On Nov 15 2016, Randy Rue <[email protected]> wrote: > Hello All! > > I've been flailing on finding a way to replicate a large POSIX file > system off site and found this project, unless I'm missing some > dealbreaker it looks like S3QL is a great candidate for what we need. > > I've just spent some time reading as much of your documentation as I > can find but have a few questions, forgive me if I missed this > information: > * What scale are folks running this at? Is a PB typical, unusual, or > unheard of?
I haven't heard of it, i.e. I'm not aware of any positive or negative results with that much data. I expect that it well strongly depend on the number of files. > * What back ends are you running it on? Standard S3 only or is IAS > possible? IAS is an option, cf. http://www.rath.org/s3ql-docs/backends.html#cmdoption-s3_backend-arg-ia. > * What kind of GET/PUT traffic would a PB incur? Whatever you upload/download, rounded-up to the smaller of (blocksize, file-size), plus metadata upload/download on every mount/umount and every --metadata-upload-interval seconds. How much that amounts to depends on the number of files and the block size. > Our use case would be > as a DR replica for a PB space on-premise traffic would be almost all > writes. But would deletes count as reads? No. > Is every disk I/O to the NFS > export a PUT/GET, or are IOs somehow aggregated into less traffic on > the S3 objects? They are aggregated locally. By default, writes are committed when a block has not been touched for 5 seconds, or when the maximum cache size has been reached. > * Do I understand correctly from the docs that snapshots are RW unless > they're locked? Yes. Best, -Nikolaus -- GPG encrypted emails preferred. Key id: 0xD113FCAC3C4E599F Fingerprint: ED31 791B 2C5C 1613 AF38 8B8A D113 FCAC 3C4E 599F »Time flies like an arrow, fruit flies like a Banana.« -- You received this message because you are subscribed to the Google Groups "s3ql" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
