User Tools

Site Tools


cluster:174

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
cluster:174 [2018/08/27 14:21]
hmeij07
cluster:174 [2018/08/28 08:38]
hmeij07
Line 4: Line 4:
 ==== SQL on GPU ==== ==== SQL on GPU ====
  
-MapD built the first ever open source SQL engine to harness GPU computing for analytics. Designed for maximum performance,​ the MapD SQL engine dynamically compiles SQL to run across multiple GPUs and CPUs.+MapD built the first ever open source SQL engine to harness GPU computing for analytics. Designed for maximum performance,​ the MapD SQL engine dynamically compiles SQL to run across multiple GPUs and CPUs. Massively parallel database servers.
  
   * http://​www.mapd.com   * http://​www.mapd.com
Line 10: Line 10:
 ** Browser ** ** Browser **
  
-Log into one of the tails like petaltail, swallowtail or cottontail2 preferably, and with x11-forwarding enabled go to port 9092 of node ''​n37''​ and create a dashboard.+Log into one of the tails like petaltail, swallowtail or cottontail2 preferably, and with x11-forwarding enabled go to port 9092 of node ''​n37''​ and create a dashboard. ​Use the "​data"​ subnet (10.10) eth1. 
 + 
 +  * http://​n37-eth1:​9092
  
-  * http://​10.10.102.47:​9092 
   - Click New Dashboard.   - Click New Dashboard.
   - Click Add Chart.   - Click Add Chart.
   - Click SCATTER.   - Click SCATTER.
   - Click Add Data Source.   - Click Add Data Source.
-  - Choose the flights_2008_10k ​table as the data source.+  - Choose the flights_2008_7M ​table as the data source.
   - Click X Axis +Add Measure.   - Click X Axis +Add Measure.
-  - Choose depdelay.+  - Choose depdelay. ​(departure delay)
   - Click Y Axis +Add Measure.   - Click Y Axis +Add Measure.
-  - Choose arrdelay.+  - Choose arrdelay. ​(arrival delay)
  
-7 million records, amazingly fast I mean  like < 50 ms. There are pie, bar, stacked bar, bubble etc charts and point maps or chloropleth maps. A SQL editor. ​ And I do not even observe a blip on gpu utilization.+7 million records, amazingly fastI mean  like < 50 ms (milliseconds). There are pie, bar, stacked bar, bubble etc charts and point maps or chloropleth maps. A SQL editor. ​ And I do not even observe a blip on gpu utilization. Each of these K20 nodes still has 7 drive bays available which would allow for a large database storage platform if we had a mapd project.
  
 {{ :​cluster:​dep-arr-scatterplot.png?​nolink&​400 |}} {{ :​cluster:​dep-arr-scatterplot.png?​nolink&​400 |}}
Line 31: Line 32:
  
   * https://​www.mapd.com/​docs/​latest/​3_mapdql.html   * https://​www.mapd.com/​docs/​latest/​3_mapdql.html
-  * can be scripted ​too+  * can be scripted ​using mapdql
  
 <​code>​ <​code>​
Line 98: Line 99:
 </​code>​ </​code>​
  
 +  * https://​www.mapd.com/​docs/​latest/​4_centos7-yum-gpu-ce-recipe.html
  
 \\ \\
 **[[cluster:​0|Back]]** **[[cluster:​0|Back]]**
cluster/174.txt · Last modified: 2018/08/28 08:38 by hmeij07