From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============5033661671614692360==" MIME-Version: 1.0 From: Jonas Pfefferle Subject: Re: [SPDK] Controller Memory Buffer Date: Sun, 11 Mar 2018 12:31:59 +0100 Message-ID: In-Reply-To: D37E9B71-5724-4A84-A2F2-9B9D2D11129D@raithlin.com List-ID: To: spdk@lists.01.org --===============5033661671614692360== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Hi Stephen, Thanks for the hints. Do you happen to know which Intel drives these are? I = have a few Intel drives: DC P3600/3700 and 900p/P4800X but neither of those = devices have CMB. I'm experimenting with Mellanox ConnectX-5 target offloading so it would be = nice to see the effects of having SQ/CQ and data staging on the device. Fro= m = the experiments I've done so far it looks like the latencies are higher wit= h = the offloaded target compared to SPDK or the Kernel target. I assume this i= s = partially due to the additional PCIe round trip times. Thanks & Cheers, Jonas On Sat, 10 Mar 2018 21:11:01 +0000 "Stephen Bates" wrote: > Jonas > = > There are some Intel drives that include a small (a few MB) CMB that = >supports SQS only. Both Eideticom and Everspin have NVMe devices that = >have CMBs that support the WDS and RDS modes as well as SQS. > = > I know that other vendors are pretty actively looking at this so I=E2=80= =99d = >expect more NVMe SSDs with a variety of CMB modes later this year. = >I=E2=80=99d suggest pinging those companies and getting roadmap updates fr= om = >them to understand their timelines. > = > Cheers > = > Stephen > = >> On Mar 9, 2018, at 9:15 AM, Jonas Pfefferle = >>wrote: >> = >> Hi @all, >> = >> Is there any NVMe drive available which supports the optional = >>"Controller Memory Buffer" feature introduced in NVMe 1.2? >> = >> Thanks & Regards, >> Jonas >> = >> _______________________________________________ >> SPDK mailing list >> SPDK(a)lists.01.org >> https://lists.01.org/mailman/listinfo/spdk > _______________________________________________ > SPDK mailing list > SPDK(a)lists.01.org > https://lists.01.org/mailman/listinfo/spdk --===============5033661671614692360==--