Jonas-Kress.de

Embed Images in Latex

Jonas M. Kress — Thu, 26 Apr 2012 13:13:47 +0000

When writing papers Latex is the most common tool. Usually it’s very hard to use, but with Lyx a graphical interface it’s pretty easy.
Nevertheless there are some things to keep in mind especially when you work with images.
If you insert an image in a common format (png, gif, jpeg) directly to your document the quality will be extremely bad, because its downsized and compressed while generating the PDF from the Latex document.
So here are my best practices:

Images created in Open Office Draw

File -> Export as PDF
Remove the white space by cutting with PDF Scissors
Use Latex tag “\includegraphics[page=1]{fileName}” with the parameter page you can set the page to be displayed

Voilà your images are inserted to your document without quality loss, because they are inserted as curves.

Images in common formats (png, gif, jpeg)

Create a new Inkscape document
Paste your image
File -> Document Settings -> Fit Page to Selection
File -> Save As -> PDF
Use Latex tag “\includegraphics{fileName}”

Your image quality will not be reduced during Latex to PDF compilation.

That’s the way how you should embed your graphics in Latex to achieve the best image quality.

Evaluating Hive HBase Table Mapping Performance

Jonas M. Kress — Tue, 20 Mar 2012 12:15:12 +0000

As part of my master thesis I am evaluating the performance of Hive queries that use tables stored in HBase.
Although the HBase table mappings in Hive are very experimental and features like server side filtering and partitions have still not been implemented, the performance and scaling seems to be very good.

In my evaluation I am measuring time queries take and compare that to the data set size.
Furthermore the Hadoop cluster is monitored by a Ganglia to evaluate the used resources.

The data set is still very small but growing…

Network:

RAM:

Load:

HBase Requests:

Disk writes:

Disk reads:

CPU:

Hive Table Column Mapping

Jonas M. Kress — Fri, 16 Dec 2011 11:56:13 +0000

When creating a column mapping in hive all the data is being saved to the metastore.

Example of a hive hbase column mapping:
CREATE EXTERNAL TABLE hivetableName(Col String, Col1 String) STORED BY 'org.apache.hadoop.hive.hbase.HBaseStorageHandler' WITH SERDEPROPERTIES ("hbase.columns.mapping" = "colFamily:Col, colFamily:Col") TBLPROPERTIES("hbase.table.name" = "hbasetableName")

If you use a vast amount of columns that are being mapped the table column of the serde params will reach its limit.

You can fix this problem with the following statement if your metastore uses a mysql database:
ALTER TABLE hive.SERDE_PARAMS MODIFY COLUMN PARAM_VALUE text;

Virtualbox Image Resize Disk (VDI, VMDK)

Jonas M. Kress — Fri, 16 Dec 2011 11:29:08 +0000

When resizing a VirtualBox image you may face the error “Resize hard disk operation for this format is not implemented yet!”.
You can fix this by doing a convert to the VDI disk format before trying to resize it.
It could be necessary to boot the converted image before doing a resize!

Convert format of virtual hard disk:
VBoxManage clonehd disk.vmdk disk.vdi --format VDI

Resize virtual hard disk:
VBoxManage modifyhd disk.vdi --resize [SIZE_IN_MB]

You may also want to resize your partitions.
This can be easily done by using the GParted live cd

Master Thesis in Progress

Jonas M. Kress — Mon, 28 Nov 2011 15:53:25 +0000

Started working on my master thesis with the title:

Analysis of migrating a MySQL based data warehouse and monitoring system to hadoop

Most relevant books and papers:

Hadoop The Definitive Guide 2nd Edition and
HBase The Definitive Guide
Hive – A Petabyte Scale Data Warehouse
High Performance MySQL

Further links:
http://www.cloudera.com/ Hadoop distribution and installation package
http://hadoop.apache.org/ Hadoop is an open source Apache project