|
it was more data warehousing and updated once a day/once a week, and not constant writes, so i dunno
|
#
?
Jul 21, 2012 22:01
|
|
|
|
| # ? May 15, 2024 04:55 |
|
|
tinselt0wn posted:it was more data warehousing and updated once a day/once a week, and not constant writes, so i dunno data warehousing is pro, and it's much smarter to have your application & db design made for performance/scaling or whatever you need, and have a real data warehouse for reporting.
|
|
#
?
Jul 21, 2012 22:13
|
|
|
tinselt0wn posted:i dunno ive dealt with sql databases up to about half a billion rows and it was ok and at least people understand it rather than the nosql du jour so whatever works for your particular application i guess im a big fan of considering your design carefully and using the right tool for the job. im a tool eheh heh heh
|
|
#
?
Jul 21, 2012 23:06
|
|
|
just like to say this thread rocks and has introduced me to some cool new topics. too bad I'll never get to use them because I'm one of three programmers supporting a 30 year old enterprise system written in *sigh* Progress 4GL/OpenEdge ABL. I'm still trying to teach one of the programmers that's been with the company 20+ years how to use OO.
|
|
#
?
Jul 22, 2012 02:23
|
|
|
Pendragon posted:just like to say this thread rocks and has introduced me to some cool new topics. if you're not having fun and being intellectually challenged at your job you should quit and find one that does challenge you
|
|
#
?
Jul 22, 2012 02:37
|
|
|
BonzoESC posted:if you're not having fun and being intellectually challenged at your job you should quit and find one that does challenge you
|
|
#
?
Jul 22, 2012 03:30
|
|
|
Also curious to know about the mongo hate too. Using it now for realtime analytics and it seems p sweet
|
|
#
?
Jul 22, 2012 03:30
|
|
|
mongo is fine if your data doesnt matter
|
|
#
?
Jul 22, 2012 03:40
|
|
|
Shaggar posted:mongo is fine if your data doesnt matter
|
|
#
?
Jul 22, 2012 04:22
|
|
|
Shaggar posted:mongo is fine if your data doesnt matter is it just the retarded sharding issue? or does it have other data loss problems
|
|
#
?
Jul 22, 2012 04:29
|
|
|
Shaggar posted:mongo is fine if your data doesnt matter shaggar was indeed right and this why google created f1 to run their adwords biz to replace mysql, which is still better than nosql if your data matters using nosql for important data is like using floats to store your financial data Cold on a Cob fucked around with this message at 04:37 on Jul 22, 2012 |
|
#
?
Jul 22, 2012 04:34
|
|
|
iamthexander posted:Also curious to know about the mongo hate too. Using it now for realtime analytics and it seems p sweet the foursquare post I linked last page is basically that they shared based on ranges of user id and one of the nodes got overloaded and slowed way down, and they couldn't recover without a big offline recovery process because of fragmentation on the overloaded node what a real distributed database does (Cassandra and riak) is sharded based on a cryptographic hash of the key, which means that distribution is more random; broken into vnodes, which makes distribution more flexible to modify while the cluster is online; and has tunable redundancy, so you can say that values in a given bucket are stored on three nodes, writes don't succeed until it's been flushed to disk on at least two, and if you want fast reads you just want to wait for an answer from one node instead of three the indexing features that mongo's had since the beginning are initially attractive to idiot web devs, but the redundancy, distribution, and integrity features that riak has had since the beginning are better for sustained operations you hear about lots of apps (like bump) moving from mongo to riak as they grow, but never anybody moving from riak to mongo and it's because riak is easy to keep running
|
|
#
?
Jul 22, 2012 04:40
|
|
|
BonzoESC posted:you hear about lots of apps (like bump) quote:Instead, we've written a thousand-line Haskell program that acts as an interface between the app nodes and the database, which exclusively deals with conflict resolution and sibling merges. Every Riak interaction is done through this tool that we've dubbed /magicd/, which guarantees the app nodes always see a consistent truth. This obviates the largest pain point of eventually consistent databases. Ugh e: but the whole "just throw more nodes in the cluster and do read-repair" seems nice salted hash browns fucked around with this message at 05:06 on Jul 22, 2012 |
|
#
?
Jul 22, 2012 05:01
|
|
|
iamthexander posted:Ugh this is a fat idiot talking about crdts: http://vimeo.com/45433300
|
|
#
?
Jul 22, 2012 05:14
|
|
|
BonzoESC posted:yeah that's the trade off since you can't do distributed locks, you have to use CRDTs to merge parallel writes woah were you there when tef was there
|
|
#
?
Jul 22, 2012 05:22
|
|
|
iamthexander posted:woah were you there when tef was there
|
|
#
?
Jul 22, 2012 05:24
|
|
|
iamthexander posted:woah were you there when tef was there goon meet
|
|
#
?
Jul 22, 2012 05:26
|
|
|
Shaggar posted:mongo is fine if your data doesnt matter mongo is fine if you want to spend an order of magnitude more time on operations than development.
|
|
#
?
Jul 22, 2012 05:28
|
|
|
BonzoESC posted:if you're not having fun and being intellectually challenged at your job you should quit and find one that does challenge you or slack off at your current job and do interesting things while no one is looking. Pendragon posted:just like to say this thread rocks and has introduced me to some cool new topics. most of the cool poo poo you never get around to using. when you do, it sorta loses its shine. Pendragon posted:because I'm one of three programmers supporting a 30 year old enterprise system written in *sigh* Progress 4GL/OpenEdge ABL. I'm still trying to teach one of the programmers that's been with the company 20+ years how to use OO. you're getting paid more than him, right ? tef fucked around with this message at 05:32 on Jul 22, 2012 |
|
#
?
Jul 22, 2012 05:29
|
|
|
on the whole lol web2.0, people have been using things that speak memcache as persistent stores. but the libraries treat caches being offline as the key not being present, cos it assumes it's speaking to a cache. so they either have to rewrite the library, or write more code around the library to check the server is up. that is, if they want the website not to fall over. quote:I remember Mike, Robert and I having great fun asking the same question over and over again: "what happens if it fails?" -- the answer we got was almost always a variant on "our model assumes no failures." We seemed to be the only people in the world designing a system that could recover from software failures.
|
|
#
?
Jul 22, 2012 05:36
|
|
|
who the gently caress designed the html5 canvas api, it's awful
|
|
#
?
Jul 22, 2012 05:41
|
|
|
apple
|
|
#
?
Jul 22, 2012 05:46
|
|
|
ugh not those guys again they just seem to ruin everything they touch
|
|
#
?
Jul 22, 2012 05:46
|
|
|
why would you include a way to get the width of a block of text but not the height
|
|
#
?
Jul 22, 2012 05:48
|
|
|
yaoi prophet posted:why would you include a way to get the width of a block of text but not the height lol its trivial just
|
|
#
?
Jul 22, 2012 05:52
|
|
|
yaoi prophet posted:why would you include a way to get the width of a block of text but not the height what api
|
|
#
?
Jul 22, 2012 05:55
|
|
|
yaoi prophet posted:why would you include a way to get the width of a block of text but not the height at a guess: the underlying draw text primitive doesn't handle word-wrapping to a length the measuring text works by rendering the text without drawing it, but checking where the new x,y coordinates are. so because it doesn't wrap, the y coordinate will never change.
|
|
#
?
Jul 22, 2012 06:02
|
|
|
Getting the height of a block of text can be really tricky these days. You've got quite a few different values that could be considered to be the text's height. Better to just pretend height doesn't exist, otherwise webdevs will get confused.
|
|
#
?
Jul 22, 2012 12:18
|
|
|
BonzoESC posted:yeah that's the trade off since you can't do distributed locks, you have to use CRDTs to merge parallel writes was this intentionally confusing or what?
|
|
#
?
Jul 22, 2012 12:23
|
|
|
how dumb are you really?
|
|
#
?
Jul 22, 2012 12:53
|
|
|
reallyprettydubm
|
|
#
?
Jul 22, 2012 13:13
|
|
|
Stringent posted:was this intentionally confusing or what? no but it was for a ruby thing not a php thing so i didn't feel i had to stick to one-syllable words
|
|
#
?
Jul 22, 2012 14:14
|
|
|
BonzoESC posted:if you're not having fun and being intellectually challenged at your job you should quit and find one that does challenge you working on it. tef posted:or slack off at your current job and do interesting things while no one is looking. that's kind of what I do now. "hmmm... I have some downtime...." *write python startup script for software package that can recover from failures* *write flex class to replace 400 lines of error-prone code from almost all of our programs* *document everything* ... *no one ever uses them* tef posted:you're getting paid more than him, right ? ...probably not. that must be rectified.
|
|
#
?
Jul 22, 2012 14:23
|
|
|
BonzoESC posted:it's basically way clumsier than just using SQL at a small scale, and once you need to share it, it turns out to be awful Cassandra doesn't have vnodes yet (they're coming soon afaik, though maybe my info is out of date), only hard server nodes. This creates lovely issues where you need to grow your cluster of DB servers by a number that evenly divides the currently existing hash ring; otherwise you'll be creating hot spots around it. Also, MongoDB, before v2.0 (iirc) had one global lock for all writes. Write data? All tables everywhere on the server are locked, both for reads and writes.
|
|
#
?
Jul 22, 2012 15:03
|
|
|
i like how everyone itt likes to periodically jizz over the D Language Forum's performance, and my Java web thing is pretty much that fast and i don't even do anything terribly clever with it (there's a bunch of memcached hooks in the code but i haven't actually got around to writing a Cache impl that actually remembers anything. i prob never will because tef says cache invalidation is Hard and tef is always right) postgres and the jvm ft loving w
|
|
#
?
Jul 22, 2012 15:18
|
|
|
If you're writing a website use a web cache.
|
|
#
?
Jul 22, 2012 15:23
|
|
|
Mr Dog posted:i like how everyone itt likes to periodically jizz over the D Language Forum's performance, and my Java web thing is pretty much that fast and i don't even do anything terribly clever with it (there's a bunch of memcached hooks in the code but i haven't actually got around to writing a Cache impl that actually remembers anything. i prob never will because tef says cache invalidation is Hard and tef is always right)
|
|
#
?
Jul 22, 2012 15:38
|
|
|
BonzoESC posted:the two hardest things in programming: naming things, cache invalidation, and off-by-one errors :groan:
|
|
#
?
Jul 22, 2012 15:57
|
|
|
yaoi prophet posted:why would you include a way to get the width of a block of text but not the height http://www.whatwg.org/specs/web-apps/current-work/multipage/the-canvas-element.html#textmetrics its just theres like ten different things that qualify as height
|
|
#
?
Jul 22, 2012 16:49
|
|
|
|
| # ? May 15, 2024 04:55 |
|
|
Alligator posted:you can  lol
|
|
#
?
Jul 22, 2012 17:09
|
|