BLUEDOG314
Cadet
- Joined
- Jul 10, 2019
- Messages
- 4
Hi folks. Newbie poster, but have a FreeNAS server at home used for various purposes. My question involves making a high capacity build, around 100TB give or take, for storing reconstructed medical images along with their preconstructed parts.
Generally, the device will serve like I said to store medical images, so it has to have a good level of redundancy. IOPS are not a huge concern, but the machine that produces images churns out about 3.8-4GB of data in about 30 seconds. This data is currently held on a staging computer with SSDs then moved to our current holding solution at a rate it can handle. As needed, the preconstructed images are moved back to the holding machine to be combined and rendered. The rendered images are also then pushed back to our storage solution. The current solution is also accessed by various other machines to pull reconstructions for research purposes. Currently everything is running on 1Gbps interfaces.
Along with data redundancy we would like to remove the 1Gbps bottleneck and build something that can juggle data with the holding/rendering machine faster to improve workflow. This had me thinking of including some sort of mirrored SSD write cache. For the hard drives I've heard folks talk about mirrored vdevs along with or in place of RAIDZ. Not sure what makes the most sense here.
Anyway, I suppose after trying to research on my own I quickly realized the hardware needed is beyond my current knowledge, so any advice is appreciated. Thanks!!
Generally, the device will serve like I said to store medical images, so it has to have a good level of redundancy. IOPS are not a huge concern, but the machine that produces images churns out about 3.8-4GB of data in about 30 seconds. This data is currently held on a staging computer with SSDs then moved to our current holding solution at a rate it can handle. As needed, the preconstructed images are moved back to the holding machine to be combined and rendered. The rendered images are also then pushed back to our storage solution. The current solution is also accessed by various other machines to pull reconstructions for research purposes. Currently everything is running on 1Gbps interfaces.
Along with data redundancy we would like to remove the 1Gbps bottleneck and build something that can juggle data with the holding/rendering machine faster to improve workflow. This had me thinking of including some sort of mirrored SSD write cache. For the hard drives I've heard folks talk about mirrored vdevs along with or in place of RAIDZ. Not sure what makes the most sense here.
Anyway, I suppose after trying to research on my own I quickly realized the hardware needed is beyond my current knowledge, so any advice is appreciated. Thanks!!