CLOSER is the international conference on cloud computing and service science. I had the big opportunity to present my first paper (link coming soon) about StorageCloudSim] on this conference. The conference was organized by INTSICC and the WEBIST conference took place at the same time at the same place, which was really nice since you could attend sessions and key notes of both conferences.
The most interesting talks of this conferences for me were:
Rethinking the Necessity of Rapid Elasticity
by Kevin Williams
He questioned one of the characteristics of the NIST definition on Clouds, which says that Clouds must offer a rapid elasticity. Sometimes you should not just scale out (database systems might even slow down if you have write operations that need to lock all nodes in your cluster) or you do not want to scale up (since you run inefficient queries that you should not run at all). Instead you should focus on your performance and use services as AWS to spin up test environments where you do A/B performance tests (see if a code change introduces more latency / computing time / …) before releasing it to Prod.
Keynote on OpenNebula
by Ignacio Martín Llorente
Keynote on Agile Infrastructure at CERN
by Helge Meinhard
Moving 9000 Servers into the Cloud. You can find some slides of a similar talk here
Keynote on Software is eating the world APIs are eating software
He said, software moved from libs that are included into applications to a service oriented architecture and will move on to some kind of API centric paradigm. We will see something like a MVC pattern, but on a level of organizations (model APIs deliver data, like twitter, view APIs deliver content to the user as for example an email sending API and we have controller APIs like ifttt.
Performance Prediction for unseen virtual machines
by John O’Loughlin and Lee Gillam
Amazon’s Computing Unit is not an accurate measurement for computing power. They did a lot of benchmarking and determined the four different CPU types in the standard instance types of one generation and found that the benchmarking of for example the extraction of a zip differs a lot. Different CPU types are more likely in some amazon data centers.
Also very nice - the location of the conference: Barcelona!
blog comments powered by Disqus