Hello cloud gurus and welcome to
GCP this month for February, 2020. I’m your host, Mattias Andersson and this month we’ll
cover a number of enhancements to GKE as well as some things that improve
running Microsoft workloads on GCP and sometimes both at the same time.
We’ll also look at dataflow, dialogflow and GPU’s, oh my! Plus
a bunch of compute engine releases, so let’s get right to it. Google’s managed service from Microsoft
active directory is now generally available. You can let Google run and completely
manage actual Microsoft 80 servers for you. A lot of organizations use Microsoft
active directory so this can smooth their migrations to GCP and
reduce their ops burden. Cloud SQL for SQL server is now
completely out of beta globally. This means that you can
have Google manager, Microsoft SQL server 2017 database
servers for you with a 99.95% uptime SLA. Dataflow monitoring has gotten much
better now that it includes charts and alerts for both CPU utilization
and element throughput. The best thing is that these are broken
down for each step of your pipeline. You can then zoom into interesting
parts of the timeline to help diagnose issues. Now you have even more
reason to go with the flow, right? This next update is pretty intense. The dialogflow mega agent now supports
up to 20,000 intense dialogflows, a platform for building chatbots and
interactive voice responses and it’s what powers Google’s contact center AI product. Now that it can handle 10 times as many
intenses before your virtual agent can be set up to guide your users
through tons more situations. Dialogflow has also gotten
some improvements to help
you develop better agents more reliably. In case you already use Nvidia T4 GPUs
or you were maybe considering them for your AI or 3d workloads, you should know that Google has just
slashed their prices more than 60% across the board. These GPUs offer incredible value and a
benchmark run on ML engine showed these T4s costing less than half as
much as the next closest GPU type. Google’s new N2D machine type family
based on AMD Epic processors is now in beta. These are general purpose instances
and offer all the same compute engine features as the Intel
based and two instances, but at a lower price and
with higher performance. N2Ds are available right
now in US central one, Asia Southeast one and Europe West four. Google has released improvements to how
you can manage sole tenant nodes such as you might need for stringent compliance
or licensing requirements if you need to minimize your physical core usage because
of windows licensing requirements. You’ll appreciate the new support for
live migration within a fixed node group and if you’re auto-scaling VMs, you’ll have new auto scaling support
to add or remove sole tenant nodes as needed. Finally, you can now migrate VMs into and out
of sole tenancy by stopping them, updating their note affinity
and restarting them. Sadly, there’s no live migration
support for that last case, but it’s still much better than
having to recreate them from scratch. Historically, managed instance groups or MIGs have
only been a good match for stateless workloads. We’re creating and
destroying instances is no big deal, but now Google has gone into
beta with the stateful MIGs. In which instances keep all of their
persistent disc and metadata the same. This can be great for things like
database nodes or those stateful legacy applications that I’m sure you’ll get
around to replacing any day now, right? Config connector is a new GKE plugin
that can let you manage GCP resources through Kubernetes,
including RBAC and events. How config connector works is that it
provides a collection of Kubernetes, custom resource definitions
and controllers for them. Then those will create and update your
Google cloud resources when you configure and apply custom objects to your
cluster application manager. Another new enhancement to
basic GKE offers an opinionated
and simple way to set up your app with a GitOps based CI/CD
flow that follows many of Google’s best practices. It uses tools like Customize internally
and because it is easily scriptable, you can call it from build tools like
cloud build or Jenkins. Check it out. But now let’s take a look
at this months GCP Gems. So have you heard the
news? Well, probably not, which is why you’re watching this
video, right? Well, Stackdriver is dead! Long live Stackdriver! Okay. This
might seem a bit grim at first, but let me explain. Back in 2014 Google acquired the
Stackdriver company and product. Ever since then, they’ve
been doing two things. One, adding significant new
products and functionality
to the Stackdriver family and Two, integrating more and more
of the Stackdriver console
directly into the normal GCP console. Well, now that ends. All of the Stackdriver
functionality remains, but every last bit of it has finally
been fully integrated into the normal GCP console and the Stackdriver
brand is being, well, retired might be a good term I guess. So let’s take a look at what
this means for terminology. Stackdriver logging is
known as cloud logging. This change happened a while
ago, but it’s where we start. Stackdriver monitoring
is now cloud monitoring. Stackdriver trace is now cloud
trace, Stackdriver debugger is now, well you get the picture right, pretty much just replace Stackdriver
with cloud. But wait, there’s more. Not only is cloud monitoring
now completely integrated
into the Google cloud console, it also gets some
very nice upgrades. For one, Metrics will now be
retained for two full years, meaning you can compare much further back
when investigating issues. And second, metrics will now be written at
up to a ten second granularity, meaning you can get quicker and more
detailed information. And that’s not all. There’s also a new API for
creating and managing dashboards, alerts routing through
Pub/Sub and workspaces that
let you monitor and control hundreds of GCP projects all together. Follow our link to the blog post
to check out all the details. Google has just released into beta support
for running Windows server containers on GKE, even in the same cluster
as your Linux containers. Now such a cluster will need to have a
node pool for Windows server nodes in addition to one for Linux nodes, but this setup can reduce your Windows
servers licensing costs by running multiple windows container applications
on each Windows server node, even with windows server nodes
and containers in your cluster. GKE still supports things
like node auto upgrades, private clusters and regional clusters
and more Windows specifically. It also supports group managed service
accounts and the ability to select the Microsoft servicing channel that works
best for you. In their announcement, Google also highlights
some GCP partnerships with
companies that support CI/CD container management of windows based
workloads on GKE, including CircleCI, CloudBees, and GitLab. There’s definitely value here for
organizations that have containerizable Windows server workloads.
So what do you think? Are Windows and GKE better together? How about cloud sequel
and Microsoft SQL server? Stackdriver and the GCP
console? All of these are great, but it is my admittedly biased opinion
that the best better together story is still A Cloud Guru and Linux
Academy joining forces. And one of the upcoming benefits of this
is that you’ll see more Google cloud news brought to you by the talented
training architects from both original organizations, so be
sure to keep tuning in. Finally, we have our
GCP guru of the month. This month’s winner is Adarsh Prasad,
a senior cloud DBA based in Singapore. Congratulations. We’ll be sending you
some awesome swag, including a snazzy, a cloud guru t-shirt, cool
stickers, and a hand signed card. And all the rest of you should check out
this month’s question available at the link below. And remember,
we choose the best response, not just the first one. So you
do all have a chance. Well, that’s all for this month’s episode. I look forward to seeing many of you in
person at Google cloud next 2020 in San Francisco in April. If you see me
or any of the other Gurus attending, be sure to flag us down and say hi. We love to connect with our
students like that. Also, just a heads up that the
session guide is now posted, so don’t forget to plan your time in
advance so you get the most out of the event. And until our next
episode of GCP This Month, keep being awesome cloud gurus.