Hi there, I hope it is OK to ask for opinions, recommendations, experiences around $SUBJECT here. I am a long time SUSE user and I am currently designing what should be my next high-end system. This system will have to act as - Virtualization host - Storage provider - Desktop system - Mail server - Web server None of the services are accessible to the Internet, services are only shared/exposed within my local GigE network. The new system will get an up-to-date Intel CPU (probably the Intel Core i7-6700 - I am not sure about the extra 26 Watts the 6700K will bring with it). For comparison, this is my current setup: - Intel(R) Core(TM) i7-3770 CPU - 32 GiB RAM - 1x bootable SSD 512 GB; used to multi-boot between 4 different installations (13.1, 13.2, Leap 42.1, Tumbleweed) - 4x spinning 2 TB hard disks in MD-raid10 mode - NFSv4 as server and client - Latest kernel from Kernel:stable:standard on all distros (currently 4.6.3) As the new system will most likely use an NVMe or M-SATA device as the boot device, Tumbleweed with its 4.6 kernel is the proper choice from my point of view to get all new devices (such as Intel NICs etc) supported; nevertheless, I have installed the lastest 4.6 kernel from Kernel:stable:standard on all remaining distributions and use it successfully on the above (old/current) setup. So, the new kernel is probably just used in preparation what will come up with the new system. The performance of the current system is almost OK, but the spinning disks are what attracts me to build a new system. Replacing all hard disks with SSDs of similar size is not reasonable due to the price point, though. I have therefore looked at caching technologies such as "bcache" and found it useful. The problem with it however is, that (a) it is not supported by the installer and (b) the required packages (bcache-tools) are not included with the rescue part of the installer image. (a) doesn't bother me, as I would like to copy over my current installation anyway - which would just need support for the network interface and some other relevant devices - but (b) forms sort of a problem, because I believe on such a current system I would need "nvme-cli" to be available in addition to the mentioned "bcache-tools" ("bcache-tools" contains the files required by "dracut" and "udev" to get such devices "/dev/bcache0" work transparently). I have tried it out on a virtual system, copied the "bcache-tools" RPM onto the rescue system inside the VM, ran "udevadm trigger" - and voila - the VM which had been setup before to use a bcache'd drive for its /home was ready to use from inside the rescue system. Please note that I'm only describing this here as a vehicle to copy over a fully configured and working system to a new one which will simply boot from rescue system and I will do all partitioning etc. there, followed by rsync'ing the old system to the new one. On the other hand, "bcache" may appear to be one option, if you know of others being able to replace the MD-raid10 I'm currently using, achieving better/similar throughput, but definitely better latencies, I am completely open to look at them, too. If, however, you think the "bcache" way might be a good option to host very fast storage for root and /boot (with several versions for multiple distros in multi-boot) and offer a fast caching method for a slower MD-raid1 of two 4 TB hard disks behind, then I believe we should think about inclusion of "bcache-tools" at least, and "nvme-cli" for newer systems anyway. What do you think about this? Your comments are very well appreciated! TIA, cheers. l8er manfred