...1588v2..

On Thu, Jan 9, 2020 at 7:36 PM Ronen Friedman <rfriedma@redhat.com> wrote:
Just a note re time sync standards:
In my previous job, we had been playing with the TSN standards, trying to achieve synchronization in the 10s of microseconds
realm. The specific TSN standard (802.1AS) is a specific PTP profile (more or less 1588 v3, if I remember correctly).
To achieve sub-millisecond accuracy one should better have hardware timestamping, but we did achieve very good
results with just software times.

Ronen


On Thu, Jan 9, 2020 at 5:09 PM Brett Niver <bniver@redhat.com> wrote:
I played with vectorial clocks 10-15 years ago.  I think in 4) you have a typo, casual should be causal, right?  I really like Lamport clocks conceptually, I personally found implementing based on that concept to have more corner cases than I liked but that could have been just me.

What are your time sync requirements with respect to NTP?  Aren't there better algorithms out today?


On Wed, Jan 8, 2020 at 10:43 PM Xuehan Xu <xxhdx1985126@gmail.com> wrote:
On Thu, 9 Jan 2020 at 02:45, Matt Benjamin <mbenjami@redhat.com> wrote:
>
>Historically, and the discussion has been about the perf. and complexity trade-offs, I think.

Thanks for reviewing, Matt, I totally agree:-)

>
> Matt
_______________________________________________
Dev mailing list -- dev@ceph.io
To unsubscribe send an email to dev-leave@ceph.io