How to make a Raid 6 w/ SSD Cache

[TechLabs]

2 hours ago, jkirkcaldy said:

The problem is with things like the benchmark tests is that they do great usually because the files that they are writing and reading are all still contained on the ssd. So if you are reading a lot of files shortly after writing them you will see a huge increase in performance like you did above. But I would wager that if you were to copy a large file onto the array then copy it back again immediately, record the results, then wait until the file isn't on the cache any more and copy it off the array you will see far less impressive numbers. 

 

But if it's as much a technical exercise as it is a practical one, and you can spare the SSD then it's worth it. 

When I was transfer files off my main computer to the server via ethernet cable, I was getting ~100 MB/s. HDD TO RAID 6.

I will also test it the other way around as well. Is there any other test that I should do?

[scottyseng]

52 minutes ago, TechLabs said:

When I was transfer files off my main computer to the server via ethernet cable, I was getting ~100 MB/s. HDD TO RAID 6.

I will also test it the other way around as well. Is there any other test that I should do?

Ah, when you test using crystal disk mark, try at least 5 passes (I usually do 9) and at 4GiB instead of the normal 1GiB. It's possible that if the file is small enough, the RAID card will just benchmark against the memory cache the RAID card has (Some have 512MB, these days the newer ones have 1 to 2+ GB of DDR3)

 

Also make sure your RAID card has a working battery. Speeds really take a hit when you have to change to write through mode (bypass RAID card cache). 

[TechLabs]

13 minutes ago, scottyseng said:

Also make sure your RAID card has a working battery.

I like to live life to the edge.. yeah, I need a battery unit for my Raid Controller 

[TechLabs]

6 hours ago, jkirkcaldy said:

The problem is with things like the benchmark tests is that they do great usually because the files that they are writing and reading are all still contained on the ssd. So if you are reading a lot of files shortly after writing them you will see a huge increase in performance like you did above. But I would wager that if you were to copy a large file onto the array then copy it back again immediately, record the results, then wait until the file isn't on the cache any more and copy it off the array you will see far less impressive numbers. 

 

But if it's as much a technical exercise as it is a practical one, and you can spare the SSD then it's worth it. 

 

^ The server to my main pc on the SSD

 

^ The server to my main pc on the HDD

 

* I saw a few dips on the HDD transfer. The SSD kept a steady flow of 100 MB/s. If I had a 10GbE SFP NIC to test with, I might have gotten higher~ish number with my SSD transfer test.

 

^This is on my server machine transferring from RAID 6 array to my SSD (HP M700). 

 

2 hours ago, scottyseng said:

Ah, when you test using crystal disk mark, try at least 5 passes (I usually do 9) and at 4GiB instead of the normal 1GiB. It's possible that if the file is small enough, the RAID card will just benchmark against the memory cache the RAID card has (Some have 512MB, these days the newer ones have 1 to 2+ GB of DDR3)

 

Also make sure your RAID card has a working battery. Speeds really take a hit when you have to change to write through mode (bypass RAID card cache). 

1

 

^ Most of the movies are around 16 GB in size.