Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Ard van Breemen wrote: > Hi, > On Thu, Apr 27, 2006 at 09:36:04AM +0200, JMWorld - Dpto. T?cnico wrote: > >>Im testing drbd before using it and I see (at /proc) that after creating >>partitions, and starting systems, it begins to synchronize, but partitions >>are about 60GB and we have to pay for every GB transmitted on the final >>systems. >>Since partitions are inittially empty, is there any way to mark volumes as >>synchronized in order to avoid transmitting such null and huge information >>or even shorten the size transfering only main parts or meta-data? > > > Sorry for this nasty suggestion, but it even helped me getting > plain mysql replication down to 10% of the original traffic > (which was needed because of a 410kb/s VPN): use vtun with a tcp > tunnel with zlib compression factor 9 or so. You can even do VoIP > over that and not notice it. > Allright, it is not the answer that you seek, but if you pay per > gig, compress it. It will probably save that much money within a > year that you can buy faster processors to begin with (to support > the extra load of compression), or more memory (so things do not > get flushed that often, which in the end also saves you gigs). > Sounds like an interesting solution, thanks for the info. An additional thing would be to do the initial sync while the machines are within ~3 meters of each other with a crossover cable, and then put one at the remote data center. Once they (drbd 0.[78].x) have done the initial sync, they should only transfer the data that has changed, unless something catastrophic happens or _you_ tell one to invalidate itself. -- Todd Denniston Crane Division, Naval Surface Warfare Center (NSWC Crane) Harnessing the Power of Technology for the Warfighter