Understanding the Meta, Comcast, and LinkedIn Outages


Google Cloud NEXT ‘18: Rainbow Unicorns and Multi-Cloud

By Angelique Medina
| | 8 min read


A few weeks ago at ThousandEyes Connect in Santa Clara, in addition to great speakers from Atlassian and Caesars, we had the pleasure of welcoming Joshua Cha from our partner Google Cloud to present on their cloud vision and specifically Google Cloud Interconnect.

So it was apropos that we in turn were able to give a talk at Google Cloud NEXT Partner Summit last week. My colleague Ameet Naik and I spoke at the Monday session, Cracking the Code: Building a Multi-cloud Network Strategy.

Google Cloud NEXT '18 photo
Figure 1: Google Cloud NEXT '18 at the Moscone Center in San Francisco.

Let Me Google “Multi-Cloud” For You

Multi-cloud is becoming a really significant enterprise challenge. Modern applications are highly atomized and distributed, and applications are getting built around cloud-based API services, like Google Maps. So a lot of inter-service communication ends up happening across regions and even across cloud providers, which means that you’re now faced with a complex matrix of network paths. Enterprises have to consider not only where to place their workloads, but also how they are going to interconnect all of these services—over the public Internet or connecting through an exchange provider.

The reality is that many enterprises don’t understand how to manage performance in this environment. Questions like “Is it the network, or the application” become that much harder to answer in the cloud, because there are so many different domains and paths. Most enterprises are also not prepared for the operational shift that needs to happen when you move to the cloud. Because everything changes. You’re no longer relying on infrastructure that you own and manage. When you own the infrastructure, if you can find the problem, you can fix it. In the cloud, you’re relying on thousands of networks and services that you don’t have any control over—many ISPs, DNS services, CDNs, SaaS and public cloud providers. You have to have very clear evidence of who’s at fault, so you can get an escalation and avoid finger pointing.

During our Google Cloud NEXT session, we covered how you would go about addressing these challenges using data. Without data, you can’t answer critical questions around your cloud deployment, like:

  • What cloud regions should host my workloads?
  • How should I connect to my cloud providers?
  • Should I directly peer with the cloud provider?
  • Should I use an overlay to connect between different clouds?

When you have deep network visibility that’s also app-aware, you can get the data you need to make informed network architecture decisions that will support your business goals. It also enables you to tune your decisions over time and quickly troubleshoot issues when they occur. This is the essence of why we’re partnering with Google Cloud, to contribute visibility that helps Google customers make the best possible decisions and to operate effectively in the cloud.

For more multi-cloud goodness, check out Ameet’s blog post on Monitoring Multi-Cloud Performance.

A Few Gems from the Sessions

Following our talk on Monday, Ameet and I had the opportunity to attend several Google Cloud NEXT sessions throughout the week. There’s lots to unpack from the show and many takeaways, but the one that we found to be particularly interesting was an announcement at the session, A Year in GCP Networking, that GCP will be breaking out their network transport into “Standard” and “Premium” network tiers. Before the announcement, all traffic was effectively treated as “Premium,” in that it was allowed to ride Google’s backbone. Following this announcement, unless you’re paying for premium, your traffic will get hot-potato routed to a transit provider, rather than riding Google’s backbone. What this means is that much more GCP user traffic will be transported via the public Internet. While any network is subject to performance issues and outages, the Internet can be a hairy place, and figuring out when and where something’s gone wrong can often be very challenging. This announcement is an important reminder that there’s no steady state in the cloud. You have no guarantee that traffic will flow a certain way just because of how it did yesterday or even an hour ago.

Other highlights from the sessions included great GCP customer stories. The New York Times detailed their GCP journey, Scotiabank discussed how they autoscale their apps, and Target took the stage multiple times to share their GCP story.

See, Understand and Optimize Your Cloud Journey

This is an exciting time for cloud adoption. What we’ve seen is that enterprises who successfully operate in the cloud take a lifecycle approach, where they implement a readiness phase that will help them make informed decisions around connectivity, set performance expectations, and begin to adopt new processes for dealing with external providers. Enterprises often don’t have the data or methodology to realize a multi-cloud strategy, but with visibility and deep insight, you can make the foundational choices that will ensure cloud success—and have the visibility you need to resolve issues effectively.

As a Google Cloud partner, we’re really proud to be helping cloud consumers successfully deploy and operate in the cloud. Hopefully, we’ll see you at next year’s Google Cloud conference—enjoying a very CLOUDy San Francisco summer.

Subscribe to the ThousandEyes Blog

Stay connected with blog updates and outage reports delivered while they're still fresh.

Upgrade your browser to view our website properly.

Please download the latest version of Chrome, Firefox or Microsoft Edge.

More detail