Project

General

Profile

Edit History

Wiki » Planning » CDS Dumpling » Chat Logs »

-1C - Erasure Encoding as a Storage Backend » History » Version 1

Jessica Mack, 06/22/2015 05:44 AM

1 1 Jessica Mack 
h1. -1C - Erasure Encoding as a Storage Backend
2 





    3
    
    
    
<pre>





    4
    
    
    
xiaoxi	I cannot see anyone & any words,it's all blank	09:00





    5
    
    
    
loicd	sjust: \o	09:00





    6
    
    
    
loicd	that's what the summit is for, right ? ;-)	09:01





    7
    
    
    
*** rturk is now known as rturk-away	09:01





    8
    
    
    
*** rturk-away is now known as rturk	09:01





    9
    
    
    
loicd	http://wiki.ceph.com/01Planning/02Blueprints/Dumpling/Erasure_encoding_as_a_storage_backend	09:01





    10
    
    
    
loicd	http://pad.ceph.com/p/Erasure_encoding_as_a_storage_backend	09:02





    11
    
    
    
*** sjustlaptop has joined #ceph-summit1	09:02





    12
    
    
    
sagewk	anybody else want to join the hangout?	09:03





    13
    
    
    
nwl	in future, i wonder if it's better to keep the blueprint out of the etherpad	09:04





    14
    
    
    
nwl	so one can just see the changes/comments etc	09:04





    15
    
    
    
rturk	the etherpad only contains the work items	09:04





    16
    
    
    
rturk	but ya, maybe it should just be for notes?	09:04





    17
    
    
    
nwl	rturk: yeah, so the notes can be then fed back into the blueprint, redmine etc	09:05





    18
    
    
    
rturk	(ideas welcome -> community@ceph.com, now or anytime!)	09:05





    19
    
    
    
saras	have you look at http://www.opencompute.org/projects/storage/	09:05





    20
    
    
    
*** liwang has joined #ceph-summit1	09:06





    21
    
    
    
dmick	saras: I have not; did you mean relative to the management API work specifically, or in general?	09:09





    22
    
    
    
nwl	saras: sage is involved in the project	09:09





    23
    
    
    
saras	thanks	09:10





    24
    
    
    
nwl	saras: we are not doing anything very actively though (as Inktank)	09:10





    25
    
    
    
saras	can you guy setup lower thirds	09:11





    26
    
    
    
saras	please	09:11





    27
    
    
    
rturk	lost your audio, loicd	09:12





    28
    
    
    
yehuda_hm	in practice an osd should not become full, unless something is misconfigured, or the entire cluster filled up	09:12





    29
    
    
    
loicd	damned	09:13





    30
    
    
    
loicd	ahahah	09:13





    31
    
    
    
loicd	I'm back	09:13





    32
    
    
    
saras	append would be great	09:13





    33
    
    
    
*** maclin has joined #ceph-summit1	09:16





    34
    
    
    
hskinner	erasure encoding is of course most valuable with geo-dispersment	09:17





    35
    
    
    
sagewk	hskinner: true	09:17





    36
    
    
    
sagewk	it's also not latency-sensitive in this use case...	09:17





    37
    
    
    
hskinner	yes, ability to migrate to optimal region locality would be bonus	09:18





    38
    
    
    
ksp	erasure encoding is also valuable in metro cluster/distance; e.g. EMC Isilon OneFS is using such algorithms	09:21





    39
    
    
    
saras	is their paper that i can read about erasure encoding	09:22





    40
    
    
    
dmick	saras: a bunch of resources on http://wiki.ceph.com/01Planning/02Blueprints/Dumpling/Erasure_encoding_as_a_storage_backend	09:24





    41
    
    
    
saras	dmick: thanks	09:24





    42
    
    
    
*** michael__ has joined #ceph-summit1	09:24





    43
    
    
    
*** michael__ has quit IRC	09:24





    44
    
    
    
*** yanzheng has quit IRC	09:26





    45
    
    
    
*** michael_dreamhost has joined #ceph-summit1	09:26





    46
    
    
    
ksp	http://en.wikipedia.org/wiki/Erasure_code	09:27





    47
    
    
    
*** yanzheng has joined #ceph-summit1	09:27





    48
    
    
    
saras	pg to be  clear	09:28





    49
    
    
    
</pre>