From carrier at cray.com Wed Jun 6 19:42:46 2012 From: carrier at cray.com (John Carrier) Date: Wed, 6 Jun 2012 19:42:46 +0000 Subject: [Twg] no TWG concall 6/7/2012 Message-ID: Dave is on vacation and I'm on the road. We will not have a TWG meeting tomorrow. --jc -------------- next part -------------- An HTML attachment was scrubbed... URL: From carrier at cray.com Thu Jun 7 16:04:37 2012 From: carrier at cray.com (John Carrier) Date: Thu, 7 Jun 2012 16:04:37 +0000 Subject: [Twg] twg requirements doc Message-ID: I have updated the requirements doc with some minor edits. Dave and I received initial feedback from some of the board members who asked that we match the order of the features in the gathered requirements section of the document with the order of the categories and requirements in our recommendation table. I did not add or delete any text. I just changed the order of items in the description section. Let me know if you have any questions or comments. --jc -------------- next part -------------- An HTML attachment was scrubbed... URL: From carrier at cray.com Thu Jun 14 14:26:26 2012 From: carrier at cray.com (John Carrier) Date: Thu, 14 Jun 2012 14:26:26 +0000 Subject: [Twg] no TWG today (6/14/2012) Message-ID: Hi all, We will not have a TWG meeting today. The Board has not yet made a decision on our recommendation concerning the 2012 requirements we submitted. There have been a couple of questions raised by Board members. The issues and our proposed resolution are below. 1) Order of requirements in the "gathered requirements" section As I mentioned in a previous email, I rearranged the categories and requirements within each category to match the order described in our "recommendations" section. 2) 2011 Required Rates and Capacities The table was created as part of our 2011 requirements investigation. I saved it as an appendix so that work wouldn't be lost. Some of the Board members want us to update the table to reflect current limits, since some of the data there is no longer accurate. For example, the Lustre 2.2 column was a best guess in 2011 and could be updated with real data now that 2.2 is released. Do we need to spend the time updating the table? I proposed to the board that we delete the table from the appendix and revise it as part of next year's requirements document. What do you think? 3) Lack of explicit recommendations Last year there was broad agreement that metadata performance was the single category requiring the greatest effort. This year there were several categories fairly equally represented in the members' top 10 picks. Our recommendation was to have OpenSFS initiate projects in each of the top four categories so that we can make incremental progress on all fronts this year rather than choosing just one category and waiting until next year to address the other categories. However, if the Board prefers to focus on just one category, then "File system availability and robustness" was the very clear winner this year, as the voting spreadsheet shows (http://goo.gl/Lqg7s). Some of the LNET work could be embodied under this umbrella since LNET contributes to availability and robustness of Lustre. I have asked the Board to clarify if they require this refinement. The current version of the doc is on google docs (http://goo.gl/63u9Q). Please send feedback to the lists. --jc -------------- next part -------------- An HTML attachment was scrubbed... URL: From spitzcor at cray.com Thu Jun 14 15:09:30 2012 From: spitzcor at cray.com (Cory Spitz) Date: Thu, 14 Jun 2012 10:09:30 -0500 Subject: [Twg] [Discuss] no TWG today (6/14/2012) In-Reply-To: References: Message-ID: <4FD9FEAA.5000000@cray.com> John, I think that we should keep the requirements table up to date. It shouldn't be much work. It just might mean that we just update the 2.2 column with a few changes and delete the Q2 2012 column altogether. I think most of the 2.2 column is correct, but the maximum file size should be 31.25 PB since the new max stripe count is 2000. We should also add a row for maximum # of MDTs too. Thanks, -Cory On 06/14/2012 09:26 AM, John Carrier wrote: > Hi all, > > > > We will not have a TWG meeting today. The Board has not yet made a > decision on our recommendation concerning the 2012 requirements we > submitted. > > > > There have been a couple of questions raised by Board members. The > issues and our proposed resolution are below. > > > > 1) Order of requirements in the “gathered requirements” section > As I mentioned in a previous email, I rearranged the categories and > requirements within each category to match the order described in our > “recommendations” section. > > 2) 2011 Required Rates and Capacities > The table was created as part of our 2011 requirements investigation. I > saved it as an appendix so that work wouldn’t be lost. Some of the > Board members want us to update the table to reflect current limits, > since some of the data there is no longer accurate. For example, the > Lustre 2.2 column was a best guess in 2011 and could be updated with > real data now that 2.2 is released. > > Do we need to spend the time updating the table? I proposed to the > board that we delete the table from the appendix and revise it as part > of next year’s requirements document. What do you think? > > 3) Lack of explicit recommendations > Last year there was broad agreement that metadata performance was the > single category requiring the greatest effort. This year there were > several categories fairly equally represented in the members’ top 10 > picks. Our recommendation was to have OpenSFS initiate projects in each > of the top four categories so that we can make incremental progress on > all fronts this year rather than choosing just one category and waiting > until next year to address the other categories. > > However, if the Board prefers to focus on just one category, then "File > system availability and robustness" was the very clear winner this year, > as the voting spreadsheet shows (http://goo.gl/Lqg7s). Some of the LNET > work could be embodied under this umbrella since LNET contributes to > availability and robustness of Lustre. I have asked the Board to > clarify if they require this refinement. > > The current version of the doc is on google docs > (http://goo.gl/63u9Q). Please send feedback to the lists. > > > > --jc > > > > > > _______________________________________________ > discuss mailing list > discuss at lists.opensfs.org > http://lists.opensfs.org/listinfo.cgi/discuss-opensfs.org From carrier at cray.com Thu Jun 14 15:19:40 2012 From: carrier at cray.com (John Carrier) Date: Thu, 14 Jun 2012 15:19:40 +0000 Subject: [Twg] [Discuss] no TWG today (6/14/2012) In-Reply-To: <4FD9FEAA.5000000@cray.com> References: <4FD9FEAA.5000000@cray.com> Message-ID: If we follow Cory's suggestion, what else should we update in the appendix table? -----Original Message----- From: Cory Spitz [mailto:spitzcor at cray.com] Sent: Thursday, June 14, 2012 8:10 AM To: John Carrier Cc: twg at lists.opensfs.org; discuss at lists.opensfs.org Subject: Re: [Discuss] no TWG today (6/14/2012) John, I think that we should keep the requirements table up to date. It shouldn't be much work. It just might mean that we just update the 2.2 column with a few changes and delete the Q2 2012 column altogether. I think most of the 2.2 column is correct, but the maximum file size should be 31.25 PB since the new max stripe count is 2000. We should also add a row for maximum # of MDTs too. Thanks, -Cory On 06/14/2012 09:26 AM, John Carrier wrote: > Hi all, > > > > We will not have a TWG meeting today. The Board has not yet made a > decision on our recommendation concerning the 2012 requirements we > submitted. > > > > There have been a couple of questions raised by Board members. The > issues and our proposed resolution are below. > > > > 1) Order of requirements in the "gathered requirements" section > As I mentioned in a previous email, I rearranged the categories and > requirements within each category to match the order described in our > "recommendations" section. > > 2) 2011 Required Rates and Capacities > The table was created as part of our 2011 requirements investigation. I > saved it as an appendix so that work wouldn't be lost. Some of the > Board members want us to update the table to reflect current limits, > since some of the data there is no longer accurate. For example, the > Lustre 2.2 column was a best guess in 2011 and could be updated with > real data now that 2.2 is released. > > Do we need to spend the time updating the table? I proposed to the > board that we delete the table from the appendix and revise it as part > of next year's requirements document. What do you think? > > 3) Lack of explicit recommendations > Last year there was broad agreement that metadata performance was the > single category requiring the greatest effort. This year there were > several categories fairly equally represented in the members' top 10 > picks. Our recommendation was to have OpenSFS initiate projects in each > of the top four categories so that we can make incremental progress on > all fronts this year rather than choosing just one category and waiting > until next year to address the other categories. > > However, if the Board prefers to focus on just one category, then "File > system availability and robustness" was the very clear winner this year, > as the voting spreadsheet shows (http://goo.gl/Lqg7s). Some of the LNET > work could be embodied under this umbrella since LNET contributes to > availability and robustness of Lustre. I have asked the Board to > clarify if they require this refinement. > > The current version of the doc is on google docs > (http://goo.gl/63u9Q). Please send feedback to the lists. > > > > --jc > > > > > > _______________________________________________ > discuss mailing list > discuss at lists.opensfs.org > http://lists.opensfs.org/listinfo.cgi/discuss-opensfs.org From adilger at whamcloud.com Thu Jun 14 16:04:21 2012 From: adilger at whamcloud.com (Andreas Dilger) Date: Thu, 14 Jun 2012 10:04:21 -0600 Subject: [Twg] [Discuss] no TWG today (6/14/2012) In-Reply-To: References: Message-ID: <9772D829-88E7-4DE6-B96F-A15CDFD3CF6B@whamcloud.com> On 2012-06-14, at 8:26 AM, John Carrier wrote: > 2) 2011 Required Rates and Capacities > The table was created as part of our 2011 requirements investigation. I saved it as an appendix so that work wouldn’t be lost. Some of the Board members want us to update the table to reflect current limits, since some of the data there is no longer accurate. For example, the Lustre 2.2 column was a best guess in 2011 and could be updated with real data now that 2.2 is released. Actually, I added the 2.2 table after 2.2 was released. It wasn't in the original 2011 version of the table. > 3) Lack of explicit recommendations > Last year there was broad agreement that metadata performance was the single category requiring the greatest effort. This year there were several categories fairly equally represented in the members’ top 10 picks. Our recommendation was to have OpenSFS initiate projects in each of the top four categories so that we can make incremental progress on all fronts this year rather than choosing just one category and waiting until next year to address the other categories. > > However, if the Board prefers to focus on just one category, then "File system availability and robustness" was the very clear winner this year, as the voting spreadsheet shows (http://goo.gl/Lqg7s). Some of the LNET work could be embodied under this umbrella since LNET contributes to availability and robustness of Lustre. I have asked the Board to clarify if they require this refinement. Having only a single option for development will unnecessarily restrict the features that could be developed. Last year there were a number of different projects that were worked on, in different (though related) areas of the code (client for statahead, ldiskfs/libcfs/ptlrpc for SMP scaling, and MDS for DNE). I would recommend that at least the #3-ish feature (#1 if going by total votes) "OST migration/rebalancing" be included in the recommendation to the board. This was the only feature on every board member's list, including the community seat. Cheers, Andreas -- Andreas Dilger Whamcloud, Inc. Principal Lustre Engineer http://www.whamcloud.com/ From carrier at cray.com Thu Jun 21 15:02:55 2012 From: carrier at cray.com (John Carrier) Date: Thu, 21 Jun 2012 15:02:55 +0000 Subject: [Twg] no TWG concall today (6/21/2012) Message-ID: Dave and I haven't received any further feedback from the Board, many of whom were at ISC this week. Hence, we have no actions or agenda topics for today's TWG call. --jc -------------- next part -------------- An HTML attachment was scrubbed... URL: From carrier at cray.com Thu Jun 28 14:25:23 2012 From: carrier at cray.com (John Carrier) Date: Thu, 28 Jun 2012 14:25:23 +0000 Subject: [Twg] no TWG concall today Message-ID: We still haven't heard from the Board and haven't received any agenda topics. So I am canceling today's TWG concall (6/28). We'll also cancel next week's call (7/5) since many in the US may be extending the July 4th holiday. Please send discussion topics to the list! Thanks, --jc -------------- next part -------------- An HTML attachment was scrubbed... URL: