Hi friends, I got the system is working and it shows not bad performance comparing even to direct write to HDD. The very basic comparation (dd if=/dev/zero of=/tmp/t.bin bs=1G count=3 oflag=dsync) on HDD-based 1Gbps system with two replicas shows the following result:
- while HDDs directly give about 170MBytes/s - VM, which reside on "disk" node, writes at about 100-110MBytes/s (this is a bit below of ceiling of 1Gbps) - while "diskless" VM writes at twice less rate - about 45-50Mbytes/s The last result is so worst because of diskless station writes two copies of data, thus giving about 1Gbps of total output on originating host, which is divided between two flows. For small systems like 2 "disk" nodes and 3rd "diskless" node (tie breaker role) this isn't a problem, while for more scaled systems, where most of VMs will reside on diskless nodes, this will have a notable performance impact, which will as more as more replicas will be used (e.g. for system, configured for 3 replicas, the test above will show 30MB/s write speed). The question is - whether DRBD supports multicast in order to generate a single flow from originating node? This will allow to dramatically improve performance of scaled system and will allow to build more scaled systems. Thank you. -- Volodymyr Litovka "Vision without Execution is Hallucination." -- Thomas Edison _______________________________________________ Star us on GITHUB: https://github.com/LINBIT drbd-user mailing list [email protected] https://lists.linbit.com/mailman/listinfo/drbd-user
