Results of running 'specjbb2005' on increasing (from 1 to 8) number of (PV, for now) guests on the same shared hosts.
Numbers come from the execution of 5 runs of the benchmak for each configuration. Stats of the aggregate performance results of running
the benchmark on all the VMs (i.e., the sum of each result achieved by each VM divided by the number of VMs involved in the specific run)
are shown in this page. As for thes runs,
NUMA support patches) were on the job in these runs.
Host and guests characteristics were the following:
- HOST: 16 CPUs, 2 NUMA nodes, 12GB RAM (2GB reserved for Dom0);
- GUESTS: 4 VCPUs, 1GB RAM.
The various lines in the plots comes with the following meaning:
- "default" is just the defaul Xen/xl behaviour when creating the various guests with 'xl create' and letting them run free
of any specific constraint wrt VCPU scheduling and memory allocation.
- "first fit" means that _all_ the VMs were fit on the host's nodes with the First Fit algorithm at creation time
(i.e., created with "node=ffit" in their config file).
- "best fit" means that _all_ the VMs were fit on the host's nodes with the Best Fit algorithm at creation time
(i.e., created with "node=bfit" in their config file).
- "worst fit" means that _all_ the VMs were fit on the host's nodes with the Worst Fit algorithm at creation time
(i.e., created with "node=bfit" in their config file).
VMs are all equal and are always created in the very same order (which is definitely not the best configuration
for stressing the placement heuristics, more numbers coming).
