On 02/24/2017 01:54 PM, Peter Lieven wrote: > > Changing the compression level via environment or a QCOW2 parameter is not a > big deal. However, it helps not that much regarding speed. What helps is > increasing the window size, but this is an incompatible change. And if we go > for an incompatible change I would think about using a better compression > algorithm than zlib deflate.
Indeed, lzma (think xz), or zstd, or $compression_of_the_day... may be better, and it may be disk-image dependent which algorithm is best. It sounds like we need to improve the qcow2 spec to add a header extension that describes the current compression algorithm and parameters; if omitted the image uses zlib with fixed window size, but if present, we can put everything we need to be able to choose other compression tunings and be able to decompress (still, it has to be an all-or-none). We'd also need to set an incompatible_features flag if the compression header extension is present with non-default values so that older qemu don't go corrupting the compressed data. And then you start having to worry about modularity (do you compile qemu with support for all compression libraries, or can you pick and choose which ones to compile in at the risk of which images you can't open, or do you figure out how to dlopen the right libraries as needed...) -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
