[pbs-devel] applied: [PATCH backup 1/2] client: avoid division by zero in avg speed calculation, be more accurate
Thomas Lamprecht
t.lamprecht at proxmox.com
Fri Jul 24 14:37:09 CEST 2020
Am 7/24/20 um 2:01 PM schrieb Dietmar Maurer:
> Why don't you use a single duration.as_secs_f64() and compute speed using floting point?
>
Because usize -> f64 is a loss of precision and mainly this is moving back
to exactly how you used to do:
https://git.proxmox.com/?p=proxmox-backup.git;a=commitdiff;h=82ab72304efd651d2afa771c499d4bcec2787f64
thought you had similar/some reasons...
And HumanByte has no from<f64> yet so either adding that or casting back would be
required, but sure, have no hard feelings between:
... = ((uploaded as f64 / duration.as_secs_f64()) as usize).into()
vs.:
... = ((uploaded * 1_000_000) / (duration.as_micros() as usize)).into();
>> On 07/24/2020 10:16 AM Thomas Lamprecht <t.lamprecht at proxmox.com> wrote:
>>
>>
>> using micros vs. as_secs_f64 allows to have it calculated as usize
>> bytes, easier to handle - this was also used when it still lived in
>> upload_chunk_info_stream
>>
>> Co-authored-by: Stoiko Ivanov <s.ivanov at proxmox.com>
>> Signed-off-by: Thomas Lamprecht <t.lamprecht at proxmox.com>
>> ---
>> src/client/backup_writer.rs | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/src/client/backup_writer.rs b/src/client/backup_writer.rs
>> index 7e5adb3..2344045 100644
>> --- a/src/client/backup_writer.rs
>> +++ b/src/client/backup_writer.rs
>> @@ -264,7 +264,7 @@ impl BackupWriter {
>> crate::tools::format::strip_server_file_expenstion(archive_name.clone())
>> };
>> if archive_name != CATALOG_NAME {
>> - let speed: HumanByte = (uploaded / (duration.as_secs() as usize)).into();
>> + let speed: HumanByte = ((uploaded * 1_000_000) / (duration.as_micros() as usize)).into();
>> let uploaded: HumanByte = uploaded.into();
>> println!("{}: had to upload {} from {} in {}s, avgerage speed {}/s).", archive, uploaded, vsize_h, duration.as_secs(), speed);
>> } else {
>> --
>> 2.27.0
More information about the pbs-devel
mailing list