Hadoop or Postgresql for effective processing -

-

i student trying use of machine learning algorithms large data set.we have 140 million records in our training set(currently in postgresql tables) , there 5 tables 6 million records exhibit primary key - foreign key relationships.

we have 2 machines following configurations 1) 6gb ram 2nd generation i5 processor 2) 8gb ram 2nd generation i7 processor

we right planning split them logical groupings before running our statistical analysis since turnaround time quite high.

1) should split them separate tables in postgresql , them use matlab or r programming or 2) should use hadoop hbase porting database 3) should combine , use them(i.e) decompose them based on logical groups , dump in postgresql database , setup hadoop +hbase analysis , use based on necessary algorithms.

thanks

it hard believe in such small cluster hadoop effective. if can parralelize task without - more effective sure
consideration take account - iteration time in learning process. if iteration takes dozens of seconds - hadoop job overhead (which 30 seconds) much.
can hadoop - effective external parralel sort - shuffle stage is. if need - consider using hadoop.
please note in general not easy port relational schema hbase - since joins not supported.


Comments

Post a Comment

Popular posts from this blog

jasper reports - Fixed header in Excel using JasperReports -

-
how can create excel report fixed header using japserreports? mean need header fixed when scroll excel file. this possible adding couple properties in jrxml file each report. take @ advanced excel features freeze panes. if wanted freeze after first column header (down left side basically) this: <statictext> <reportelement style="sans_bold" mode="opaque" x="0" y="60" width="104" height="20" forecolor="#ffffff" backcolor="#666666"> <property name="net.sf.jasperreports.export.xls.auto.filter" value="start"/> <property name="net.sf.jasperreports.export.xls.column.width" value="110"/> <property name="net.sf.jasperreports.export.xls.freeze.column.edge" value="left"/> </reportelement> &
Read more

media player - Android: mediaplayer went away with unhandled events -

-
i need duration of audio file series of voice announcements need play app. have added audio files resources , play fine. sample code below works perfect intended purpose: return duration of audio files. here code: float getdurationofaudioresource(locationenum loc, context context){ float duration = 0; try { mediaplayer mp; mp = mediaplayer.create(context, getaudioresource(loc)); duration = mp.getduration(); mp.release(); mp = null; } catch (illegalstateexception e) {e.printstacktrace(); logerror(25, "testdescitem:fault::could not open mediaplayer object audio resource.");} return duration; } here's weird thing. code called in main activity prepares set of audio instructions given test. there no errors within activity. second activity called, long string of errors on logcat. 03-07 13:23:43.820: i/actionlogger(21435): gentest_info_test #0 created. 03-07 13:23:43.830: i/actionlogger(21435): gente
Read more

python - ('The SQL contains 0 parameter markers, but 50 parameters were supplied', 'HY000') or TypeError: 'tuple' object is not callable -

-
import pyodbc,nltk,array cnxn = pyodbc.connect('driver={mysql odbc 5.1 driver};server=127.0.0.1;port=3306;database=information_schema;user=root; password=1234;option=3;') cursor = cnxn.cursor() cursor.execute("use collegedatabase ;") cursor.execute("select * sampledata ; ") cnxn.commit() s=[] j=[] x=[] words = [] w = [] sfq = [] pos=[] entry in cursor: s.append(entry.injury_type),j.append(entry.injury_desc) nltk.tokenize import punktwordtokenizer nltk.corpus import stopwords tokenizer = punktwordtokenizer() english_stops = set(stopwords.words('english')) in range(0,26): # filter stop words tokenizer.tokenize(j[i]) w.append([word word in tokenizer.tokenize(j[i]) if word not in english_stops]) in range(0 , 26):#converting tokenzied text ito string sfq.append(" ".join(w[a])) replacers import regexpreplacer replacer = regexpreplacer() in range (0,26):#pos tagging replacer.repl
Read more