What’s Hot in Clouds? Analyze (superficially) the ~140 Papers/Short papers/Workshops/Posters/Demos in CloudCom Each paper may fall in more than one category Core Cloud/Virtualization/Reliability 26 plus ~2 keynotes and 5 tutorials Applications 40 plus 1 (adopting clouds) panel plus ? keynotes Scheduling /resource allocation 16 Security/Privacy 25 plus 1 Panel plus 1 Keynote MapReduce 31 plus 3 tutorials Other Programming models (e.g. workflow) 9 Storage 14 plus 2 tutorials Discovery/semantics 5 Portal/clients 7 Interoperability/Federation 3 plus 1 panel Green IT 6

Tutorials -- MapReduce Yahoo on Hadoop – 40,000 nodes running Hadoop Indiana on Twister – Iterative Chicago on Sphere – General user processing in Map and Reduce; put communication at end of “map” and not beginning of “reduce” as in Hadoop

Tutorials -- Storage Yahoo on HDFS Chicago on Sector – Does not spread blocks of files around, so faster than Hadoop Bret Piatt on OpenStack – has open source service based object store like Amazon S3 (not released yet); supports CDMI Cloud Data Management Interface standard from SNIA (Industry Storage association)

Tutorials -- Core Cloud Platforms Microsoft on Azure Spanish group on OpenNebula RackSpace on OpenStack which has a compute (~EC2) component from NASA that is released; supports OCCI standard from OGF Indiana on FutureGrid including Eucalyptus. FutureGrid also supports Nimbus and expects to support OpenStack and OpenNebula Chicago on Nimbus

Adopting Clouds Panel Agreement that clouds are good for some applications – especially data-intensive and bioinformatics Complaints about details of Hadoop – language and need for text input Software as a Service important; – Globus has data transfer as a Service (Foster keynote) Fortes emphasized need for SLA’s (Service Level Agreements) and multi-cloud systems (Called Sky Computing) Gannon emphasized following day cloud – client interaction e.g. cloud backend to Excel