Posted on by Simon Thompson

Trip Report: Meet the Devs – Oxford February 24, 2016

In January I took over the advocate role for the Spectrum Scale UK user group from Ross Keeping and was asked to join the ‘Meet the Devs’ in Oxford on February 24th, 2016. ‘Meet the Devs’ is an informal meeting with about 20 customers and partners. Typically there are 3-5 topics which will be discussed. This format was new to me, so I was very interested to travel to Oxford. Rick Welp of the IBM Manchester Lab volunteered to come to Oxford to help with the transition. In addition I was accompanied by Simon Lorenz and Stefan Schmidt of the Mainz Development team.

After the opening by Simon Thompson, the chairman of the UK user group, I started with Spectrum Scale Update and Directions. My key message was that Spectrum Scale is a very successful cluster file system for HPC and that now IBM is adding enhancements for more general data intensive file and object workloads. In the last years we have integrated various access protocols, in particular NFS, SMB, object (Swift and S3) and Hadoop in addition to the native POSIX access, when a Spectrum Scale is mounted locally on a Linux, AIX or Windows server. This supports workflows, where for instance data will be injected via object protocol, analyzed via Hadoop and HDFS and finally backed up using a standard back-up application. Spectrum Scale provides a central file and object repository for all those access methods, eliminating the need to copy files to support multiple access methods.

Then I shared the general customer feedback that those new features are great, but more guidance is needed to configure, tune, monitor and troubleshoot all those new features and Spectrum Scale in general. This was the point, where Stefan Schmidt took over to present upcoming GUI enhancements. He gave an overview of the Spectrum Scale GUI and then showed mockups of new GUI panels for monitoring and problem determination. To our surprise only a few people in the audience have seen the GUI so far, because many are still running GPFS 3.x. We encouraged people to download the Spectrum Scale 4.2 Evaluation Image for VirtualBox to take a look into the GUI and to involve in the Continuous Open Beta for the Spectrum Scale GUI. Both are available on DeveloperWorks.

Next, Simon Lorenz did a deep dive of Spectrum Scale’s object capabilities. Simon used the above mentioned evaluation image to demonstrate object on file by injecting a file via Swift protocol and reading it via SMB. There have been many questions on the technical details and then the session evolved to an intensive discussion on use cases such as metadata search and a couple of other nifty suggestions. The beta code for Object Meta Data Search is available on DeveloperWorks.

Finally Jez Tucker of Pixit Media shared his experience with the beta of the upcoming ‘Transparent Cloud Tiering’ feature. Transparent Cloud Tiering allows to configure cloud storage as a Spectrum Scale Storage Pool and to migrate cold data via configurable policies into the cloud. Jez was telling that the code is pretty stable, but still needs some touch. The beta code for Transparent Cloud Tiering is available on DeveloperWorks.

Overall the meeting was well worth the time. It was good to connect to the UK user group community and to see the faces behind the names on the mailing list. Thanks to Rick Welp for taking care for the ordering of the pizzas and Simon Lorenz and Stefan Schmidt for traveling with me from Germany. I am looking forward to meet the UK user group again at the main meeting in May in London.

Resources: