| 982a1bb7 | 21-Feb-2023 |
Michal Berger <michal.berger@intel.com> |
test/vhost: Make sure $disk_map exists
Also, simplify the way how it's read. As a benefit, this gets rid of the xargs complaining about a NUL being sent out to its input which was quite verbose.
Si
test/vhost: Make sure $disk_map exists
Also, simplify the way how it's read. As a benefit, this gets rid of the xargs complaining about a NUL being sent out to its input which was quite verbose.
Signed-off-by: Michal Berger <michal.berger@intel.com> Change-Id: Iaeb09298c2255404273bb3fc6910bc6b93c2d7eb Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16892 Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
show more ...
|
| 47c2ee64 | 20-Feb-2023 |
Michal Berger <michal.berger@intel.com> |
test/vhost: Remove cpuset limit enforced upon vhost cpus
This limit didn't do much in a first place. It was creating separate cgroup with mem nodes already set to 0-1 nodes - in practice these are a
test/vhost: Remove cpuset limit enforced upon vhost cpus
This limit didn't do much in a first place. It was creating separate cgroup with mem nodes already set to 0-1 nodes - in practice these are all the NUMA nodes available for mem allocation by the very default. Regarding cpus, vhost is already limited by its own cpumask's affinity hence there's no need to enforce this limit via a dedicated cgroup. Lastly, this was not taking into consideration the fact that other processes may still be scheduled on the vhost cpus as the cgroups they belong to were not modified in any way (like in case of the scheduler tests for instance). That said, the amount of jitter coming from these 3rd party processes would not have much bearing on vhost anyway - the only processes that could be more noisy are QEMU's but each VM instance is already put into a separate cgroup (see test/vhost/common.sh).
Signed-off-by: Michal Berger <michal.berger@intel.com> Change-Id: I1de24bfc9e24f8f6391207e579cc599ea5c82094 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16890 Reviewed-by: Karol Latecki <karol.latecki@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
show more ...
|
| 98d98ecb | 20-Feb-2023 |
Michal Berger <michal.berger@intel.com> |
test/vhost: Switch from msdos to gpt
Disks used under the vhost benchmarks can be > 2TB so the msdos pt is not very suitable here. Use something more robust like gpt.
Signed-off-by: Michal Berger <
test/vhost: Switch from msdos to gpt
Disks used under the vhost benchmarks can be > 2TB so the msdos pt is not very suitable here. Use something more robust like gpt.
Signed-off-by: Michal Berger <michal.berger@intel.com> Change-Id: I3e98bcb655c2f55a515f4000b0668b26d71c8fca Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16889 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com>
show more ...
|
| c1d2bdfe | 23-Nov-2022 |
Michal Berger <michal.berger@intel.com> |
test/vhost: Add perf collection support
Signed-off-by: Michal Berger <michal.berger@intel.com> Change-Id: I059658c477be4122e7b04f33a796f732746b7c90 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/
test/vhost: Add perf collection support
Signed-off-by: Michal Berger <michal.berger@intel.com> Change-Id: I059658c477be4122e7b04f33a796f732746b7c90 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15603 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
show more ...
|
| 61ecba1a | 19-Jan-2021 |
Karol Latecki <karol.latecki@intel.com> |
test/vhost_perf: set test block device scheduler to "none"
In recent systems default is set to bfq which lowers performance when using SPDK vhost.
Change-Id: Iefa3b8b514dac2f7b028cf06c18cee3f59c5a9
test/vhost_perf: set test block device scheduler to "none"
In recent systems default is set to bfq which lowers performance when using SPDK vhost.
Change-Id: Iefa3b8b514dac2f7b028cf06c18cee3f59c5a9ac Signed-off-by: Karol Latecki <karol.latecki@intel.com> Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5982 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Maciej Wawryk <maciejx.wawryk@intel.com> Community-CI: Mellanox Build Bot
show more ...
|
| 757eddeb | 23-Jul-2020 |
Karol Latecki <karol.latecki@intel.com> |
test/vhost: add option for fio gtod_reduce
Enable gtod_reduce fio option on demand for better IOPS/BW results at cost of reducing latency measurements.
Signed-off-by: Karol Latecki <karol.latecki@i
test/vhost: add option for fio gtod_reduce
Enable gtod_reduce fio option on demand for better IOPS/BW results at cost of reducing latency measurements.
Signed-off-by: Karol Latecki <karol.latecki@intel.com> Change-Id: I0511a5fdc09bb20b5a769ad55f2306b87e5d2dde Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3502 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Community-CI: Mellanox Build Bot Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
show more ...
|