For the past ~2 decades, Windows has had a little option when you right-click a local drive and hit properties, to 'Compress Drive to Save Space'. It's a tempting little button, I mean who wouldn't want to get more disk space out of their existing drive? But there's gotta be a catch, nothing is truly free- so most of us never touch that button.
There's been plenty of speculation around it over the years, with some recommending never to enable it as it'll drastically slow down your system or drive performance. Others claim the compression it performs really isn't all that spectacular from the point of space saving to begin with. This is all really derived from common sense with basic knowledge of how file compression works. And honestly, that speculation does translate to reality.
I spent around a week and a half experimenting with a test bed trying to settle these theories once and for all, and to separate fact from fiction. The results were interesting and in some ways surprising, but showed a very clear trend that, at least for me personally, settled my wonders about this topic once and for all quite definitively. With the tests conducted on relatively high capacity drives like these, we can see just how effective (or ineffective) Windows software compression truly is in this day and age.
SCENARIO 1: Compression ON vs OFF
Compression enabled vs disabled on ~500GBs of diverse real-world data on a 1TB external SSD on Win10 client
WITHOUT Compression:
Summary:
1TB D: Drive= 476GB Total Used Space, 454GB Still Free (49% free space)
Savings- 0GB (uncompressed)
WITH Compression:
Summary:
1TB D: Drive= 451GB Total Used Space, 479GB Still Free (52% free space)
Savings- 25GB (5% savings)
Filetype contents and ratios:
NOTE: .CAS files are game data files for Call of Duty- this dataset contains a lot of videogame data files:
This dataset- although very realistic for a modern gamer, is not representative of ALL high-data-use data found on all systems, however most common space-consuming data types (such as a drive filled with tons of video or photo data for video editors or photographers) are already compressed in their own codecs and are not natively very compressible after they have already been encoded. So these results should be representative *enough* for use as general result set.
SCENARIO 2: DEDUPLICATION WITH COMPRESSION
Compression leveraged alongside pre-deduplicated volumes on 1TB SSD and 8TB HDD on WinServer 2019
Dedupe WITHOUT Compression:
Summary:
1TB D: Drive= 995GB Total Saved, 555GB Still Free (71% savings)
8TB F: Drive= 947GB Total Saved, 4.55TB Still Free (25% savings)
Dedupe WITH Compression:
Summary:
1TB D: Drive= 971GB Total Saved, 591GB Still Free (72% savings)
8TB F: Drive= 951GB Total Saved, 4.52TB Still Free (25% savings)
Filetype contents and ratios on each drive:
D:\ Drive (VMDK's are virtual machine disks, VBK's are backup files- both highly dedupe'able):
F:\ Drive (CCC's are deduplication-specific files):
Conclusion:
As you can see, these results are in some ways surprising but when it all boils down, quite negligible. These drives consist of a wide variety of filetypes, and plenty of each. On both solid state and mechanical drives. These are not synthetic files for benchmarks but real-world contents that are likely to be seen in other environments. So when we find that even over multiple Terabytes of data, we find almost no compression savings volume to volume, we can see it's really just not worth it. It adds CPU cycles, decreases performance when reading some filetypes, increases the possibility of file corruption/data inconsistency, decreases compatibility when copying over to other computers, and generally just adds headache to filesystems when trying to query or use them day to day.
The fact that in some cases we LOST space with compression in enabled, is the most shocking find in this test.
Childhood Story 1:
It reminds me of conducting real-world tests on RAID arrays back when mechanical drives ruled, and in practice, even though RAID 0's should have seen noticeable (if not extreme on-paper) performance gains as they scaled with multiple drives, actual tests would always show very very minimal changes if any at all (as is burned into my brain from my tests). I saw 97-103% performance conducted on three 2008-era 7200RPM RAID 0'd drives compared to a baseline non-RAIDed 100% number- and 98-102% with 2 drives. In some cases you actually lost performance, and it made sense because of physical drive seek times. Sometimes multiple drive heads looking for fragmented data may do worse than a single drive sequentially looking for its own data.
However, just to finish that RAID 0 statement... the paper performance gains are realized extremely well with solid state drives, since you eliminate that physical seek time delay.
/endrant
In general though, it is my opinion and from these findings, I'm never going to be enabling Windows compression on volumes in my systems going forward (and never really have before, as I was aware of the drawbacks and was never sure how effective the actual space savings could end up being). I would confidently tell others the same.
Childhood Story 2:
I remember in the early 2000's, I had this amazingly vast and detailed music collection I spent a solid year getting all the metadata exactly right, hand-typed for each and every track I'd Kazaa'd over the years in middleschool. I'd ripped every one of my CD's and even some of my parent's just for the sake of it. All beautifully accessible through my Windows XP MCE 2004 (Kazaa'd that too) desktop from Real Jukebox (remember that? Still one of my favorite music apps of all time, lightweight and straightforward- in my opinion superior to WinAmp... but it was shortlived as other Real software like RealOne, which sucked, became priority and they retired Real Jukebox before its time).
But Microsoft was marketing .wma as the total future, just as much fidelity as .mp3 at half the bitrate (128kbps MP3's were apparently equivalent to 64kbps WMA's). So I bought into it, and not having much local space anyway I deleted my MP3 collection after I converted everything without even listening to a single WMA conversion beforehand. And my god WMA's were un-listenable. Utterly butchered. The biggest data loss mistake I remember making from my childhood. But it did teach me to treat my data well and always make sure I had multiple copies before I jumped the gun or put data at risk in any way.
/endrant
So in summary, it's my professional and personal recommendation to not bother with Windows compression on any of your drives, and I hope these tests put this mystery to bed once and for all.
Comments