Last week the third 'meet the devs' event took place, this time in Warwick. This was my first visit to IBM Warwick and after a short walk from the train station the first thing that struck me was the size of the site. It always reminds me of the sheer size of IBM when I visit these sites I barely knew existed but are many times the size of our lab in Manchester.
Having arrived and got set up it was great to see a mixture of familiar faces and first time attendees. This makes me confident that people are finding these sessions valuable if they are attending multiple sessions, whilst also ensuring we are involving and getting input from a widening range of users.
The release of Spectrum Scale 4.1.1 was the first release that our development team in Manchester was involved with. With that work now a month behind us, this meeting was an opportunity to look forward to what is coming in 4.2 - targetted in for release in Q4 this year.
The first topic of interest was how IP addresses are handled when using the new protocol support added in 4.1.1. I came into the meeting with a rough idea of what we are planning and this session provided some great ideas for how we can continue to expand the capabilities of our IP management. The main take-away is that customers would love to have the ability to have control over what determines the way in which addresses move based on a wide variety of factors, ranging from tying specific IP addresses to specific nodes up to dynamically managing them to balance the addresses based on CPU load, network bandwidth and other system statistics. Hopefully the first steps towards this will be made in 4.2.
Next up was a disussion of the work we're doing in the area of big data analytics. You might have heard that IBM is backing Apache Spark and as part of this investigation is ongoing into using Spark with GPFS. The focus at the moment is on tuning GPFS to provide optimal performance with various Spark workloads. Based on the feedback from the group we will be making as many of these workloads and datasets available as possible, so that users have examples for how to reproduce our results on their own systems.
The final thing to cover on the day was to take a look at an early version of the GPFS GUI that will be making its appearance as part of 4.2. This was definitely the topic that had the most interest and generated the most feedback. It was great to hear that although there are plenty of things that might be improved and extended over time, the specifics of which I can't really go into here, everyone seemed to agree that this is a huge step forward in our goal of making GPFS easier to manage.
In amongst the discussions on the agenda topics we had the usual conversations that sprang from people's questions and comments and took us off on various tangents. These free-form conversations are a major reason why I think these days are so valuable - they give us a chance to hear about things that our users really care about in their day-to-day workings with GPFS. A few of the topics covered and requests we received from these conversations were: making some filesystem statistics more easily accessible rather than relying on heavyweight mm commands; adding more statistics for metadata performance monitoring; and looking at optimising what we collect for debug. All of these are things that I can go back and discuss with the rest of the development team and see what we can do to address the issues.
To end I'd like to thank everyone who attended - it's always great to hear from people using the product day-to-day and the feedback is incredibly useful for guiding our priorities going forward. I hope you will start seeing action on at least some aspects of it as soon as 4.2.
Look out for information about the next event later in the year and I look forward to seeing you there.
Disclaimer: All of the above are my own views and I can't make any commitment as to what will be delivered and when.