That’s a wrap!
Yes, we’ve made it to the end of another successful Spectrum Scale user group! This year has been our most highly attended group with over 110 registrations and around a 90% attendance rate. We had representation from 56 organisations and delegates from 13 countries as well as 29 speakers over 28 sessions. A big thank you goes out to all our IBM, sponsor and user speakers.
For 2018, we headed back to London and the IBM Client Centre London on the South Bank. The client centre team rightly deserve our thanks – where nothing was too much trouble for them. Finding us cables to plug our streaming kit and generally making sure the logistics of the venue were all dealt with.
A few of us kicked off the group with an informal meal and beers the evening before, and the networking opportunities with development are one of the key features that makes the user group so successful.
Day One
Just before starting, we had a bit of trouble getting our live streams online and the audio working… it’s the first time we’ve tried live streaming the event and hopefully, some people found it useful – we know we had watchers out there. We kicked off in earnest on Wednesday morning with a welcome from myself and Claire. Ted Hoover from IBM also stepped in with opening remarks. Mathias Dietz then picked up the baton with an update on the changes in the latest 5.0.0 release stream and Chris Maestas updated on ESS changes and a little on some changes to the support process.
Following on from his previous successful UG talks, Rob Basham led a session on Transparent Cloud Tiering (TCT), covering the changes to enable multi-cloud support and giving an update on some future plans. Of particular note, Rob mentioned that whilst at last year’s event, a conversation led to an idea and breakthrough in developing TCT. So it’s not just us users of Scale who benefit from the group!
Our first sponsor talk, Stuart Campbell, E8 Storage was up next with some pretty impressive performance numbers using the NVMe storage systems. Our attendees then hungrily headed out for the “grazing bowl lunch”, which made a great change from the usual sandwich fare of events.
After lunch were a group of user talks, starting with EMBL-EBI talking on a performance issue they have been having with MMAP, this dove-tailed into a short talk from Sven on the code improvements coming in 5.0.1 that he’s worked with the site to develop. NEC’s sponsor slot was taken by Jan Erik Sundermann from Kalsruher Institute of Technology describing their needs and use of NEC’s GxFS storage appliance to support this.
Unfortunately, our planned speaker for the next slot had to drop out at the last minute, but we filled the slot with a discussion on tooling of Spectrum Scale. Ulf is keen to see the community contribute some of the code that we are all developing in-house and opening it to the community. It’s something that we all think could be of great benefit to the community and would ensure the IBM Development team can focus on core technology.
Oliver Gappa, Rohde and Schwarz then gave us a pretty cool insight into some of their technology being built to support broadcast systems, with some scary numbers on the cost to broadcasting of a storage failure (think advertising revenue!). The technology they are developing is based on Spectrum Scale to support high availability storage. He also gave some fascinating insight into why there is a move to IP in broadcasting, not least due to the lack of skilled engineers in traditional technologies.
Before the break, Laurence, from sponsor OCF presented his pilot work with NooBar to provide a transparent backend to TCT functions. Allowing NooBar to be used to transparent (to TCT) way to migrate data between cloud providers.
Returning from a much-needed coffee break, Rosemary Francis, from sponsor Ellexus gave a talk on the 10 commandments of good IO, basically a “lessons learned” from the application of their technology to help improve client applications use of IO. Michael Holliday from The Francis Crick Institute then presented some of the challenges and use of AFM when moving into the Institute’s new building – bringing together numerous separate organisations under one roof.
The last two slots were filled by IBM speakers, first up Luis Bolinches with stories of pool and storage migrations with some pretty important warnings on changing block sizes. Carl Zeite then did an “Ask me (nearly) anything” session. Originally this was planned to be an RFE community voting process, but we only had 1 submission. Carl talked about why the new process, where attendees would get 3 votes is important as it requires us users to consider which RFE is most important to us. He also gave examples of good and bad RFEs. Most importantly he wants the community to explain the problem they have, rather than presenting a solution without defining the problem.
Our evening social/networking event took place at The Wharf, just a few minutes walk along the river. The sunny weather was certainly welcome and we had the use of their terrace. A big “Thank-You” goes to the staff at The Wharf who were friendly and helpful for us. Hopefully our delegates appreciated the food and drinks, and a big thank-you to all our sponsors for helping make this happen.
Day Two
Day two always starts a bit earlier as generally most people are already in town. And doing this means we can finish a bit earlier and give people time to travel home. Our parallel break-out sessions filled the morning, we tried this out a couple of years and have found it gives us the chance to get more topics on the agenda so people get a bit of choice. We even included a slot on Spectrum Scale on System Z – a first for the user group! Other slots covered the GUI, genomics workloads, audit and a deep dive into AFM. Talks on network flows and system health were also packed in, giving a great insight into helping to troubleshoot deployments.
After lunch, Peter Childs from QMUL presented an overview of their Spectrum Scale deployment for research data and Michael Hennecke from sponsor Lenovo gave an update on DSS-G systems. This segued into an unplanned discussion over the impact of Spectre and Meltdown.
And before coffee, we had an update from Sven Oehme. We listened to your feedback and moved him from the end of the day. We figured before coffee would give people the option of listening to more and asking questions of grabbing coffee. Hopefully you found this a reasonable compromise?
After the break, we headed into the home straight with user talk from MAX IV, a light source location. Then Vic Cornell from sponsor DDN presented some technical quirks they have learned over the past year using AFM for DR (not AFM-DR). Luckily we had Venkat from the AFM team on hand to answer some of the questions as well!
The day wrapped up with Yong ZY Zheng presenting on the HDFS connector for Spectrum Scale as well as some best-practice architectures for using Spectrum Scale with Machine Learning applications.
Thank you all who attended, spoke, sponsored, organised and generally made it work. Once we’ve got back and recovered we’ll be checking out the evaluation forms and starting the planning for 2019! I’ll also take a look at the streams and see if we can cut the videos up into each talk to make it easier for those not able to attend to review the content.
The 2018 user group meeting was sponsored by:
DDN, E8 Storage, Ellexus, IBM, Lenovo, NEC and OCF