So take a look at my key takeaways, with with a focus on the Day #2 keynote (don’t miss my Day #1 takeaways here, if you haven’t seen them, watch them first):
No time to watch – read on:
The keynote was lead by Greg DeMichillie who promised that it would be ‘not marketing and all technology’ and it lived certainly up to that, in detail the audience learned:
Google Datacenters - Joe Kava led us through the principles of Google Data Center setup and design, the session was the first time Google shared to a public and wide audience what these principles are. And they are
- Availability – Google design for very high availability, and it needs it for its very own services, achieves it with building own hardware, networking etc.
- Security – Google is committed to security, e.g. was the first major cloud provider to enable HTTPS / TLS by default (as one example).
- Performance – Google’s proprietary designs are aided by e.g. by high efficiency cooling and innovative hardware designs, e.g. the new OCP announcements around 48V rack specification.
- Culture – Google runs its datacenters with its own employees only. From design to built – it’s all done by Google itself, from ‘chip to chiller’. Kava shared a chart that showed that humans make more machines than machines, so the Google built monitoring systems are crucial.
- Sustainability - Google is the only cloud provider running exclusively on renewable energy, it helps that Google is the largest private investor in renewables. Own R&D in cooling technology is another strength.
- Innovation – Google is using its advances on Machine Learning to harden its data center operations further, all along the software supported operations.
Security matters – Next was Niels Provos walking us through the security framework of Google and then GCP. Too much for a blog post – but compare the below two pictures. Google of course re-uses its own Google App security stack, and then extends it for the needs of the GCP customers. No surprise here, good to see the synergies.
|Google Security Stack|
|Google Cloud Platform Security Sack|
Configuration matters – Next up was Eric Brewer, updating us that Google has been using containers since over 10 years over Borg, then Omega and now Kubernetes. He walked us through recently announced Kubernetes 1.2 and then tackled the challenge of configuration, a key DevOps problem. Making the config available via a mounted volume and gives many advantages, and that’s what Google is introducing with Helm, the package manager of Kubernetes, now on GitHub.
- Customer Panel – We had the chance to talk to Google customers Coca-Cola, Snapchat and Spotify. All have very different use cases, Snapchat has been native to Google from the very start, Spotify just choose GCP – mainly for its Analytics / MachineLearning abilities and Coca-Cola built a webscale soccer worldcup picture sharing application. When asked what they would like GCP to improve the mentioned IAM (Spotify), better server farm management (Snapchat) and more partner and SI support (Coca-Cola). Good to see happy customers.
- Partner Panel – Next was a panel of a very different set of partners with Avere (Hybrid storage, burst to cloud), PWC (building new applications on GCP), BitNami (providing the app stack to ISVs) and xxx (a classis ISV, building a content management system). Good to see GCP being able to serve such a diverse set of partners, and good to see happy partners.
- Google pushes MachineLearning further – Google machine learning platform Tensorflow became more interesting given the announcements made on assisted learning. Going beyond the data scientist, machine learning specialist is a key step for advancing the overall industry. And Google know that Machine Learning creates substantial need for storage and compute – something every cloud provider wants. The portability of the Tensorflow models makes the product interesting compared to the competition where more lockin is typically in play.
MyPOVA good event for Google that has shown some very compelling arguments to use GCP.
There can be no doubt at this point that Google understands and operates at cloud scale. Allowing GCP customers to use the same infrastructure and architecture is very attractive for enterprises, as long as their needs fit into the Google use cases. In the executive Q&A Diane Greene was very adamant that it wants all of the enterprise load, and chairman Eric Schmidt made it clear that Google has now understood it needs to come more ‘to you’ (i.e. the enterprises). But there are loads in the enterprise (thing e.g. ERP, CRM etc. systems) that Google has no answer for – except for a re-write of the load into new applications. But that is not palatable for many enterprises, so Google is banking on a multi-cloud future. And we can agree that the future if multi cloud, the question for enterprises is – how much automation / load gets put into which cloud. Naturally they gravitate towards simplicity and to the ‘one butt to kick’ decision making, that has worked well in the past for them – and there is no reason why the same principles will not work in the future.
Google now needs to think hard how wide it wants to make its product offering to capture enterprise load – e.g. Google can and could go after basic cloud load like storage and DR. Not a peep in the keynotes at GCP. For now enterprises should look at GCP for their next gen Application needs that are aligned with Google’s core competencies – that by itself is a great synergy in what enterprises need to build and what Google can offer. If this will be enough for Google to catch up to current players ahead of Google in the cloud race (AWS and Azure) remains to be seen.