Rookie here again with my second question. I appreciate all the advice you guys gave me on my hard drive cooling thread btw.
While researching read/write speeds expectations of a RAIDZ2 pool I came across this website (https://calomel.org/zfs_raid_speed_capacity.html) that had some pretty eye-opening comparisons. I have tried to find other places that provide the same types of results in order to verify that this isn't a one-off odd setup but I have been unable to locate anything.
About 2/3 of the way down the page the author sets up a test comparing different types of SATA controllers and the title of the section is "All SATA controllers are NOT created equal". The test shows that using the LSI MegaRAID SAS 9265-8i RAID Controller provides a very large performance advantage compared to onboard controllers in most cases. It even uses a Supermicro mboard as one of the examples. So I read through the stickied LSI thread in this forum and felt pretty good about going with the recommended
IBM ServeRAID M1015 card that you convert to an LSI SAS2008 but when I read through some of the reviews and some comments in the review threads everyone was talking about how the SAS2008 had great read performance but was lacking on the write side... I did some more digging and found out that the cheaper converted card doesn't have a cache on it but the better LSI card that was used on the calomel article does and that was the reason why it was so much better. I looked at prices for both and it looks like a $400 purchase to get the better card....
My first question is whether anyone is using the LSI MegaRAID SAS 9265-8i card and has an idea of possible pitfalls I might run into if I use it on a Supermicro board compared to the cheaper card.
My second question is whether the differences in performance in the article are in the same ballpark for what I could expect while running FreeNAS? Does their test setup have any peculiarities that would lead you to believe that the gains he saw would not translate to a FreeNAS server? While performing a write operation the FreeNAS OS would take advantage of a write cache on an LSI card right? It wouldn't bypass the cache if I put the card in JBOD mode?
Any insight would be most appreciated! Thanks in advance!
While researching read/write speeds expectations of a RAIDZ2 pool I came across this website (https://calomel.org/zfs_raid_speed_capacity.html) that had some pretty eye-opening comparisons. I have tried to find other places that provide the same types of results in order to verify that this isn't a one-off odd setup but I have been unable to locate anything.
About 2/3 of the way down the page the author sets up a test comparing different types of SATA controllers and the title of the section is "All SATA controllers are NOT created equal". The test shows that using the LSI MegaRAID SAS 9265-8i RAID Controller provides a very large performance advantage compared to onboard controllers in most cases. It even uses a Supermicro mboard as one of the examples. So I read through the stickied LSI thread in this forum and felt pretty good about going with the recommended
IBM ServeRAID M1015 card that you convert to an LSI SAS2008 but when I read through some of the reviews and some comments in the review threads everyone was talking about how the SAS2008 had great read performance but was lacking on the write side... I did some more digging and found out that the cheaper converted card doesn't have a cache on it but the better LSI card that was used on the calomel article does and that was the reason why it was so much better. I looked at prices for both and it looks like a $400 purchase to get the better card....
My first question is whether anyone is using the LSI MegaRAID SAS 9265-8i card and has an idea of possible pitfalls I might run into if I use it on a Supermicro board compared to the cheaper card.
My second question is whether the differences in performance in the article are in the same ballpark for what I could expect while running FreeNAS? Does their test setup have any peculiarities that would lead you to believe that the gains he saw would not translate to a FreeNAS server? While performing a write operation the FreeNAS OS would take advantage of a write cache on an LSI card right? It wouldn't bypass the cache if I put the card in JBOD mode?
Any insight would be most appreciated! Thanks in advance!
