So, I bought myself a Bee-link ME Mini…
As read in this post, I need to make sure that the SSDs used (together with CPU, NIC, …) are not exceeding 45W of power. I trust there is a reason Bee-link teamed up with Crucial and on some tests I read that the P3 has a average wattage of 1,1W with peak at 5,2W. Using 6 of these should still keep me within this power limit.
This brings me to me question. How to configure TrueNas? Below my considerations. I would love to hear your thoughts on this and/ or your configuration.
| vDev type | Usage | Configuration | Considerations | |
|---|---|---|---|---|
| CACHE | Additional cache behind the ARC cache for caching files which are accessed a lot. ARC sits in RAM so it is much faster. No need for redundancy: upon failure/swapping no loss of data. | Not | As this vDev will have the same read/write speed as the other drives, there will be none to very little performance boost. Therefore, I prefer extra storage in the DATA vDev. Also, the Bee-link comes with 12GB of RAM and I read that 1GB for each TB stores is optimal. So, the optimal for the ARC is 8GB, leaving 4GB for TrueNas itself and for read/write operations. | |
| Data | Where the actual data is stored. Redundancy & resilience is key. | 6x Crucial P3 2TB RAIDz2 | Balance between redundancy & resilience while maintain 75% storage efficiency having a max storage of 8TB. | |
| Dedup | Stores the same data which is stored on multiple datasets/folders/… In case of small deviations the ‘main’ file is stored in the Dedup vDev with a file on each location in the DATA vDev holding the deviations. | Not. | Mainly, I am storing personal data with some logs files. Personal data is already deduplicated. | |
| Hotspare | Disk TrueNas will use upon Resilvering (when a disk in the Data vDev is no longer reachable. (NB does not ‘activate’ by bad sectors, etc.) | Not | I choose the keep a additional disk in the box. So, I swap this ‘cold’ disk when the Scrub or S.M.A.R.T. tests indicate errors. | |
| LOG | Stores the ZIL file for all async writes. Async is at a maximum of 5 seconds and given an input of (2x 2.5 Gbps) ~600MB/s a maximum of 3 GB is needed for these drives. However, use mirroring for redundancy | Not | Crucial P3 has a write speed of 3000MB/s and I use 6 of them. The NICs are the bottleneck here. Also, it seems a waste to by the smallest NVMe SSD (250 GB) twice while it will only use 3GB of it. | |
| METADATA | Stores the filenames, permissions and timestamps. Therefore, can be (a lot) smaller than the DATA vDev | Not. | As this vDev will have the same read/write speed as the other drives, there will be none to very little performance boost. Tricky to me as without this vDev the whole DATA vDev becomes unusable. |
Hope to hear what you think.
Would you choose the have 2x 1TB mirrored disks as a METADATA vDev and chosing a RAIDz1 for the remaining 4 disks in the DATA vDev? (Giving 6 TB of actual data at 75% storage efficiency.)
Or do you think configuring 2x 250GB mirrored disks as a LOG vDev will give a significant performance boost?
Love to see your considerations.