Hot Deal - Intel Optane 905P 960GB w/U.2 to M.2 adapter cable $219.99
- Thread starter Thomas H
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Optane is great. But Optane is also attacked by multiple fronts by RAM and the NAND SSD.
I would say that TODAY, OPTANE vs FLASH SSD
1. Low queue depths random read, Optane has still a big advantage (like 3x~4x faster).
2 Low queue depths random write, Optane is on par with pSLC drives and better than TLC drives.
3. High queue depths random read/write, flash SSDs are better.
4. Sequential random read/write, flash SSDs are better.
5. Has great endurance, but nothing that can't be replicated with sufficient overprovisioning on the flash side.
So you really need to think about an application where you need to flush to permanent storage immediately, there's no time to batch things (make it sequential operation), and the access pattern is random. The answer to this is probably a mission critical database.
For L2ARC, which is a ZFS optimization for read operation, it's more economical and performant to just have more ram.
For ZIL/SLOG, which is a ZFS optimization for synchronous write operations (random and sequential, but mostly small stuff), I think it's toss up between optane and good modern flash drive.
For metadata vdev, which is a filesystem map/database with small entries, I think optane is still a top choice.
I would say that TODAY, OPTANE vs FLASH SSD
1. Low queue depths random read, Optane has still a big advantage (like 3x~4x faster).
2 Low queue depths random write, Optane is on par with pSLC drives and better than TLC drives.
3. High queue depths random read/write, flash SSDs are better.
4. Sequential random read/write, flash SSDs are better.
5. Has great endurance, but nothing that can't be replicated with sufficient overprovisioning on the flash side.
So you really need to think about an application where you need to flush to permanent storage immediately, there's no time to batch things (make it sequential operation), and the access pattern is random. The answer to this is probably a mission critical database.
For L2ARC, which is a ZFS optimization for read operation, it's more economical and performant to just have more ram.
For ZIL/SLOG, which is a ZFS optimization for synchronous write operations (random and sequential, but mostly small stuff), I think it's toss up between optane and good modern flash drive.
For metadata vdev, which is a filesystem map/database with small entries, I think optane is still a top choice.
The thing is, its mostly database applications where transactional workloads are common that you need synchronous write. Like MySQL, PostgreSQL, SQLite, RocksDB, FoundationDB, Kafka, Pulsar and so on. For special metadata, a normal SSD is plenty for spinning drives.