GregRahn
-
OH at Bay to Breakers: "No shortage of naked dudes, but I have yet to see a naked chick."
12:04 PM May 20th
via Tweetbot for iOS
from San Francisco, CA
-
@ Suggestion: Instead of AWR screen shots from SLOB runs, how about a results page? Nice to see more than one number at a time.
7:30 PM May 18th
via Twitterrific for Mac
-
Dear @: Graph with 3 axes and no labels (slide 15)? No labels makes me sad.
11:39 AM May 18th
via Twitterrific for Mac
-
Thinking I need to upgrade that last tweet's to a .
4:32 PM May 17th
via Twitterrific for Mac
-
OH: "We're seeing issues with […] when we run with more than 10000 connections on our Oracle database."
4:27 PM May 17th
via Twitterrific for Mac
-
HP internal documents on Itanium released:
10:21 PM May 16th
via Twitterrific for Mac
-
Emails show HP hid details from customers about Itanium end of life.
9:06 PM May 16th
via Zite Personalized Magazine
-
@ G1 is not the default GC in JDK 7u4. Requires -XX:+UseG1GC See
1:05 PM May 16th
via Twitterrific for Mac
in reply to timurakhmadeev
-
@ @ Xeons.
11:38 AM May 16th
via Twitterrific for Mac
in reply to kevinclosson
-
@ It didn't mention output because the discussion is cost on pre-assembled vs cost on DIY for same components.
11:32 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ In 2008 Yahoo sorted 1TB on 910 nodes in 209 seconds. Today, I think you can beat that using 2s Xeon 5600 nodes with 1/25 the HW.
11:31 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ The argument is not cheap nodes vs expensive node. It's cheap clusters vs expensive clusters. Then tie that to work output.
11:23 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ BTW, it's painful, but probably not for the reasons you think. :) /cc @
10:26 AM May 16th
via Twitterrific for Mac
in reply to kevinclosson
-
@ Sure, but that "truth" is over a year old. You need something new to complain about. /cc @
10:21 AM May 16th
via Twitterrific for Mac
in reply to kevinclosson
-
@ Meh. /cc @
10:16 AM May 16th
via Twitterrific for Mac
in reply to kevinclosson
-
@ The goal should be the *right* hardware. More of the wrong, cheap stuff doesn't result in a better hadoop cluster.
10:13 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ Depends if you find it easier to manage 100 nodes or 10. Also depends on density. In 2U you can get 12 x 3TB drives.
10:10 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ What's the typical node config for that 1GbE setup? CPUs, memory, # disks.
8:48 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ …and something else. What's you other hadoop cluster configuration?
8:38 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
-
@ Sounds like you need some time on a BDA and something else. You'll learn quickly where the bottlenecks are and what's important.
7:32 AM May 16th
via Twitterrific for Mac
in reply to gwenshap
|
- Name Greg Rahn
- Location San Francisco, CA
- Web http://structured...
- Bio big data nerd, database junkie, machine learning enthusiast, data scientist in training, vldb performance ninja, root cause addict, mac lover
|
