I had done a test run with about 6 million rows. That took 8 times as long as the real move with 3.5 million rows. I think InnoDB is not that great for more than a couple of million rows per table.
@RickMeasham we could have a chat. Read and single insert performance we are happy with, but moving large amounts of data to a table that already has a couple of million rows seems problematic. We run into that if we want to make any schema changes.
Well I wouldn't call the critique we got totally undeserved. We definitely didn't react to the probs (that appeared around xmas holidays) quickly enough.
There are a number of excuses for that (andy was sick, I was really busy moving to London) but we should have got our act together earlier.
@mikie not all of it was undeserved, true, but Mashable were still a bit over the top. If Jaiku users had stuck around through all the 504 problems, another 24 hours was hardly going to "drive users in to the arms of Twitter and Pownce".
I think the lesson about communicating with users has been learned, and it could have been a much harsher lesson. I know I look forward to whatever improvements are planned for Jaiku.
45 comments so far
NIce!
2 years ago by yerga
I had done a test run with about 6 million rows. That took 8 times as long as the real move with 3.5 million rows. I think InnoDB is not that great for more than a couple of million rows per table.
2 years ago by mikie
great
2 years ago by CiaranR
it does look like it :D
2 years ago by PolarFox
Awesomeness. Feels faster, too.
Now, for those in the Bay area, get some sleep :)
2 years ago by novemberborn
Well, when expecting 24h downtime these were some really fast hours :-)
2 years ago by localhost
Yay! That was quick
2 years ago by ruiwen
Jippie!
2 years ago by intosi
cool, often it's the other way round!
2 years ago by gerikson
Cool!
2 years ago by pr1001
I know time goes fast - but that fast... WOW!
2 years ago by map
bravo jaiku
2 years ago by oryza
simply well done
2 years ago by topgold
i'm impressed. great job guys!
2 years ago by cybette
Jaiku feels much more responsive now! Excellent! (Using Mauku in N810.)
2 years ago by hhedberg
hellu. didn't even notice it was down hah :)
2 years ago by cervus
Yeah! Well done! It's better to estimate these things conservatively than optimistically.
2 years ago by enfors
wooooooooooooooooooosh. good work!
2 years ago by dw
@mikie: We run innoDB on tables of over 80 million rows ... no problems at all :-D
Fantastic to see you back so fast! WELL DONE! Go team Jaiku!
2 years ago by RickMeasham
w00t w00t~!!
2 years ago by samin
@RickMeasham we could have a chat. Read and single insert performance we are happy with, but moving large amounts of data to a table that already has a couple of million rows seems problematic. We run into that if we want to make any schema changes.
2 years ago by mikie
yea this is crazy fast, keep it this way!
2 years ago by constantine
Really appreciating the new speed ripping through my Jaiku screen. Hoping the robin wasn't killed.
2 years ago by topgold
the robin lives a happy life in our back pocket, feeding on lint, and only has to come out and tweet to you when we have something serious going on
2 years ago by mikie
wow, it's like Fast & Furios IV now !
2 years ago by kOoLiNuS
Nice work Jaiku! You prepare us for a whole day and it's just a few hours! Sweet.
2 years ago by mjohnson
/me not holding breath waiting for Mashable to admit that it was being just a tad melodramatic.
I'd like to see Twitter try and achieve an overestimated downtime like this one :P
2 years ago by smperris
Well I wouldn't call the critique we got totally undeserved. We definitely didn't react to the probs (that appeared around xmas holidays) quickly enough.
There are a number of excuses for that (andy was sick, I was really busy moving to London) but we should have got our act together earlier.
2 years ago by mikie
well we're glad to hear google didn't eat up jaiku and leave it to die, that is what we were all thinking.
2 years ago by constantine
@mikie not all of it was undeserved, true, but Mashable were still a bit over the top. If Jaiku users had stuck around through all the 504 problems, another 24 hours was hardly going to "drive users in to the arms of Twitter and Pownce".
I think the lesson about communicating with users has been learned, and it could have been a much harsher lesson. I know I look forward to whatever improvements are planned for Jaiku.
2 years ago by smperris
So was this a switch to the Google servers or 'just' optimization?
2 years ago by BlueAce
@constantine Dodgeball, anyone?
2 years ago by smperris
Awesome!
2 years ago by Itkovian
@BlueAce 'just' optimization
2 years ago by mikie
Ok. Doesn't change the fact that it feels pretty speedy right now, so gj! :)
2 years ago by BlueAce
Very nice job guys!
2 years ago by Bishop
Nice. If I could just get my guys at work to be this quick.
Also, the "Montgomery Scott" model of time estimation for repair is not a bad thing either. :P
2 years ago by AndyG
San Francisco waking up: great job team :)
2 years ago by jyri
w00t!
2 years ago by silversurfer
Bravissimo!
2 years ago by sachaqs
loving the performance, it's better than ever
2 years ago by silversurfer
Thank you so much, Jaiku! Hope everything stays this smooth from here on out.
2 years ago by pathawks
That sounds great!..@.@ now I can update my account again..^^
2 years ago by yojspew
Imagine a non-beta service doing this to your comms lifeline...
2 years ago by WillemBoijens
muito legal esse assunto mais legal ainda é ter seu site na internet http://www.bhsite.com.br/
1 year, 12 months ago by Janderson