@sparx Yep, Sweden.
@marcoi These are great for SLOG. Superb endurance, power loss protection, high and consistent IOPS, and low and very consistent write latency. Just over provision the drives by making the partition small, like 8-20 GB, enough to cover 2-3 times the size of one transaction group. The size you need will vary based on your RAM size and other hardware + settings (flush time, network speed, etc).
I don't use mine for SLOG. My two DC S3700 400GB drives are configured in RAID 1 with a very simple, non-caching, hardware RAID controller (SAS2008). They perform great for all my VMs.
ZFS SLOG is only useful for synchronous writes, which itself is a good idea when running VMs from the array. RAM will not help performance for synchronous writes. The benefit of an SLOG is that you can treat the writes as asynchronous and temporarily store them in the regular transaction log in RAM after they have been committed to the SLOG device, instead of having to wait while each is committed to the spinning media.
SLOG is a dedicated device for ZIL. Without SLOG, the ZIL for synchronous writes is stored on the same media as long-term storage, only it has to be written twice. First quickly to prevent data loss, and then again as part of the regular write flush of a transaction group.