From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andreas Bluemle Subject: Re: FileStore performance: coalescing operations Date: Thu, 19 Mar 2015 15:59:02 +0100 Message-ID: <20150319155902.70551366@doppio> References: <20150226152807.5b71a93c@doppio> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: Received: from mail.itxperts.de ([212.202.108.166]:20599 "EHLO mail.itxperts.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756235AbbCSO7K (ORCPT ); Thu, 19 Mar 2015 10:59:10 -0400 In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Sage Weil Cc: Ceph Development Hi Sage, as discussed yesterday during the performance meeting, I have integratged pull reuest 3878 into v0.93 source code and re-tested: I can confirm that the changes show a significant benefit in the CPU consumption of the ceph-osd daemon for a given workload: the reduction in CPU cosumption for the FileStore workers is 22% and 10% for the overall CPU consumption of the ceph-osd daemon. Regards Andreas Bluemle On Thu, 5 Mar 2015 15:04:41 +0800 Haomai Wang wrote: > I think the performance improvement can be refer to > https://github.com/ceph/ceph/pull/2972 which I did a simple benchmark > comparison. > > This coalescing should get a lighter improvement but I think obviously > it should be better. > > On Thu, Mar 5, 2015 at 8:10 AM, Sage Weil wrote: > > On Tue, 3 Mar 2015, Sage Weil wrote: > >> I think we should try to avoid the dups in the first place in the > >> upper layers before resorting to deduping in the FileStore. > > > > I took a stab at this: > > > > https://github.com/ceph/ceph/pull/3878 > > > > Here's what the txns look like now: > > > > { > > "ops": [ > > { > > "op_num": 0, > > "op_name": "omap_setkeys", > > "collection": "0.2_head", > > "oid": "2\/\/head\/\/0", > > "attr_lens": { > > "0000000005.00000000000000000002": 178, > > "_epoch": 4, > > "_info": 729, > > "can_rollback_to": 12, > > "rollback_info_trimmed_to": 12 > > } > > }, > > { > > "op_num": 1, > > "op_name": "write", > > "collection": "0.2_head", > > "oid": > > "8b08fa52\/benchmark_data_maetl_15609_object5\/head\/\/0", > > "length": 123, "offset": 0, > > "bufferlist length": 123 > > }, > > { > > "op_num": 2, > > "op_name": "setattr", > > "collection": "0.2_head", > > "oid": > > "8b08fa52\/benchmark_data_maetl_15609_object5\/head\/\/0", "name": > > "_", "length": 269 > > }, > > { > > "op_num": 3, > > "op_name": "setattr", > > "collection": "0.2_head", > > "oid": > > "8b08fa52\/benchmark_data_maetl_15609_object5\/head\/\/0", "name": > > "snapset", "length": 31 > > } > > ] > > } > > > > Care to give that a spin and see if it give syou a similar (or > > better) improvement? > > > > sage > > -- > > To unsubscribe from this list: send the line "unsubscribe > > ceph-devel" in the body of a message to majordomo@vger.kernel.org > > More majordomo info at http://vger.kernel.org/majordomo-info.html > > > -- Andreas Bluemle mailto:Andreas.Bluemle@itxperts.de ITXperts GmbH http://www.itxperts.de Balanstrasse 73, Geb. 08 Phone: (+49) 89 89044917 D-81541 Muenchen (Germany) Fax: (+49) 89 89044910 Company details: http://www.itxperts.de/imprint.htm