ORIGINALLY POSTED 15th August 2018

IBM Technical University – Sydney Wrapup

This time last week, IBM’s Sydney Technical University was in full flow. I wanted to provide a quick summary of last week, including our ‘stand room only’ AP FlashSystem 9100 launch event and thank everyone involved, from the organising committee, through our local and international speakers, and of course out clients and partners for all making the events such a success.

It’s actually partly due to the AP TechU events that I now live in New Zealand, after attending the 2014 events and ‘finding a new job’ via the Technical Sales team management. I still remember walking back into Hursley after our family trip ‘down under’ and my then boss lamenting that he almost didn’t let me go as he had a horrible feeling I would want to move (partly due to my Kiwi wife of 15 years!)

Anyway, last weeks 3 day event in Sydney was one of the best we’ve run, with every year the feedback and client satisfaction ratings going up and up. The best thing about IBM’s TechU events is the quality of the sessions – with over 400 sessions across the 3 days, there is plenty to choose from across all of IBM’s Systems – Storage, Power, Z

A couple of stand out sessions for me were (obviously in the storage arena) from Brian Sherman, IBM Distinguished Engineer, with his ‘Advanced Introduction to NVMe’ – with all the discussion and interest in NVMe technologies, Brian gave a great ‘state of the nation’ session, explaining how NVMe is being used, where its ‘enterprise ready’ and what areas are in catch up – mainly the Operating Systems. James Harris from Bendigo Adelaide Bank also gave a great ‘client perspective’ session covering their recent upgrades to SV1 hardware (as well as a retrospective of their SVC clusters that have evolved since 2005) and their implementation of the NPIV capabilities on the latest levels of Spectrum Virtualize code – combining with port masking to simplifying the environment – not only from a day to day perspective – but when trouble shooting – but isolating workloads (node to node, backend, front end, and replication) Its great to hear these sessions, as its ‘from the horses mouth’ so to speak – we as IBM can tell clients and partners the benefits of various features and platforms, but to hear it first hand from an end user means so much more.

Finally Evelyn Perez and Andrew Martin gave an excellent deep technical dive into the new FlashSystem9100 product family. These are the other type of session that are extremely popular at TechU – where you get to hear from the developers who have lived and breathed a product for years and years. This session delved down into the heart of the hardware and platform side of the new FS9100 and set the scene for the launch event on Thursday night.

FlashSystem 9100 Asia/Pacific Launch Event – Sydney

Often, after an event like TechU people will filter away during the last afternoon, but this year we had the launch event straight afterwards, and only next door. Here we had Doug Balog, GM IBM Storage Client Success, Eric Stouffer, VP IBM Storage Offering Management, AJ Casamento from Broadcom and of your myself presenting various aspects around the new all NVMe FlashSystem with our unique FlashCore Modules (FCMs)

During the Technical University we had a demo FS9150 system running, with a very early pre-pre-release version of code running that showcased our NVMe-FC front end driver stack. This system, combined with Broadcom’s G620 switches, and the ex-Emulex 310002 HBA was running NVMe direct from the SuSE SLES12SP3 server right into the FS9100 and down to the NVMe FCM modules – so true end-to-end NVMe all the way. While this wasn’t a performance demo, the system was achieving 60,000 IOPs running 70/30 mix to just one volume, via one path at some 125us latency. But the real point of the demo was to show the reduction in CPU loading on the server.

Since NVMe replaces SCSI as the logical protocol running over the FC transport, the server can get benefit from the reduction in ‘kernel’ side processing. That is, with the NVMe driver being slimline, potentially user space’ and more parallel, we see between 30 and 50% reduction in the amount of server CPU needed to process the same I/O – freeing that CPU to run more VM’s or reduce costs via less CPU licensing for your applications.

The demo was recorded, and when the video is available I will post here, but for now, here are a few images of the FCM modules – 19.2TB with hardware compression – 38TB+ in the palm of your hand.

Thanks again to everyone involved and attending last week – and see you again next year in Melbourne.

---