[ovs-dev] [PATCH per-port ingress scheduling 2/2] ingress scheduling: Provide per interface ingress priority

interfaces from packet loss as PMDs get overloaded. Signed-off-by: Billy O'Mahony --- include/openvswitch/ofp-parse.h | 3 + lib/dpif-netdev.c | 188 +--- lib/netdev-dpdk.c | 10 +++ 3 files changed, 170 insertions(+), 31 deletions

[ovs-dev] [PATCH per-port ingress scheduling 1/2] ingress scheduling: documentation

Signed-off-by: Billy O'Mahony --- Documentation/howto/dpdk.rst | 15 +++ vswitchd/vswitch.xml | 15 +++ 2 files changed, 30 insertions(+) diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst index ab3d576..83284e7 100644 --- a/Documentation

[ovs-dev] [PATCH per-port ingress scheduling 0/2]

* applies cleanly on 9b4f08c RFCv2: * Keep ingress prio config in netdev base rather than in each netdev type. * Account for differing rxq lengths * Applies cleanly to 4299145 RFCv1: Initial version. Billy O'Mahony (2): ingress scheduling: documentation ingress scheduling: Provide per interface

[ovs-dev] [PATCH v3] OVS-DPDK: Change "dpdk-socket-mem" default value.

higher. Patch contains documentation update. Signed-off-by: Marcin Rybka <marcinx.ry...@intel.com> Co-authored-by: Billy O'Mahony <billy.o.mah...@intel.com> Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.r

[ovs-dev] [PATCH v3] OVS-DPDK: Change "dpdk-socket-mem" default value

Hi All, This change is the outcome of this discussion: https://mail.openvswitch.org/pipermail/ovs-discuss/2017-April/044110.html. I am posting this on behalf of Marcin who is the author but no longer works for Intel. I made a few changes hence the double sign-off but please remove my sig if

[ovs-dev] [PATCH v2] OVS-DPDK: Change "dpdk-socket-mem" default value.

higher. Patch contains documentation update. Signed-off-by: Marcin Rybka <marcinx.ry...@intel.com> Co-authored-by:: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.rst | 3 ++- lib/dpdk.c | 29 - vswi

[ovs-dev] [PATCH v2] OVS-DPDK: Change "dpdk-socket-mem" default value

Hi All, This change is teh outcome of this discussion: https://mail.openvswitch.org/pipermail/ovs-discuss/2017-April/044110.html. I am posting this on behalf of Marcin who is the author but no longer works for Intel. I made a few changes hence the double sign-off but please remove my sig if

[ovs-dev] [PATCH] OVS-DPDK: Change "dpdk-socket-mem" default value.

higher. Patch contains documentation update. Signed-off-by: Marcin Rybka <marcinx.ry...@intel.com> Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.rst | 3 ++- lib/dpdk.c | 35 ++

[ovs-dev] [PATCH] Change "dpdk-socket-mem" default value.

Hi All, This change is teh outcome of this discussion: https://mail.openvswitch.org/pipermail/ovs-discuss/2017-April/044110.html. I am posting this on behalf of Marcin who is the author but no longer works for Intel. I made a few changes hence the double sign-off but please remove my sig if

[ovs-dev] [RFC v2 2/2] ingress scheduling: Provide per interface ingress priority

from loss and latency as PMDs get overloaded. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 ++ lib/dpif-netdev.c | 103 +++- lib/netdev-bsd.c| 1 + lib/netdev-

[ovs-dev] [RFC v2 1/2] ingress scheduling: schema and docs

Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/howto/dpdk.rst | 18 ++ vswitchd/vswitch.ovsschema | 9 +++-- vswitchd/vswitch.xml | 40 3 files changed, 65 insertions(+), 2 deletions(-) diff

[ovs-dev] [RFC v2 0/2] Ingress Scheduling

base rather than in each netdev type. * Account for differing rxq lengths * Applies clean to 4299145 Billy O'Mahony (2): ingress scheduling: schema and docs ingress scheduling: Provide per interface ingress priority Documentation/howto/dpdk.rst| 18 +++ include/openvswitch/ofp

[ovs-dev] [RFC 2/2] ingress scheduling: Provide per interface ingress priority

from loss and latency as PMDs get overloaded. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 ++ lib/dpif-netdev.c | 47 +- lib/netdev-bsd.c| 1 + lib/netdev-dpdk.c

[ovs-dev] [RFC 1/2] ingress scheduling: schema and docs

Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/howto/dpdk.rst | 18 ++ vswitchd/vswitch.ovsschema | 9 +++-- vswitchd/vswitch.xml | 40 3 files changed, 65 insertions(+), 2 deletions(-) diff

[ovs-dev] [RFC 0/2] Ingress Scheduling

) Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz One pmd Port fwding port 0<->1, 2 <-> 3 Frame 64B, UDP 221 streams per port. OvS base - 4c80644 http://github.com/istokes/ovs dpdk_merge_2_9. Added 600cycles approx pkt processing in order to bring per packet cost to ~1000 cycles. DPDK v17

[ovs-dev] [RFC 1/2] dpif-netdev: Fix per packet cycles statistics.

From: Ilya Maximets DP_STAT_LOOKUP_HIT statistics used mistakenly for calculation of total number of packets. This leads to completely wrong per packet cycles statistics. For example: emc hits:0 megaflow hits:253702308 avg. subtable lookups per

[ovs-dev] [RFC 2/2] dpif-netdev: RFC EMC load-shedding

When EMC hit rate goes down start start shedding load from the EMC. --- lib/dpif-netdev.c | 107 -- 1 file changed, 103 insertions(+), 4 deletions(-) diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index e3a5590..f77e79a 100644 ---

[ovs-dev] [RFC 0/2] EMC load-shedding

r testing :). [1] https://mail.openvswitch.org/pipermail/ovs-dev/2017-August/336509.html [2] https://mail.openvswitch.org/pipermail/ovs-dev/2017-September/338380.html [3] https://mail.openvswitch.org/pipermail/ovs-dev/2017-August/337309.html Billy O'Mahony (1): dpif-netdev: RFC EMC load-shedding Ilya Max

[ovs-dev] [PATCH v2 4/4] docs: Document ingress scheduling feature

Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/howto/dpdk.rst | 31 +++ vswitchd/vswitch.xml | 31 +++ 2 files changed, 62 insertions(+) diff --git a/Documentation/howto/dpdk.rst b/Documentation

[ovs-dev] [PATCH v2 3/4] dpif-netdev: Add rxq prioritization

If an rxq is marked as 'prioritized' then keep reading from this queue until there are no packets available. Only then proceed to other queues. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/dpif-netdev.c | 24 ++-- 1 file changed, 18 insertions

[ovs-dev] [PATCH v2 2/4] netdev-dpdk: Apply ingress_sched config to dpdk phy ports

Ingress scheduling configuration is given effect by way of Flow Director filters. A small subset of the ingress scheduling possible is implemented in this patch. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 + lib/dpif-ne

[ovs-dev] [PATCH v2 1/4] netdev: Add set_ingress_sched to netdev api

Passes ingress_sched config item from other_config column of Interface table to the netdev. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/netdev-bsd.c | 1 + lib/netdev-dpdk.c | 21 + lib/netdev-dummy.c| 1 + lib/netdev-linux.c| 1

[ovs-dev] [PATCH 0/4] prioritizing latency sensitive traffic

Regards, Billy. billy O'Mahony (4): netdev: Add set_ingress_sched to netdev api netdev-dpdk: Apply ingress_sched config to dpdk phy ports dpif-netdev: Add rxq prioritization docs: Document ingress scheduling feature Documentation/howto/dpdk.rst| 31 +++ include/openvswitch/ofp

[ovs-dev] [PATCH 3/4] dpif-netdev: Add rxq prioritization

If an rxq is marked as 'prioritized' then keep reading from this queue until there are no packets available. Only then proceed to other queues. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/dpif-netdev.c | 24 ++-- 1 file changed, 18 insertions

[ovs-dev] [PATCH 4/4] docs: Document ingress scheduling feature

Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/howto/dpdk.rst | 31 +++ vswitchd/vswitch.xml | 31 +++ 2 files changed, 62 insertions(+) diff --git a/Documentation/howto/dpdk.rst b/Documentation

[ovs-dev] [PATCH 2/4] netdev-dpdk: Apply ingress_sched config to dpdk phy ports

Ingress scheduling configuration is given effect by way of Flow Director filters. A small subset of the ingress scheduling possible is implemented in this patch. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 + lib/dpif-ne

[ovs-dev] [RFC v2 4/4] docs: Document ingress scheduling feature

Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/howto/dpdk.rst | 31 +++ vswitchd/vswitch.xml | 31 +++ 2 files changed, 62 insertions(+) diff --git a/Documentation/howto/dpdk.rst b/Documentation

[ovs-dev] [RFC v2 3/4] dpif-netdev: Add rxq prioritization

If an rxq is marked as 'prioritized' then keep reading from this queue until there are no packets available. Only then proceed to other queues. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/dpif-netdev.c | 23 --- 1 file changed, 20 insertions

[ovs-dev] [RFC v2 2/4] netdev-dpdk: Apply ingress_sched config to dpdk phy ports

Ingress scheduling configuration is given effect by way of Flow Director filters. A small subset of the possible ingress scheduling possible is implemented in this patch. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 ++ lib/dpif-ne

[ovs-dev] [RFC v2 1/4] netdev: Add set_ingress_sched to netdev api

Passes ingress_sched config item from other_config column of Interface table to the netdev. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/netdev-bsd.c | 1 + lib/netdev-dpdk.c | 19 +++ lib/netdev-dummy.c| 1 + lib/netdev-linux.c| 1

[ovs-dev] [RFC v2 0/4] prioritizing latency sensitive traffic

it does not take into account: * reconfiguration - of the ingress_sched config of pmd masks and so on. * ovs-appctl dpif-netdev/pmd-*-show o/p should indicate rxq prioritization. * any requirements for a more granular prioritization mechanism * ... Regards, Billy. Billy O'Mahony (4): net

[ovs-dev] [PATCH v10] dpif-netdev: Assign ports to pmds on non-local numa node.

performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> Signed-off-by: Ilya Maximets <i.maxim...@samsung.com> Co-authored-by: Ilya Maximets <i.maxim...@samsung.com> Tested-by: Ian Stokes <ian.sto...@intel.com> Acked

[ovs-dev] [PATCH v9] dpif-netdev: Assign ports to pmds on non-local numa node.

performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> Signed-off-by: Ilya Maximets <i.maxim...@samsung.com> Co-authored-by: Ilya Maximets <i.maxim...@samsung.com> --- v9: v8 missed some comments on v7 v8: Some coding

[ovs-dev] [PATCH v8] dpif-netdev: Assign ports to pmds on non-local numa node.

performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> Signed-off-by: Ilya Maximets <i.maxim...@samsung.com> Co-authored-by: Ilya Maximets <i.maxim...@samsung.com> --- v8: Some coding style issues; doc tweak v7: Incorpora

[ovs-dev] [PATCH v7] dpif-netdev: Assign ports to pmds on non-local numa node.

performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> Signed-off-by: Ilya Maximets <i.maxim...@samsung.com> Co-authored-by: Ilya Maximets <i.maxim...@samsung.com> --- v7: Incorporate review comments on docs and impleme

[ovs-dev] [PATCH v7] dpif-netdev: Assign ports to pmds on non-local numa node.

I have incorporated Darrell's doc comments and Ilya's implementation and tested with various scenarios. I had to make a few small changes to ensure that the warning messages were issued for each queue as per previous patch comments. Billy O'Mahony (1): dpif-netdev: Assign ports to pmds on non

[ovs-dev] [RFC PATCH 2/3] netdev-dpdk: Apply ingress_sched config to dpdk phy ports

Ingress scheduling configuration is given effect by way of Flow Director filters. A small subset of the possible ingress scheduling possible is implemented in this patch. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- include/openvswitch/ofp-parse.h | 3 ++ lib/dpif-ne

[ovs-dev] [RFC PATCH 1/3] netdev: Add set_ingress_sched to netdev api

Passes ingress_sched config item from other_config column of Interface table to the netdev. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/netdev-bsd.c | 1 + lib/netdev-dpdk.c | 19 +++ lib/netdev-dummy.c| 1 + lib/netdev-linux.c| 1

[ovs-dev] [PATCH v6] dpif-netdev: Assign ports to pmds on non-local numa node.

a pmd on the 'wrong' numa node albeit with lower performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- v6: Change 'port' to 'queue' in a warning msg v5: Fix warning msg; Update same in docs v4: Fix a checkpatch error v3: Fix warn

[ovs-dev] [PATCH v5] dpif-netdev: Assign ports to pmds on non-local numa node.

a pmd on the 'wrong' numa node albeit with lower performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.rst | 10 + lib/dpif-netdev.c| 43 +++

[ovs-dev] [PATCH v4] dpif-netdev: Assign ports to pmds on non-local numa node.

a pmd on the 'wrong' numa node albeit with lower performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- v4: Fix a checkpatch error v3: Fix warning messages not appearing when using multiqueue v2: Add details of warning messag

[ovs-dev] [PATCH v3] dpif-netdev: Assign ports to pmds on non-local numa node.

a pmd on the 'wrong' numa node albeit with lower performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.rst | 10 + lib/dpif-netdev.c| 39 +++

[ovs-dev] [PATCH v2] dpif-netdev: Assign ports to pmds on non-local numa node.

a pmd on the 'wrong' numa node albeit with lower performance. Local pmds are still chosen when available. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- Documentation/intro/install/dpdk.rst | 10 ++ lib/dpif-netdev.c| 36 +++

[ovs-dev] [PATCH] dpif-netdev: emc comments

Add a concrete example of how a flow's hash determines the set of possible storage locations in the EMC. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/dpif-netdev.c | 24 +--- 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/lib/dpif-ne

[ovs-dev] [PATCH] netdev-dpdk: Enable INDIRECT_DESC on DPDK vHostUser.

This gives much better performance for linux apps in the guest without affecting dpdk applications in the guest. Test details in cover letter. Signed-off-by: Billy O'Mahony <billy.o.mah...@intel.com> --- lib/netdev-dpdk.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git

[ovs-dev] [PATCH] netdev-dpdk: Enable INDIRECT_DESC on DPDK vHostUser.

but before release), other versions not tested. DPDK v16.11 QEMU v2.5.0 (also seen with v2.7.1) Results: INDIRECT_DESC enabled2.75 Mpps @64B pkts (0.176 Gbit/s) INDIRECT_DESC disabled 2.75 Mpps @64B pkts (0.176 Gbit/s) Billy O'Mahony (1): netdev-dpdk: Enable INDIRECT_DESC