Table of Contents

Dr. B. in da SPECTORS project

Cloud Computing

The idea stems from Prof. Sandro Leuchter. Together with Matthias Krauledat he purchased 60 Dell R220 power edge servers to set up a cloud computing environment: Infrastructure as a Service (IaaS).

The setup of each Dell R220 as of 2014:

Components

Software

Service

R220

CPU Options for PowerEdge R220

Intel® Xeon® processor E3-1200 v3 product family, Intel Pentium®, Intel Celeron®, Intel Core®

This v3 is a Haswell architecture: https://de.wikipedia.org/wiki/Intel_Xeon#Haswell-Generation

https://www.computerbase.de/2014-05/intel-xeon-e3-1200-v3-haswell-refresh-spezifikationen/

BUG in Haswell: TSX not working properly.
“In August 2014, Intel announced that a bug exists in the TSX implementation on Haswell, Haswell-E, Haswell-EP and early Broadwell CPUs, which resulted in disabling the TSX feature on affected CPUs via a microcode update. The bug was fixed in F-0 steppings of the vPro-enabled Core M-5Y70 Broadwell CPU in November 2014.”
Source: https://en.wikipedia.org/wiki/Transactional_Synchronization_Extensions#implementation

https://www.heise.de/newsticker/meldung/Intel-Bug-Vorerst-kein-Transactional-Memory-2292196.html

“Insbesondere bei Supercomputern und Datenbankservern kann die Funktion deutliche Geschwindigkeitssteigerungen bringen, denn TSX sorgt dafür, dass voneinander abhängige Threads Cores weniger oft auf gesperrte Speicherbereiche stoßen.”
http://www.golem.de/news/transactional-memory-intel-schaltet-tsx-wegen-bug-bei-haswell-ab-1408-108545.html

I would like to use Xeon E3-1240 v5 or Xeon E3-1240L v5 (low power). Version v5 is not compatible with the chipset C222 of PowerEdge R220. The v5 needs dedictated Intel Xeon chipsets C232 or C236.

:!: We are stuck to v3 and therefore the TSX bug!!!

Used Servers (refurbished?)

https://www.servershop24.de/

Apache Spark and Hadoop

http://www.computerwoche.de/a/apache-spark-versus-hadoop,3096930

http://www.infoworld.com/article/3014440/big-data/five-things-you-need-to-know-about-hadoop-v-apache-spark.html

https://www.xplenty.com/blog/2014/11/apache-spark-vs-hadoop-mapreduce/

Kevin Mader, 4Quant, ETHZ

Giachetta