• Developer

    Hi all,

    just out of curiosity has anyone done much testing with compression over download speed? especially with it now being partclone?

    it’s just space is not a concern but wondered how much benefit there is with compression regarding deploy tasks, i know it makes upload tasks quicker (when not compressed) or the image smaller (compressed) but has everyone found that the more compression the impact of deploy speed?

    just seeing what everyone elses experiences are… 🙂

  • Your findings, Lee, is about what I’d expect. This, of course, if the deploy is done (well maybe not now) from a unicast stand point rather than a Multicast.

    Compression is a fickle thing. Uploading anything and compressing before doing so always take far more time than re-obtaining the data even if the re-obtaining is highly compressed. This is because the CPU is actually doing much more work finding what can and how much it can be compressed.

    On a download, the less data you have to transfer, the faster it transfers.

  • Moderator

    Tom, what have you noticed?

  • Moderator

    Interesting. Will have to do a few tests myself on this one.

  • Developer

    initial testing (16GB image) image pushed from same server to same client, same image uploaded from same client but one image upload with no compression and the other image upload with highest(9) compression. no other traffic on 1GB network during testing.

    upload speed seems much better on partclone than partimage regardless of compression.

    [U]No Compression Results:[/U]
    Upload speed = Avg 2.02GB/Min
    Image size on Server = 16,244,393,968 (main partition)
    Deploy speed = Avg. 3.08GB/Min

    [U]Highest Compression Results:[/U]
    Upload speed = Avg 1.2GB/Min
    Image size on Server = 7,576,511,522 (main partition)
    Deploy speed = Avg. 4.51GB/Min

    Re-run both tests to make sure!!

    i actually got better deploy speeds with high compression! doesn’t seems logical to me but not complaining!! May be due to decent processor (quad core i7 4770K) on the client, might be worth testing on clients with low spec processors to see if get same outcome - also noticed deploy speeds much better if the image was uploaded using partclone rather than using existing image created by partimage.