
- I put focus on L2 (write 32,9MB/s, Datenträger 1), this is really slow for an sequential write on ssd
- the reported response time for L2 (73405ms, Durchschnittliche Antwortzeit) is also slow, should be less than 0,5ms
Any idea how to optimize this? I guess I will try to reconnect the drives.
Now, 1st controller connects system ssd, 2nd controller connects L2 ssd and hdd.
I wanna try 1st controller connects system ssd and hdd, 2nd controller connects L2.
Do you think this might help?
P.S. I know my test is not a real world test. I just try to understand how the cache should be set to get best performance with my hardware.
- Attachments
-
- PC_fillCache_slow.jpg (963.42 KiB) Viewed 7373 times
Re: Need statistics Explanation
So far L2 cache only cache read-data, so when you copy 10GB data to the target drive, L2 won't get involved.FalkH wrote:1. create 10GB of data with h2testw on the cached harddrive
When you read back 10GB data at the first time, L2 cache will try to store them for future use. So you'll see L2 cache getting filled. PrimoCache is designed to fill L2 cache on idle time in order to not disturb other normal tasks. So you see the filling speed may not be fast as you expect.FalkH wrote:3. all h2testw activities on hdd are done, now primocache starts to fill the L2 cache
Re: Need statistics Explanation
Sure, yes, yes.FalkH wrote:2. I am still in the trial period, if I buy a license for 2.5 will it cover an upgrade to 3.x? Will a license grant lifetime upgrades?

Re: Need statistics Explanation
1. I see, than I misundertood the defer-write option. I thought this affects mainly the L2 cache. Something like writeback/writetrough.support wrote:1. ..., L2 won't get involved.
3. ... PrimoCache is designed to fill L2 cache on idle time in order to not disturb other normal tasks. So you see the filling speed may not be fast as you expect.
2. I hope this gets optimized with version 3.0 soon too

Since the taskman shows 100% active disk time at 32,9 MB/s there should be some potential for making the access faster. The data produced by h2testw are 1GB files, should be sequential written with around 400MB/s.
Anyway, I will also try to reconnect my drives in a diferent way. Maybe it works better with hdd and L2 on different controllers. Thanks for your explanations.
Re: Need statistics Explanation
Just wanna give some feedback. My guess about the controller seems right. Fill the cache is 10 times faster now, since ssd cache and cached disk doesn't share the same controller.
I would strongly advice anybody to try a second controller for connecting the L2 cache disk.
I would strongly advice anybody to try a second controller for connecting the L2 cache disk.
- Attachments
-
- SpeedUp.jpg (324.17 KiB) Viewed 7081 times
Re: Need statistics Explanation

BTW, What's the type of interface to connect your disks to the computer? IDE, sata, or others?