14. 14
THE END OF THE CLOUD
50bn IOT devices by 2020
Cars, Drones & Robots are mobile data centers
Data needs to be processed in real time != Cloud
Data gravity is about where data is born
Ceph’s origins were in addressing issues around scaling storage
Initial use-cases for the community were naturally about scale too.
When you start to have at-scale deployments with brand names (telcos! Banks! service providers!) you start to see dollar signs and a business was born.
And all start-ups need an ambitious goal..
…and Inktank’s was to become the next major storage vendor.
The Red Hat playbook looms large in every open source startup’s mind and storage felt like a perfect market to run the same playbook
Practically, there were distractions where people expect you to answer every storage question under the sun.
These are the use-case which are bread and butter for somelike EMC
But there was enough opportunity without addressing these requests in the short term.
But the bigger issue in the storage industry was that public clouds were doing the disruption through commoditizaion, not open source startups.
And these were the guys actually hurting EMC and NetApp way more.
All of the mid-tier was and is moving to them
There are definitely counters to the public cloud and this is what many vendors continue to make money on, including Red Hat.
These are good reasons and will continue to resonate but with a shrinking audience, where the mid-market moves to the cloud.
Meanwhile, all the vendors are failing to a divide and conquer where we fight with each other while AWS go marching on
It isn’t sufficient to just aim to deliver to the same use-cases that EMC or NetApp did. That’s not how you will become the next major storage vendor.
And it is also a negative question to ask “what’s left”
We need to be more aggressive in our view of the future.
Imagine a world where AWS are not dominant.
It’s hard but it’s also exciting.
Telcos are never going to use the cloud. They have too much and services that need to reach out as close to the cities people live in as possible to provide the best quality of service.
While huge content repositories and core NFV are about scale, telcos are running OpenStack and Ceph in tiny pods – sometimes just a few servers – to run NFV out in small cabinets, in random offices, in small cities – the last mile. We’ll come back to this in a minute.
Peter Levine looked at how edge computing is going to mean that cloud doesn’t remain as frightening and dominant as it is now
Move from world where cloud is the center of gravity and our edge devices (phones) are thin viewers into all the cool stuff at the center (think Slack, Facebook, etc)
To a world where the edge is where all the cool things happen.
The data gets generated and used.
The TAM then becomes billions not millions.
We have had a preview of the edge with Remote NFV
Let’s break down what happens at the edge to see why the cloud doesn’t make sense
See examples with the police and long-distance trucks
Charlie Stross invented term “lifelogging” – Black Mirror episode too!
Not about Unified Storage
And not about Linux of storage to mean open source
But run everywhere in all sorts of different sizes
Data lakes are a new concept where customers who need to hold onto massive amounts of data that will eventually need analytics, host it in a single location which can either be HDFS or, to increase flexibility, object storage.
The excitement here is around fast object storage and the ability to do ephemeral analytics.