Re: Hardware Compression
2007-08-14 00:20:56
Hi Jon,
> LTO is unusual. When hwc is enabled each input block is compared
> with and without compression. The smaller of the two is recorded.
> As amtapetype feeds "random" data, data that is not compressible,
> the original input block is taped. Thus you see the same results
> with or without hwc enabled. And amtapetype can't tell whether
> hwc is enabled or disabled.
Ah, I see. So I don't have to worry about that anymore. Good.
> If some or all of your DLEs are not software compressed and you are
> using hwc, there is no single value that is accurate every day. The
> reason is that different data compress different amounts. Varying
> widely, like not at all to 90 percent. As what you backup each
> day varies, the degree of hwc will vary too.
> So pick a value and watch the daily reports. See how much data is
> actually written to your tapes. You will soon get a feel for whether
> it is too low or high. If you never get a tape overflow then how
> accurate your estimate is doesn't matter. If you regularly get
> tape overflow the reports will tell you how much amanda was able
> to write up to the end of tape. Compare several to approximate
> the tapelength. Note, this will primarily affect amanda's estimate
> and planning phase.
Sure, I will have to find a "good" value for the scaling factor within
the first test runs.
I was just confused because of the examples at zmanda-Wiki that used
IDENTICAL values for the tape length in both cases and in that case HWC
wouldn't make that much sense at all imho.
That was really a good and helpful answer. Thanks a lot!
Ralf
|
|
|