Cephalocon-seoul-2020-pc January 2020

cephalocon-seoul-2020-pc@ceph.io

22 participants
25 discussions

by Sage Weil

Hi everyone, I'm hoping we can resolve some of these questions over email and keep our discussion short. And I think we shoudl schedule a call as early monday as we can to try to finalize this. I'm sorry I couldn't do any of this last week--for the last week of my trip we only had a satellite link in the lounge area, and it was super flaky (e.g., went out when the clouds came in in the evening). I'm sending an invite.. let me know if this should be moved forward or back. Assuming we're still targetting 3 tracks (not 4, as the draft agenda Jen sent shows), then we have space for ~45 talks. If we accept the tutorial (highly rated at 4.6) then that would consume 2 or 3 slots, leaving 42-43 talk slots. I'm thinking 2 slots (1.5 hours) would be sufficient? So, 43 talks. Lightning talks would take 1 or 2 slots (probably 2), so 41 talks. If we follow a similar keynote format to barcelona, then we would have: - 1 opening keynote, what's new, etc. - 1 invited keynote - 4 platinum keynotes (15m each) - 4 gold keynotes (8m each) - CLT panel I think the simplest thing would be to pick one of the submitted talks and invite them to present in the keynote slot. The normal talks are 40m and the keynote is usually a bit shorter (30m), but we can probably be flexible. There are a few talks that have been tagged as candidates: - Matthew Leonard @ Bloomberg (we added 6 PB of storage and noone noticed) - Tom Byrne (50PB of ceph adventures). His talks are consistently good and very well received. - BY Chen (multi-pb scale storage using ceph in taiwan computing cloud) The other one that caught my eye is NAVER. Interesting because they are South Korean, and it's ceph + openstack + kubernetes (checks all boxes). I've never seen them present, though. I think Lars knows more about them? Anyway, assuming we pick one of those, then we'd have 41 additional slots to fill, for a total of 43 again. Right now that cutoff is around the average rating of 3.6. If we start breaking this down into tracks/topics, through row 44, ignoring the lightnight talks, then we have the below. I tagged a few of these as candidate for moving to CDS on Tuesday... I think that 200 is a definite, and 203 and 64 are maybes.. what do you all think? 64 especially seems like it warrants a realtime conversation and not a preso. 203 sounds like they went and implemented somethign so it may warrant a talk, but it makes me nervous since it's not upstream. 59 (deepika's observability talk) could also be during CDS but I think it might be nicer to showcase the work with a full talk.. which makes me think li's 203 should get one too? crimson: 110 crimson status (kefu/radek) 188 seastore (sam) 40 crimson-osd with alien from chunmei (i'd vote yes on this?) no: 51 crimson messenger no: 79 spdk one from intel user: 136 all nvme (at&t) 149 we added 6 pb of ceph (bloomberg) 177 50pb of ceph (STFC) 130 staying ahead of the curve (dan/CERN) 158 multi-pb at twcc (asus and suse) 55 object at scale, devops perspective (flipkart) 76 20k rbs volumes k8s on openstack (naver) 163 rgw road to production, security challenges at workday (workday) ops: 150 battling alert fatigue with prometheus (mnaser vexxhost) 31 ceph traps and pitfalls (xie zte) 23 1 year tuning ceph distilled (david byte suse) community: 88 ceph in china and how to make it bigger (intel) 111 tracking cephs in the wild (lmb) rook: 206 tutorial (msft + sk telecom) 118 rook best practices (blaine suse) 82 make storage scale with rook-ceph (seb red hat) 77 whatever can go wrong will (sagy red hat) dev: 120 ceph on windows [CDS] 200 rgw needs microserves (robin digitalocean) 90 optimizing ceph on arm (arm) [CDS?] 203 *store and persistent memory (igor and tushar) 59 reworking observability in ceph (deepika red hat) rbd: 38 disaster recovery [CDS?] 64 qos for rbd (liwang didi) rgw: 116 rgw multisite, ocotpus and beyond (yehuda and casey) 127 demystifying access mgmt (abhishek) 162 rgw with billions of objects (LINE) 115 automating data pipelines with ceph, knative, strimzi (red hat) 133 rgw capabilities and future plans (matt + uday/red hat) cephfs: 83 samba clustering improvements (david disseldorp suse) 104 schedule snapshots to async rep (jan, suse) rados: 100 bluestore best practices (vikyat red hat) 202 stretch clusters (greg red hat) 63 global dedup (myoungwon samsung) 151 disaggregating with nvmeof (zoltan ibm) misc: 197 orchestrator (swagner suse) 91 dashboard (lens suse) I think these ones are the mostly shoe-ins. If we move those 2 to CDS, and promote one user talk to a keynote, then we can accept 7 more. If we consolidate blaine and seb's rook talks (or seb and sagy's, maybe) then we can take 8 more. That's I think where it gets interesting since there are a lot of talks around the row 45-65 range (3.4-3.6 rating) where we should exercise editorial control... Listing them here: 187 gosbench (christopher blum red hat) 15 putting compute in your storage (federico red hat) 139 migrating to new hw without disruption (tyler brekke red hat) 152 fixing ceph 101 (mnaser vexxhost) 140 decoupling ceph (bluestore nvmf) (arun intel) 185 high perf cephfs for ebay loggying (xiaoxi ebay) 13 ceph security (danny DT) 92 ceph on public clouds (josh and orit, red hat) 94 demystifying perf on k8s (aakarsh, dustin red hat) 101 large scale ceph on arm (yaowei china mobile) 103 large scale mgmt (china mobile) [CDS?] 108 rbd snaps independent with cow/radix trees (roman suse) 97 teuthology status and future (kyr suse) 51 crimson messesnger (yingxin intel) 67 crash consistent client cache (lisa intel) 135 building storage ai/ml platform (sherard red hat) 142 turbocharging rook database workloads (alex calhoun red hat) 194 object bucket claims in rook (jiffin red hat) 28 ceph in the real world, what we learned (safespring) 25 ceph in ram (diamond light source) 78 cloud functions with knative (song LINE) If there are any talks below line 71 that anyone wants to nominate to rescue/consider, let's add them to this list? Skimmign the list I see the mulitiple upstream/downstream community talks, none of which made the cut. Maybe we can accept one or more of those as lightning talks? I'm ignoring the lightning talks for the moment. I think we can choose those quickly but just picking which ones (if any) we don't want. :) Doc links, in case you need them: https://docs.google.com/spreadsheets/d/1bYCepH57ZjhuE3WoFUNQLfNOL2h2ti6Kln0… https://docs.google.com/spreadsheets/d/1zSK6g7eadJ6cjT7dNLnO6yRseMLuKWNcMiP… sage

4 years, 3 months

Re: Cephalcon talk review!

by Lars Marowsky-Bree

On 2020-01-02T13:35:58, Matias Bjorling <Matias.Bjorling(a)wdc.com> wrote: Hi Matias, if I'm seeing this right, there are two "MB" columns in the voting spreadsheet. I'm not sure how that happened, but you're the only one with "MB" initials I think, at least according to the list of people Sage invited? Which column is right? Thanks, Lars -- Architect SDS, Distinguished Engineer SUSE Software Solutions Germany GmbH, MD: Felix Imendörffer, HRB 36809 (AG Nürnberg) "Architects should open possibilities and not determine everything." (Ueli Zbinden)

4 years, 3 months

Re: Cephalcon talk review!

by Dan van der Ster

It might also be useful if we had a rough outline of the session tracks before the review, so we can rank talks per subject. Do we have a rough sketch already? The timing seems fine to me -- I'll try to get my ranks in in the next days. Dan -----Original message----- > From: Sage Weil <sage(a)newdream.net> > Sent: Wednesday 18th December 2019 10:06 > To: cephalocon-seoul-2020-pc(a)ceph.io > Subject: [Cephalocon-seoul-2020-pc] Cephalcon talk review! > > Oh boy, here's the spreadsheet! > > https://docs.google.com/spreadsheets/d/1bYCepH57ZjhuE3WoFUNQLfNOL2h2ti6Kln0… > > I created reviewer columns like last time. Did we just rate from 0-5 > before? I can't remember. 0-5 seems reasonable? > > Currently the dates for this are: > > - speaker notifications Jan 13 > - schedule published Jan 17 > > I would love it if we could review these and put together a schedule > earlier. Sooner means we can publish a schedule sooner and people will > know what they are signing up for when they buy tickets and travel etc. > The challenges is the holidays. I'm going to be out Dec 26 - Jan 10, but > > I am going to try to do my reviews before that, and I can make time for a > > planning call or two the first week of Jan... or we can try to start the > discussion over email. If nothing else we can probably get the schedule > up by the 15th (W) instead of 17th (Fri) maybe? > > The other thing we need to start thinking about is keynotes. We should be > > on the looking for talk proposals that we can simply elevate to a keynote, > > or think about speakers that we could invite. > > Currently, the PC is > > - Dan (CERN) > - Josh (RH) > - Lars (SUSE) > - Matias (WD) > - Neha (RH) > - Patrick (RH) > - Sage (RH) > > I asked Florian to help again and he said no. Is there anyone else we > should invite? > > Thanks! > sage > _______________________________________________ > Cephalocon-seoul-2020-pc mailing list -- cephalocon-seoul-2020-pc(a)ceph.io > To unsubscribe send an email to cephalocon-seoul-2020-pc-leave(a)ceph.io >

4 years, 3 months

Cephalcon talk review!

by Sage Weil

Oh boy, here's the spreadsheet! https://docs.google.com/spreadsheets/d/1bYCepH57ZjhuE3WoFUNQLfNOL2h2ti6Kln0… I created reviewer columns like last time. Did we just rate from 0-5 before? I can't remember. 0-5 seems reasonable? Currently the dates for this are: - speaker notifications Jan 13 - schedule published Jan 17 I would love it if we could review these and put together a schedule earlier. Sooner means we can publish a schedule sooner and people will know what they are signing up for when they buy tickets and travel etc. The challenges is the holidays. I'm going to be out Dec 26 - Jan 10, but I am going to try to do my reviews before that, and I can make time for a planning call or two the first week of Jan... or we can try to start the discussion over email. If nothing else we can probably get the schedule up by the 15th (W) instead of 17th (Fri) maybe? The other thing we need to start thinking about is keynotes. We should be on the looking for talk proposals that we can simply elevate to a keynote, or think about speakers that we could invite. Currently, the PC is - Dan (CERN) - Josh (RH) - Lars (SUSE) - Matias (WD) - Neha (RH) - Patrick (RH) - Sage (RH) I asked Florian to help again and he said no. Is there anyone else we should invite? Thanks! sage

4 years, 3 months

Re: Cephalcon talk review!

by Dan van der Ster

Just finished my column -- I had to hide/show a couple times to paste in my scores properly, then ended up hiding all the columns which seemed filled. Apologies if I hid someone's almost complete column :-/ And I also have a conflict (my own submission), so I put a score of 'C' there too. Talk to you all soon, Dan -----Original message----- > From: Matias Bjorling <Matias.Bjorling(a)wdc.com> > Sent: Thursday 2nd January 2020 21:50 > To: Sage Weil <sage(a)newdream.net>; Dan van der Ster <daniel.vanderster(a)cern.ch> > Cc: cephalocon-seoul-2020-pc(a)ceph.io > Subject: [Cephalocon-seoul-2020-pc] Re: Cephalcon talk review! > > There is a lot of great presentations! > > We all have some relative "rating" value that we use - here is how my ratings > should be interpreted. > > 1 - No > 2 - Negative > 3 - Good - Looking for a tie-breaker > 4 - Positive - Great talk > 5 - Very Positive - Great talk and informational > > Please note that I have a conflict with Sam Just - I love to rate his talk, > unfortunately, I'm very biased (performance-geek, NVMe, ZNS, ...). I've put > the rating as C. Feel free to change it to an appropriate value. > > Happy Holidays! > > Kind regards, > Matias > > -----Original Message----- > From: Sage Weil <sage(a)newdream.net> > Sent: Friday, 20 December 2019 22.59 > To: Dan van der Ster <daniel.vanderster(a)cern.ch> > Cc: cephalocon-seoul-2020-pc(a)ceph.io > Subject: [Cephalocon-seoul-2020-pc] Re: Cephalcon talk review! > > We are currently planning 3 tracks. If we find we really want more talks, > then we could possibly go back to 4, but it'll cost more. Let's see where > we end up. > > As for how to organize the tracks... I'm not sure. Some of the general categories > I have in my head are: > > - crimson stuff > - roadmap stuff (new features we want to talk to users about) > - user talks (experiences runnign ceph at scale etc) > - developer-focused talks. some of these might be better suited for the Tuesday > developer summit if they are really just trying to start a conversation. > - ops stuff. kubernetes, rook, deployment tools, etc. > - tutorial.. there's a rook tutorial proposal (this time from Microsoft and > SK Telecom) ? > > I added a red column that is supposed to add up the scores, but I can't figure > out why =sum(...) isn't working. But maybe we don't need that yet until we > have the scores in. > > I hid my score column. I suggest after you finish entering your scores you > hide yours as well so that we have more independent scoring. > > I added a orange keyword column that has things like "user", "crimson", "rgw", > "cephfs", "tuning", etc. Feel free to update/adjust that.. I wasn't very > careful. I also was tempted to add a 'vendor' keyword for talks from vendors > (red hat, suse, intel mostly) but I couldn't tell if I being unfair; sometime > these talks come from tech sales people who don't have much hands-on experience > running real clusters. OTOH, there were a few talks from proserv folks who > have the associated premature gray hair to prove it and I wanted to hear from > them. Somehow they seemed like they'd be less vendor-ey, but not sure how > to draw the line. > > There were several duplciate submissions. I marked them DUP and put a strikeout > through them. > > I highlighted the submission type when it was not a normal talk (i.e., lightning, > panel, or tutorial) so that they were easy to spot. > > Happy reviewing! > sage > > > > On Wed, 18 Dec 2019, Dan van der Ster wrote: > > > It might also be useful if we had a rough outline of the session tracks > before the review, so we can rank talks per subject. > > Do we have a rough sketch already? > > > > The timing seems fine to me -- I'll try to get my ranks in in the next days. > > > > Dan > > > > -----Original message----- > > > From: Sage Weil <sage(a)newdream.net> > > > Sent: Wednesday 18th December 2019 10:06 > > > To: cephalocon-seoul-2020-pc(a)ceph.io > > > Subject: [Cephalocon-seoul-2020-pc] Cephalcon talk review! > > > > > > Oh boy, here's the spreadsheet! > > > > > > > > > https://docs.google.com/spreadsheets/d/1bYCepH57ZjhuE3WoFUNQLfNOL2h2 > > > ti6Kln0x_vWPrik/edit?usp=sharing > > > > > > I created reviewer columns like last time. Did we just rate from > > > 0-5 before? I can't remember. 0-5 seems reasonable? > > > > > > Currently the dates for this are: > > > > > > - speaker notifications Jan 13 > > > - schedule published Jan 17 > > > > > > I would love it if we could review these and put together a schedule > > > earlier. Sooner means we can publish a schedule sooner and people > > > will know what they are signing up for when they buy tickets and travel > etc. > > > The challenges is the holidays. I'm going to be out Dec 26 - Jan > > > 10, but > > > > > > I am going to try to do my reviews before that, and I can make time > > > for a > > > > > > planning call or two the first week of Jan... or we can try to start > > > the discussion over email. If nothing else we can probably get the > > > schedule up by the 15th (W) instead of 17th (Fri) maybe? > > > > > > The other thing we need to start thinking about is keynotes. We > > > should be > > > > > > on the looking for talk proposals that we can simply elevate to a > > > keynote, > > > > > > or think about speakers that we could invite. > > > > > > Currently, the PC is > > > > > > - Dan (CERN) > > > - Josh (RH) > > > - Lars (SUSE) > > > - Matias (WD) > > > - Neha (RH) > > > - Patrick (RH) > > > - Sage (RH) > > > > > > I asked Florian to help again and he said no. Is there anyone else > > > we should invite? > > > > > > Thanks! > > > sage > > > _______________________________________________ > > > Cephalocon-seoul-2020-pc mailing list -- > > > cephalocon-seoul-2020-pc(a)ceph.io To unsubscribe send an email to > > > cephalocon-seoul-2020-pc-leave(a)ceph.io > > > > > _______________________________________________ > > Cephalocon-seoul-2020-pc mailing list -- > > cephalocon-seoul-2020-pc(a)ceph.io To unsubscribe send an email to > > cephalocon-seoul-2020-pc-leave(a)ceph.io > > > > > _______________________________________________ > Cephalocon-seoul-2020-pc mailing list -- cephalocon-seoul-2020-pc(a)ceph.io > To unsubscribe send an email to cephalocon-seoul-2020-pc-leave(a)ceph.io > _______________________________________________ > Cephalocon-seoul-2020-pc mailing list -- cephalocon-seoul-2020-pc(a)ceph.io > To unsubscribe send an email to cephalocon-seoul-2020-pc-leave(a)ceph.io >

4 years, 3 months

2024

2023

2022

2021

2020

2019

Cephalocon-seoul-2020-pc January 2020