From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============2771342984439874508==" MIME-Version: 1.0 From: Walker, Benjamin Subject: [SPDK] Re: SPDK socket abstraction layer Date: Tue, 05 Nov 2019 18:08:26 +0000 Message-ID: <36cd5c96-700c-4fe0-999e-f04ae52891bb@email.android.com> In-Reply-To: 20191105052925.2696.40119@new-ml01.vlan13.01.org List-ID: To: spdk@lists.01.org --===============2771342984439874508== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Can you outline your system set up so I can reproduce this? On my system cu= rrently the network stack is electing to do a deferred copy so the performa= nce is not good. I'd love to reproduce your exact set up. I'm not surprised there are bugs at higher queue depth. The patch is defini= tely still a work in progress. On Nov 4, 2019 9:26 PM, allenz(a)mellanox.com wrote: Hi, With suggestion of Sasha, we synced the codes to patch (https://review.gerr= ithub.io/c/spdk/spdk/+/471752). We run tests on two X86-64 servers connecte= d with Mellanox CX-5 100G. For Perf with 16 QD/4K IO, we found 14% improvement with the zero-copy patc= h when 1 or 2 cores were used, and 6% improvement when more cores (e.g., 8)= were used. Unfortunately, when we tried to use queue depth more than 16, and bigger I= O than 4K, Perf hung or got CQ error. But without the zero-copy patch, 64QD= or 64K IO was ok. Best regards, Allen _______________________________________________ SPDK mailing list -- spdk(a)lists.01.org To unsubscribe send an email to spdk-leave(a)lists.01.org --===============2771342984439874508==--