Well we did it, we held our first “Meet the Devs” in the US and I’ll venture to say it was successful. Let me emphasize the “we” involved in this event. This was a great joint effort between users and IBM with special thanks to Janet Ellsworth of IBM for all she did to make this event happen, and Doris Conti for making sure lots of IBMers, including lots of developers, came down from Poughkipsie. Thanks also to the US co-principal Bob Oesterlin and Pallavi Galgali for keeping things moving right along behind the scenes. We’ll see Bob and Pallavi in person at the SC15 meeting in Austin.
There’s a poll we’re running to get feedback from the attendees and we’ll see how my own feelings match up with the group’s. The discussion was lively, and everyone felt comfortable to ask direct questions and talk about it all —the good, the bad and the ugly.
From an attendee perspective there was reasonable diversity as well with representatives from financial, medical/genetics, and university sectors.
So, what was covered? I’m glad you asked.
The first presentation was about enhanced protocol support including (NFS, Samba, Swift and Cinder). Gautam Shah talked about these features, with some planning recommendations regarding number of active connections, maximum number of protocol nodes (16 for SMB and 32 for NFS) among other things. As the use cases for GPFS expand beyond just delivering performance, and functionality needs increase these protocols of convenience, like Samba, will no doubt become more important, and I for one am happy to see them get official support.
Next we got a live demo of the Spectrum Scale Dashboard from Thorne Ventura, and while this product is in its early days it will definitely be helpful especially to those who aren’t seasoned command line GPFS admins. As I understand it, the pieces that make up the Spectrum Scale (there I said it!) GUI will become the common underpinnings for most IBM GUI interfaces to storage. One nice feature was that when you run a command in the GUI it can be extracted and saved for scripting or future reference. The dashboard includes views into, and the ability to change, several things including: snapshots, per protocol statistics, performance, quotas and ILM policies. A request was made to include tracking GPFS waiters in the tool, and the ability to view which specific nodes are having issues. A beta of the GUI is available.
Zach Giles from Mount Sinai was our first user presentation. Zach’s presentation generated a lot of discussion, and like many of us Zach is seeing a deluge of small files, whee! Zach had, at the end of Zach’s presentation, several of the issues he had run into and some suggestions for IBM and lots of dialog surrounded this part of the presentation.
Lastly, Paul Sanchez of DE Shaw talked to us about his experiences, including big gains with using flash for metadata and hopes for new developments with GNR (aka declustered RAID).
We ran out of time for a full hour of Q&A, because there were so many questions peppered throughout the afternoon, but I think that’s a good sign that the topics were engaging to many.
Thanks to all who participated, and hope to see some of you at a future event.
PS - I know it’s difficult for some to travel to events, especially if they’re just half a day long. So, I encourage you to express what you would like to see the user group do either online, via webinar or some clever thing only you can come up with. :-) Topics for discussion are always welcome. This is your users group so do participate.