Hi, Since a few days ago, I have been trying to create a glusterfs demo to show my boss that this piece of software has potential to us. I have been using the following hardware and software combination: hardware: Dell Studio XPS 16, Intel i7 CPU (4 core), 8GB RAM, 128 GB Samsung PW80 SSDOS: Ubuntu 10.04LTS but running the backported 2.6.35-23 kernelglusterfs 3.1.2 deb package downloaded from the gluster Web site. I created 3 KVM guests, all on a LVM volume. Each guest resides in its own logic volume (16GB each), has 512MB RAM, and 2 vcpus. All guests use virtio for both storage and networking.
Two of the guests are glusterfs servers in replicated configuration (replica=2), the 3rd guest acts as a client. The glusterfs native protocol is used for mounting in the client guest. I then try to run bonnie++ in the client machine. When running directly, the following takes less than a minute wall time to finish: /usr/sbin/bonnie++ -d /tmp/foo But when pointing the -d to a subdirectory in the imported volume, even the first part: Writing a byte at a time... takes forever. In the Virtual Machine Manager, the client shows a lot network traffic, but little is accomplished with such traffic. Currently, the glusterfs configuration is default for everything. I was hoping to use these defaults to generate a set of baseline data, and go from there. But looks like the way it is, I will have to take a different approach. Question: How can I make bonnie++ run faster with my "demo" setup? I would love to hear some tuning tips that can make my demo run faster! Thanks, -- Zack
_______________________________________________ Gluster-users mailing list [email protected] http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
