What you might have missed in analytics since last time…

I feel bad.  I promised a weekly analytics roundup and this one has been festering for a couple of weeks or so.  In a way that's good, otherwise I might have looked like a bit of an Amazon fanboy - but Amazon did announce a lot of new stuff at re:Invent.  But in truth, other work has got in the way of my attempts at bloggerization.  Anyway, enough of me, here's your roundup...

• Microsoft Azure add GPUs:  In my last post on machine learning I touched on the role that Graphics Processing Units (GPUs) can play in neural networks.  Microsoft isn't the first cloud vendor to add GPUs to its cloud, and the details here aren't complete.  But, we still live in an era when exactly what you get for your money in the cloud isn't always easy to figure out.


• Basho revs up Riak TS for the Internet of Things.  I'm keen on the whole NoSQL database concept, I think it can have a big role to play in IIoT infrastructure.  Riak is just such a beast, and this latest release provides explicit support for time series data.  Other time series options for those interested in NoSQL include Open TSDB, DataStax, and MongoDB.


• OK, here comes the Amazon re:Invent stuff, let's try and get it done in one bullet.  This Fortune article tickled me - apparently Amazon got busy locking down a bunch of potential IoT domain names.  That's quite possibly the cheapest way to enter the market without actually doing anything of substance...But onto the real stuff, Amazon beefed up it's analytics offerings with QuickSight.  As Amazon's own website proclaims, "Very fast, easy-to-use, cloud-powered business intelligence for 1/10th the cost of traditional BI solutions".  QuickSight starts at $9 per month per seat, but could cost more.  I've possibly been a tad over critical on the lack of transparency in cloud pricing models in the past.  You could argue if the cost is so low, a lack of transparency and predictability isn't that big a deal.  But, it bugs me nonetheless - and so Amazon QuickSight has the mysterious SPICE ("Super-fast, Parallel, In-memory Calculation Engine") which is somehow going to factor into your costs, but it's not at all clear how.OK, so that's the analytics taken care of.  Amazon also took a crack at the data ingestion problem too with Kinesis Firehose which allows IoT devices to stream data directly into Amazon's cloud storage.  Encryption and compression are options, but otherwise the data is transported in raw form with no processing.  And to extend further, the company also announced 3 Linux based IoT device kits.


• And just when you thought we were done with Amazon...Walmart open-sources its internal PaaS to stick it to Amazon.  Not clear to me what Walmart's motivations are - I can't imagine that spite is one of them.  But, either way, it may be too little too late.


• SAP makes its first foray into the SaaS BI and analytics market.  SAP's had analytics bits and piece in the cloud for a while, but this is a more unified approach.  Given the push to migrate everyone to the cloud over time, this was predictable in everything but the timing.  Interesting to note that the offering includes planning too.   It makes perfect sense to try and integration BI with forecasting and planning - that's why most major planning and BI vendors have tried it over the years.  Here's hoping it's more successful this time - it deserves to be.


• And finally, an example of Industrial IoT in action from the trucking industry.  This case study combines the cloud (Amazon again, sorry...), Hadoop and machine learning.  The potential benefits are many fold:  Predictive maintenance, fuel savings, and helping customers to pick more appropriate truck configurations for their workload.