From: Tariq Toukan <tariqt@mellanox.com>
To: Jonathan Lemon <jonathan.lemon@gmail.com>,
Saeed Mahameed <saeedm@mellanox.com>
Cc: "ilias.apalodimas@linaro.org" <ilias.apalodimas@linaro.org>,
Tariq Toukan <tariqt@mellanox.com>,
"brouer@redhat.com" <brouer@redhat.com>,
Netdev <netdev@vger.kernel.org>, kernel-team <kernel-team@fb.com>
Subject: Re: [PATCH 01/10 net-next] net/mlx5e: RX, Remove RX page-cache
Date: Sun, 20 Oct 2019 07:29:07 +0000 [thread overview]
Message-ID: <b73c8c15-3f10-d539-f648-5a0c772d9fc0@mellanox.com> (raw)
In-Reply-To: <7C9F38DB-6164-4ACB-A717-1699ACC9DCB0@gmail.com>
On 10/19/2019 3:10 AM, Jonathan Lemon wrote:
> I was running the updated patches on machines with various workloads, and
> have a bunch of different results.
>
> For the following numbers,
> Effective = hit / (hit + empty + stall) * 100
>
> In other words, show the hit rate for for every trip to the cache,
> and the cache full stat is ignored.
>
> On a webserver:
>
> [web] # ./eff
> ('rx_pool_cache_hit:', '360127643')
> ('rx_pool_cache_full:', '0')
> ('rx_pool_cache_empty:', '6455735977')
> ('rx_pool_ring_produce:', '474958')
> ('rx_pool_ring_consume:', '0')
> ('rx_pool_ring_return:', '474958')
> ('rx_pool_flush:', '144')
> ('rx_pool_node_change:', '0')
> cache effectiveness: 5.28
>
> On a proxygen:
> # ethtool -S eth0 | grep rx_pool
> rx_pool_cache_hit: 1646798
> rx_pool_cache_full: 0
> rx_pool_cache_empty: 15723566
> rx_pool_ring_produce: 474958
> rx_pool_ring_consume: 0
> rx_pool_ring_return: 474958
> rx_pool_flush: 144
> rx_pool_node_change: 0
> cache effectiveness: 9.48
>
> On both of these, only pages with refcount = 1 are being kept.
>
>
> I changed things around in the page pool so:
>
> 1) the cache behaves like a ring instead of a stack, this
> sacrifices temporal locality.
>
> 2) it caches all pages returned regardless of refcount, but
> only returns pages with refcount=1.
>
> This is the same behavior as the mlx5 cache. Some gains
> would come about if the sojourn time though the cache is
> greater than the lifetime of the page usage by the networking
> stack, as it provides a fixed working set of mapped pages.
>
> On the web server, this is a net loss:
> [web] # ./eff
> ('rx_pool_cache_hit:', '6052662')
> ('rx_pool_cache_full:', '156355415')
> ('rx_pool_cache_empty:', '409600')
> ('rx_pool_cache_stall:', '302787473')
> ('rx_pool_ring_produce:', '156633847')
> ('rx_pool_ring_consume:', '9925520')
> ('rx_pool_ring_return:', '278788')
> ('rx_pool_flush:', '96')
> ('rx_pool_node_change:', '0')
> cache effectiveness: 1.95720846778
>
> For proxygen on the other hand, it's a win:
> [proxy] # ./eff
> ('rx_pool_cache_hit:', '69235177')
> ('rx_pool_cache_full:', '35404387')
> ('rx_pool_cache_empty:', '460800')
> ('rx_pool_cache_stall:', '42932530')
> ('rx_pool_ring_produce:', '35717618')
> ('rx_pool_ring_consume:', '27879469')
> ('rx_pool_ring_return:', '404800')
> ('rx_pool_flush:', '108')
> ('rx_pool_node_change:', '0')
> cache effectiveness: 61.4721608624
>
> So the correct behavior isn't quite clear cut here - caching a
> working set of mapped pages is beneficial in spite of the HOL
> blocking stalls for some workloads, but I'm sure that it wouldn't
> be too difficult to exceed the WS size.
>
> Thoughts?
>
We have a WIP in which we avoid the HOL block, by having pages returned
to the available-queue only when their refcnt reaches back to 1. This
requires catching this case in the page/skb release path.
See:
https://github.com/xdp-project/xdp-project/tree/master/areas/mem
next prev parent reply other threads:[~2019-10-20 7:37 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-16 22:50 [PATCH 00/10 net-next] page_pool cleanups Jonathan Lemon
2019-10-16 22:50 ` [PATCH 01/10 net-next] net/mlx5e: RX, Remove RX page-cache Jonathan Lemon
2019-10-17 1:30 ` Jakub Kicinski
2019-10-18 20:03 ` Jonathan Lemon
2019-10-18 20:53 ` Saeed Mahameed
2019-10-19 0:10 ` Jonathan Lemon
2019-10-20 7:29 ` Tariq Toukan [this message]
2019-10-16 22:50 ` [PATCH 02/10 net-next] net/mlx5e: RX, Manage RX pages only via page pool API Jonathan Lemon
2019-10-16 22:50 ` [PATCH 03/10 net-next] net/mlx5e: RX, Internal DMA mapping in page_pool Jonathan Lemon
2019-10-17 1:33 ` Jakub Kicinski
2019-10-18 20:04 ` Jonathan Lemon
2019-10-18 21:01 ` Saeed Mahameed
2019-10-16 22:50 ` [PATCH 04/10 net-next] page_pool: Add API to update numa node and flush page caches Jonathan Lemon
2019-10-17 12:06 ` Ilias Apalodimas
2019-10-18 21:07 ` Saeed Mahameed
2019-10-18 23:38 ` Jonathan Lemon
2019-10-16 22:50 ` [PATCH 05/10 net-next] net/mlx5e: Rx, Update page pool numa node when changed Jonathan Lemon
2019-10-16 22:50 ` [PATCH 06/10 net-next] page_pool: Add page_pool_keep_page Jonathan Lemon
2019-10-16 22:50 ` [PATCH 07/10 net-next] page_pool: allow configurable linear cache size Jonathan Lemon
2019-10-17 8:51 ` Jesper Dangaard Brouer
2019-10-18 20:31 ` Jonathan Lemon
2019-10-16 22:50 ` [PATCH 08/10 net-next] page_pool: Add statistics Jonathan Lemon
2019-10-17 8:29 ` Jesper Dangaard Brouer
2019-10-18 21:29 ` Saeed Mahameed
2019-10-18 23:37 ` Jonathan Lemon
2019-10-16 22:50 ` [PATCH 09/10 net-next] net/mlx5: Add page_pool stats to the Mellanox driver Jonathan Lemon
2019-10-17 11:09 ` Jesper Dangaard Brouer
2019-10-18 20:45 ` Jonathan Lemon
2019-10-16 22:50 ` [PATCH 10/10 net-next] page_pool: Cleanup and rename page_pool functions Jonathan Lemon
2019-10-18 20:50 ` [PATCH 00/10 net-next] page_pool cleanups Saeed Mahameed
2019-10-18 21:21 ` Saeed Mahameed
2019-10-18 23:32 ` Jonathan Lemon
2019-10-21 19:08 ` Saeed Mahameed
2019-10-21 21:45 ` Jonathan Lemon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=b73c8c15-3f10-d539-f648-5a0c772d9fc0@mellanox.com \
--to=tariqt@mellanox.com \
--cc=brouer@redhat.com \
--cc=ilias.apalodimas@linaro.org \
--cc=jonathan.lemon@gmail.com \
--cc=kernel-team@fb.com \
--cc=netdev@vger.kernel.org \
--cc=saeedm@mellanox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).