Yesterday, I published my thoughts recapping Day 1 of Google Cloud Next ’17, which I attended in San Francisco last week. We heard a lot about what Google is doing in the realms of machine learning, amongst a smattering of other announcements in different areas—a new partnership with SAP, a new Engineering Support model, Google in schools, and several others. Day 2 also did not lack in announcements—here’s my rundown of the big news items from Thursday.
Google Cloud Platform updates galore
A large percentage of the announcements from Day 2 were centered around the Google Cloud Platform, its public cloud solution set for businesses. It announced a regional expansion of the GCP, to California, Montreal and the Netherlands, which brings the number of GCP regions up to 6. Both Amazon.com AWS (14) and Microsoft Azure (34) have more regions, but Google has more than 17 regions planned for the future.
Google also announced the public beta launch of Google Cloud Functions, a serverless environment to build and connect cloud services without having to manage any infrastructure. This is like Amazon.com AWS Lambda. While I dislike the industry term “serverless” as there are, in fact, servers involved, it is valuable, as it doesn’t require on-site servers, operating systems, or runtime environments. Google Cloud Functions has the capability to spin up singular functions, and spin them back down immediately, and is the smallest unit of compute currently offered by GCP. Google is targeting extending mobile back-ends, IoT (processing, storage and transformation) and data processing (ie processing images, video transcoding). Google also announced an expansion of their App Engine PaaS that targets web and mobile backends, which now supports Node.js, Java 8, and Ruby, amongst a slew of others like Microsoft’s .Net, C#, PHP and Python. I like that this is in or on top of Docker or Kubernetes, so if you do want to take your workload out of GCP, you can, which should reduce lock-in fear. Bravo to GCP. This is another example of Google “meeting customers where they are”.
Thursday also brought a fair few updates to IaaS BigQuery, Google’s “serverless” enterprise data warehouse product, similar to AWS Redshift. It announced the new BigQuery Data Transfer Service, which will enable the automation of data movement from certain Google SaaS apps (ie DoubleClick, YouTube Content) and partner integrations (ie Informatica, Talend, Tableau, Looker, Qlik) directly to BigQuery, which promises to expedite data analysis and visualization. Google also introduced a new serverless, browser-based service called Google Cloud Dataprep, which also aims to reduce the time necessary to prepare data for analysis—by connecting to the data source, identifying datatypes (and any anomalies), and suggesting data transformations. This should help with batch ingestion and save time getting the data ready to do something useful.
Also announced was that BigQuery would be extending its reach to Cloud Bigtable, Google’s NoSQL Big Data database service. Google also introduced new commercial datasets through partnerships with several companies, including Xignite (financial), HouseCanary (real estate), Remine (real estate), AccuWeather (weather), and Dow Jones (financial)—through a subscription with any of these providers, the corresponding dataset will be instantly ready to query within BigQuery. Think of these as resident data sets enterprises could combine with their own apps that have the same performance and reliability as their own apps.
Google also announced the beta availability of Cloud SQL for PostgreSQL (aka Postgres), a fully-managed SQL database service which Google says will simplify the secure connection to databases from almost any application, from any location, by storing and connecting to a user’s relational data through open standards. Cloud SQL already supported MySQL but PostgreSQL has become a defacto enterprise standard, so this is another example of Google “meeting customers where they are”. It can be scaled up to 32 processor cores, and over 200 gigabytes of RAM—giving it the flexibility to perform with small workloads as well as high-compute ones.
We also learned Thursday that Google would be extending support for Cloud Dataproc, Google’s managed data processing PaaS that uses Hadoop and Spark. Dataproc is similar to Amazon.com AWS EMR and Microsoft Azure HDInsight. Google announced new support geared towards the purpose of automatically restarting failed jobs—ideal for long-running processes. It’s now possible within Dataproc to make single-node clusters for more lightweight data processes, and on the converse, it’s now possible to attach GPUs to clusters for more intensive data workloads. It was also announced that cloud labels and regional endpoints are now available within Dataproc, and that cloud storage processes would be optimized through switching the Java SSL provider to one based on Google’s own BoringSSL instead.
I’ve been pretty impressed by Google’s ability to beef up their cloud infrastructure, without jacking up their prices substantially. They’ve doubled the number of vCPUs you can run simultaneously, from 32 to 64, and increased their memory offerings up to 416 gigabytes. For intensive, parallel workloads, they’re also now offering GPUs. Google also proudly announced that they were the first public cloud provider to successfully run Intel’s Skylake, a custom chip that brings what Google says are “significant enhancements” for compute-intensive workloads. All the while, they’ve dropped prices for Compute Engine, extended the GCP free trial from 60 days to 12 months, and introduced new “Always Free” products. With a 1 or 3-year purchase commitment, their Committed Use Discounts can cover up to 57% of the asking price of Compute Engine (no money up front, billed monthly). For the power and capabilities Google is bringing to the table, the price seems quite reasonable.
My head is still swimming from the onslaught of GCP announcements, but that doesn’t necessarily mean its easy to compare GCP to Azure to AWS. It isn’t.
Google bulks up G Suite
In addition to public cloud services, my company also tracks collaboration tool and even tests and uses tools like G Suite, Office 365, Slack and Cisco Systems Spark. At Next ’17, Google announced a large handful of updates to G Suite. Quite a few functions were added to Google Drive—Team Drives is now in GA (general availability), which helps streamline the management of permissions, ownership, and file access for organizations. Google Vault for Drive, important for many legal, compliance and regulatory reasons, is also now in GA, as well as Quick Access, which predicts what info you will need and tees it up first in Drive using machine learning. I have used Quick Access in Beta and I have yet to be wowed by what comes up, but hope to the more I use it. Google also announced that Drive File Stream is now in the Early Adopter Program, which will allow users to access to large amounts of cloud storage content without having to sync with their desktop, or having it take up space on their hard drive. This is very like Dropbox SmartSync, Apple’s iCloud but unfortunately not in Microsoft OneDrive any more. The user sees an icon on their device even though the file isn’t on the device itself, the file is in the cloud. The final Drive announcement was that Google had acquired their partner AppBridge, to assist customers with large, complicated data migrations to Drive, Gmail, and Calendar. These migrations could come from many places, including Microsoft (OneDrive, SharePoint, Exchange, Office365), WebDAV services, Box, Dropbox, Egnyte, and Citrix Sharefile.
Other workplace collaboration updates include Hangouts Meet—now in GA, it makes it possible for up to 30 people to join a video meeting in a matter of seconds, via any Android or iOS device. It doesn’t require any browser plug-ins or downloads, and it integrates with G Suite for file sharing. I’ve been using Hangouts video for years and some of the improvements include G Suite integration and a dedicated dial-in phone number for people with a lousy data connection. I haven’t used Meet yet, but I hope they have improved the video quality over standard Hangouts video.
Right on the heels of Microsoft’s disclosure of Teams, Google announced Hangouts Chat as part of the Early Adopters Program. Chat is a team messaging app like Teams, Slack, and Spark and allows G Suite file sharing and can integrate with many 3rd party tools like Asana, Box, Prosperworks and Zendesk. Google really needed to bring Chat out and to be blunt, I am surprised it took this long. I have used Hangouts for years with my team and it’s out of date and aged. Chat can also incorporate intelligent bots, such as @meet, which was unveiled at GCN ‘17, and automates the scheduling of meetings. I am looking forward to testing both Chat and Meet with my teams.
Jamboard, Google’s take on the intelligent whiteboard, also received a general release date (this coming May), starting at $4,999, with a $600/year management and support fee. I had a great conversation with the Jamboard team at Next ’17 and got a chance to use it, too. I like their approach to white-board first and think this is a differentiated approach to both Spark Board (video first) and Microsoft Hub (file collaboration first). This team gets it. I am hoping to use Jamboard for a long-term trial to really understand it better.
Add-ons for Gmail were also introduced, which will serve to integrate enterprise workflows within Gmail. One of the coolest things about this, in my opinion, is that these add-ons will be triggered by the content and context of the email. These add-ons will be available through the G Suite Marketplace later in 2017. Gmail has many add-ons today, but these specific ones are considered “enterprise-ready”. I need to get more details on Add-ons as they are in developer preview.
Google amps up security offerings
Day 2 came with a slew of security updates to Google’s portfolio. This makes sense as one of IT’s biggest fears is cloud security, whether it’s warranted or not. It announced that both their Identity-Aware Proxy (IAP) and Data Loss Prevention (DLP) API are now in beta for the Google Cloud Platform. While traditional VPN access is more of an all-or-nothing approach, the more flexible IAP allows granular access to apps, based upon risk. Google has used IAP for years and used by all Googlers. The DLP API scans for sensitive data types, so that they can be managed properly, identified, and redacted. The DLP demos are very cool and powerful.
Several other security updates are the GCP are now generally available, including Security Key Enforcement, also used by all Googlers, enabling the requirement of physical security keys as a second factor for G Suite authentication, and the Key Management System, which enables the user to create, use, switch out, and destroy symmetric encryption keys within the cloud like enterprises do on-prem today.
Lastly, Google introduced Titan, their custom security chip, designed for cloud security at the hardware level. It functions by establishing a cryptographic identity for specific pieces of hardware, and according to Google, will allow for more secure identification and authentication by establishing a hardware root of trust. This intended to safeguard BIOS attacks, which, according to Cisco’s latest security report, is on the rise.
Google will need to work very hard convincing enterprises their data is safer in the Google Cloud comingled with other data than on-prem in enterprises own datacenters.
Clearly, there was no lack of new announcements from Google Cloud Next ’17—attendees were bombarded with quite a lot. Google has made a ton of progress in such a short period of time and I’m impressed at just how much it has done in just the last year by building out the portfolio. I’m encouraged by the updates made in both the Google Cloud Platform and G Suite—though I’m still not necessarily seeing a whole lot of overt differentiation. Maybe this happens over the next year as points of comparisons are easier to be made.
I look forward to Google Cloud Next ’18, and seeing what the rest of 2017 has in store for Google Cloud.