Modified article

article.txt → GPU vs SUN Fire.txt

@@ -1,15 +1,15 @@
 Analytical queries : Nvidia GTX 260 vs. Sun Fire 
 
-The TPC Benchmark (TPC-H) is a decision support benchmark. It consists of a suite of business oriented queries that
-can be used to measure the processing power of a system. Lets compare top results achieved using Sun Fire with results obtained 
-using a GPU.
-We will use a scale of 100 which means that the size of database is around 100GB and the largest table consists of 600000000 records.
+The TPC Benchmark (TPC-H) is a decision support benchmark. It consists of a suite of business oriented queries
+that can be used to measure the processing power of a system.
+Lets compare top results achieved using Sun Fire with results obtained using a GPU.
+We will use a scale of 100 which means that the size of database is around 100GB and the largest table 
+consists of 600000000 records.
 
 System configuration 1 : 
    SunFire X4270 with
    2 Intel Xeon 5570 2.93 GHz Quad Core processors
    120 GB memory
-   1x 146GB SAS internal disk (10K RPM)
    8 x 32GB internal SSDs
 Software : Sybase IQ 15.1
 Total cost : $61,000 
@@ -29,10 +29,11 @@ the next is a join between 3 large tables followed by a group operation).
 Q1    35s        31s
 Q3     7s        29s 
 
-Predictably, query 3 runs slower on GPU because we lack the memory to do the entire table joins and have to process the data piece by piece.
-Although query 1 runs faster on GPU which is interesting given the difference in price between configurations. 
-Alenka excels at scanning large tables and when using a modern GPU with large amount of memory it is capable of outperforming
-large servers that use a state of the art database software.
+Predictably, query 3 runs slower on GPU because we lack the memory to do the entire table joins and have to process
+the data piece by piece. Although query 1 runs faster on GPU which is interesting given the difference
+in price between configurations. 
+Alenka excels at scanning large tables and when using a modern GPU with large amount of memory
+it is capable of outperforming large servers that use a state of the art database software.
 
 So what makes it fast ? 
 
@@ -48,10 +49,18 @@ Original table takes 74 GB of data. Lets see how we can get it down to a managea
 
 Technical details needed to repeat results :
 use dbgen to generate TPCH data with needed scale
-run alenka scripts to create data files (load_lineitem.sql, load_customer.sql and load_orders.sql) using segment size of 10000000 (-l 10000000)
+
+compile alenka from source
+( nvcc.exe -arch sm_13 -L "C:\Program Files\Microsoft Visual Studio 10.0\VC\lib\amd64" -lcuda C:\alenka\bison.cu -o alenka )
+
+run alenka scripts to create data files (load_lineitem.sql, load_customer.sql and load_orders.sql)
+using segment size of 10000000 (-l 10000000)
+
 run queries q1.sql and q3.sql 
+
 see the results in mytest.txt file.
 
-I have tested alenka only on 64 bit Windows 7, it reads data in a background process when using Windows so results may differ under Linux.
+I have tested alenka only on 64 bit Windows 7, it reads data in a background process when using Windows
+so results may differ under Linux.