1) Yes, absolutely! We'll make sure to reference existing VSS provider also works for LSFS.
2) Yes, absolutely! LSFS is a log-structured file system so never overwrites anything and just appends all the data. Incomplete write can make you lose few megabytes (up to one page in size, 4 megabytes by default) of recent transactions. However you should be running fault tolerant config (2-way or 3-way replication between LUs) and in this case you never lose anything.
3) You don't replace VHDX you replace FLAT with LSFS. Performance depends on your workload. If you have many VMs and they are write-active you'll get ±10x of performance increase when moving FLAT -> LSFS. There are no random writes with it. If your workload is mostly reads you'll see no performance increase (just throw in more RAM and flash for L1 and L2 caches instead).
4) There's no limit. LSFS does keep track of all the write transactions so if you're not going to trim it it can grow and grow. Basically with any log-structured file system you turn disk space into performance (TBs to IOPS). Make sure you properly configure it to allow it to clean itself to prevent uncontrolled growth.
Tim wrote:Ok I have some questions:
1) Is LSFS VSS aware? Can I do a backup of the storage files and get a consistent copy?
2) Is LSFS crash safe? Can it recover from a system crash without corruption?
3) Is it worth replacing my local thick VHDX files with thin LSFS? Any performance gains?
4) If I create a 100GB thin LSFS, autodefrag, no-dedude. What's that maximum space that will be used on the storage volume? Will it be limited to ~100GB?
Thanks
Tim
