Wednesday, 2016-01-13

*** eranrom has joined #openstack-storlets07:02
*** [1]eranrom has joined #openstack-storlets08:58
*** eranrom has quit IRC08:58
*** [1]eranrom is now known as eranrom08:58
kota_fyi, i will be absolute from meeting today, sorry.09:52
kota_I have a dinner plan with my friends09:53
*** takashi has joined #openstack-storlets12:07
*** takashi has quit IRC12:17
eranromkota_,: Have fun :)12:19
*** takashi has joined #openstack-storlets12:20
*** takashi has quit IRC12:29
*** takashi has joined #openstack-storlets12:34
eranromHi, anybody here?13:00
takashiHi :-)13:01
eranromTakashi: Hi. So I finally got a chance to go over your patches :-)13:02
eranromThanks very much for them and apologies for the late review.13:02
takashieranrom: Thanks a lot for your review! now I have to solve one conflict and after that we can totally merge patches about our first unittest13:03
takashieranrom: np13:03
eranromgreat. i will try to be more responsive this time :-)13:03
eranromDo yuo have anytyhing in mind planned?13:04
eranromI mean after those patches?13:04
takashieranrom: I have some small things like refactoring, but I think I cannot do that so soon13:05
eranromok, sure13:05
takashiI'm going to start our performance testing about storlets from this week, so I think I cannot spend so much time for that soon13:06
eranromoh, I will be VERY interested to see the results :-)13:06
eranromso far we had a chance to do very little performance testing and we found out the following:13:07
eranrom1. The storlet daemon thread pool size is crucial when it comes to mabny concurrent requests.13:07
eranrom2. The size of chunks in which the storlet reads the data from the input stream / write to the output stream is also significant. This is a classical latency / memory trade off...13:08
eranroms/mabny/many13:09
takashiThese are very useful information! I will take care about them in my testing.13:10
takashiI'm going to start with very basic situation, and then change some parameters to see a tendency about the performance of storlets13:13
eranromOne more thing. If you plan on using the identity storlet for this, then it is written in a most non-performant way. For each chunk of data it reads it transforms it into a string replacing \0 with something and logs the data13:13
eranromWe have modified it for performance tests and saw that there is no different between GET Vs. GET with X-Run-Storlet: identitystorlet-1.0.jar13:15
eranromSounds good.13:15
takashieranrom: interesting results13:15
takashiI'm going to use my image processing storlets for testing, like grayscaling, feature detection and so on to test storlets in a something like a more PoC-like way.13:17
eranromalright, will be interesting to get the results :-) That is if you will be able to share.13:17
eranromBTW - are you planning on coming to the Hackathon?13:18
takashieranrom: yes, I and kota_ will be there, and also other two members working about changing policy stuff will.13:19
eranromgreat to hear!13:19
eranromthat is to read :-)13:19
takashiI'm now preparing for my buisines trip to Bristol! :-)13:20
eranromyeah, I need to also start to get prepared.13:20
takashieranrom: I think we can share some results in our PoC, and I hope we can propose what to improve we find in it.13:22
eranromThat would be great! Looking forward.13:22
takashieranrom: I wrote some items about resource limitations in wiki as just ideas, but I think we can find if they are really needed throug that PoC.13:23
takashiI'm meaning about discussion points in hackathon, written here. https://wiki.openstack.org/wiki/Meetings/Storlets13:24
eranromI absolutely agree. Perhaps we will be able to design a solution during the Hackathon. I do not think that this is something the whole community will be interested in.13:25
takashieranrom: yes13:26
takashiI think the following two topics are what we have to discuss in swift, gate job stuff and fixes in swift13:28
takashibecause they need some help and understanding from swift community.13:30
takashiOf cause other topics are also interesting for me :-)13:30
eranromAgree. BTW - You may want to try and add pleuse as a reviewer to the patch. I do not know how active he is lately but he is the most supportive core to storlets so far...13:31
takashiOK. I'm going to ask him. and again I'm also going to ask kota_13:32
eranromgreat.13:33
eranromanything else for today?13:34
takashione thing for Austin13:34
eranromsure13:34
takashias I wrote in irc, I modified abstruct a little13:35
takashieranrom: thank you for your preparing base idea13:35
eranromsure, yes I saw the change. thanks. In IBM we need to pre-submit talks we want to give to get an approval. So I posted the abstract after your edit.13:36
takashiI'll ask again kota_ about this (because it incluse NTT), but I think we can post that abstract if we have no problem about it.13:36
eranromBTW - it is always possible to change it before submitting to Openstack13:36
takashis/incluse/includes13:36
takashieranrom: I see13:37
eranromThis is an internal IBM thing, and few poeple are seeing this so no worries.13:37
eranromI will not post anything to openstack before I get your approval.13:37
takashiwe have also same kind of things in NTT :-)13:38
eranromI am not surprized :-)\13:38
eranromHas the process started yet?13:38
takashieranrom: I think we will soon start, but we already asked our managers about this.13:39
eranromok.13:39
takashiThat's all for me. I'll inform about our updates if we have.13:41
eranromok, great. thanks!13:42
eranromhave a good night :-)13:44
takashieranrom: have a good day :-)13:44
eranrom:-)13:48
*** takashi has quit IRC13:56
*** eranrom has quit IRC16:14
*** eranrom has joined #openstack-storlets19:25
*** eranrom has quit IRC20:03

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!