Like virtually every business across the globe, The New York Times had to quickly adapt to the challenges of the coronavirus pandemic last year. Fortunately, our data system with Google Cloud positioned us to perform quickly and efficiently in the new normal. 

How we use data

We have an end-to-end type of data platform; on one side we work very closely with our product teams to collect the right level of data that they’re interested in, such as which articles people are reading, and how long they’re staying onsite. We frequently measure our audience to understand our user segments, and how they come onsite or use our apps. We then provide that data to analysts for end-to-end analytics. 

On the other side, the newsroom is also focused on audience, and we build tools to help them understand how Google Search or different social promotions play a role in a person’s decision to read The New York Times, and also to get a better sense of their behavior on our pages. With this data, the newsroom can make decisions about information that should be displayed on our homepage or in push notifications. 

Ultimately, we’re interested in behavioral analytics—how people engage with our site and our apps. We want to understand different behavioral patterns, and which factors or features will encourage users to register and subscribe with us. 

We also use data to create or curate preferences around personalization, to ensure we’re delivering to our users fresh content, or content that they may not have normally read. Likewise, our data also gets used in our targeting system, so that we can send out the right messaging about our various subscription packages to the right users.

Choosing to migrate to Google Cloud

When I came to The New York Times over five years ago, our data architecture was not working for us. Our infrastructure was gathering data that proved harder for analysts to crunch on a daily basis. We were also hitting hang ups with how that data was streaming into our system and environment. Back then we’d run a query and then go grab some coffee, hoping that the query would finish or give us the right data by the time we came back to our desks. Sometimes it would, sometimes it wouldn’t.

We realized that Hadoop was definitely not going to be the on-premises solution for us, and that’s when we started talking with the Google Cloud team. We began our digital transformation with a migration to BigQuery, their fully managed, serverless database warehouse. We were under a pretty aggressive migration timeline, focusing first on moving over analytics. We made sure our analysts got a top-of-the-line system that treated them the way that they themselves would want to treat the data. 

One significant prominent requirement in our data architecture choice was to enable analysts to be able to work as quickly as they needed to provide high-quality deliverables for their business partners. For our analysts, the transition to BigQuery was night and day. I still remember when my manager ran his very first query on BigQuery and was ready to go grab his coffee, but the query finished by the time he got up from his chair. Our analysts talk about that to this day.

While we were doing the BigQuery transition, we did have concerns about our other systems not scaling correctly. Two years ago, we weren’t sure we’d be able to scale up to the audience we expected on that election day. We were able to band-aid a solution back then, but we knew we only had two more years to figure out a real, dependable solution. 

During that time, we moved our streaming pipeline over to Google Cloud, primarily using App Engine, which has been a flexible environment that enabled quick scaling changes and requirements as needed. Dataflow and Pub/Sub also played significant roles in managing the data. In Q4 of 2020 we had our most significant traffic ever recorded, at 273 million global readers, and four straight days of the highest traffic we’ve had compared to other election weeks. We were proud to see that there was no data loss.

A couple of years ago, on our legacy system, I was up until three in the morning one night trying to keep data running for their needs. This year, for election night, I relaxed and ate a pint of ice cream because I was able to more easily manage our data environment, allowing us to set and meet higher expectations for data ingestion, analysis and insight among our partners in the newsroom.

How COVID-19 changed our 2020 roadmap 

The coronavirus pandemic definitely wasn’t on my team’s roadmap for 2020, and it’s important to mention here that The New York Times is not fundamentally a data company. Our job is to get the news out to our users every single day in paper, on apps, and onsite. Our newsroom didn’t expect the need to build out a giant coronavirus database that would enrich the news they share every day. 

Our newsroom moves quickly, and our engineers have built one of the most comprehensive datasets on COVID-19 in the U.S. With Google, The New York Times decided to make our data publicly available on BigQuery Google’s COVID-19 public dataset. Check out this webinar for more details on our evolution architecture:https://www.youtube.com/embed/mtNlrFpschU?enablejsapi=1&

Flexible approach

We have many different teams that work within Google Cloud, and they’ve been able to pick from the range of available services and tailor project requirements keeping those tools available in mind.

One challenge we think about with the data platform at The New York Times is determining the priorities of what we build. Our ability to engage with product teams at Google though the Data Analytics Customer Council allows us to see into the BigQuery roadmap, or the data analytics roadmap, and plays a significant role in determining where we focus our own development. For example, we’ve built tools like our Data Reporting API, which reads data directly from BigQuery, in order to take advantage of tools like BigQuery BI Engine. This approach encourages our analysts to be better managers of their domains around dimensions and metrics, but not have to focus on building caching mechanisms of their data. Getting that kind of clarity helps us plan how to build The New York Times in the new normal and beyond.

If you are interested to learn more about the data teams at the New York Times, take a look at our open tech roles here and you’ll find many interesting articles at NYT data blog.

As retailers strive to meet the growing expectations of shoppers, they are turning to Google Cloud to transform their businesses and tackle opportunities in an increasingly challenging industry. From optimizing inventory management to increasing collaboration between employees across locations and roles, to helping build omnichannel experiences for their customers, we are working together with retailers to help make the shopping experience as seamless and personalized as possible.  

A standout Google Cloud customer is Macy’s, one of the world’s largest retailers. Founded in 1858, Macy’s operates approximately 680 Macy’s and Bloomingdale’s, and 190 specialty stores including Bloomingdale’s The Outlet, Bluemercury and Macy’s Backstage. And through macys.com, bloomingdales.com, and bluemercury.com, it also serves millions of customers across more than 100 countries. 

By moving its infrastructure to the cloud, and taking advantage of Google Cloud data warehousing and analytics solutions, Macy’s is streamlining retail operational functions across its network.

With the opening of its new approximately 675,000 square-foot distribution center in Columbus, Ohio, Macy’s is leveraging the scalability of Google Cloud to ensure that merchandise is accurately and efficiently received, sorted, ticketed, picked, packed and shipped from the distribution center to the stores—even during peak retail seasons like back to school and the holidays.

“Powered by software developed at Macy’s technology, this new distribution center is a fantastic first step in our cloud journey. Working with Google Cloud allows us to be more nimble, efficient and flexible in how we utilize our warehouses,” 

Naveen Krishna, CTO, Macy’s

Leveraging Google Cloud’s data management and analytics solutions, Macy’s new warehouse management system will initially service 200+ Macy’s Backstage off-price stores at launch. Macy’s will begin rolling out this software solution to additional distribution centers that service its nationwide fleet of Macy’s and Bloomingdale’s department stores, as well as Macys.com and Bloomingdales.com direct-to-customer orders. 

Our continued work with Macy’s reflects their investment in technology to improve digital and mobile experiences, site stability, store technology, fulfillment, and logistics, and integrate its front line and back office to reinvent retail. I look forward to deepening our partnership with Naveen and his team to help them achieve these goals. 

Throughout the past couple of years, I have helped a good number of companies, big and small, migrate their systems to the Google Cloud Platform (aka GCP). During the course of these migrations, there are always a few of those moments where people look at a specific Google Cloud feature and say, “now, that’s cool!”.

More often than not, it is because, coming from other platforms, they have gotten used to some features requiring multiple steps, or some operations being complicated, etc. And often they find out that in GCP you can do this specific operation in a couple of clicks, or by setting up a simple text-based configuration. Then you see that light bulb turning on in their head, and there you go… happy customer.

A few of these happen so often that I compiled them in a list to share with others who might also benefit from these “aha!” moments. You could say these are the five things I wish they told me when I started using Google Cloud.

Projects: Naturally Group Resources Together

A project is a namespace where resources live. Every resource you instantiate in GCP, from load balancers to Kubernetes clusters to virtual machines, belongs to a single project, and has no access (by default) to resources in other projects. User roles and authorisations can be defined per-project and trickle down to everything in it. This has two immediate benefits: you can group things that belong together in neat logical units, and things that don’t belong together are isolated from each other (and isolation is a Good Thing)

This is powerful and quite simple, but it often takes new users off-guard. I’ve had many clients call me and ask me “How can I make sure my developers cannot access the production machines? What’s the best way to create access policies? ”

The answer to this is actually super-simple:

• have a project for development where your developers have rights,
• have a project for production where they don’t.
• That’s it.

Every machine/other resource in the production project won’t be accessible to developers.

Of course there is a lot more to it, and you can refine roles and permissions to a much greater degree using Organizations, Folders, etc. Not to mention all the crazy things you can do with per-project billing. But at least you can say “hey, if it’s a machine in the staging environment then it can be found in the “staging” project”.

Global Virtual Networks Are *Truly* Global

Imagine you are using a Cloud provider and that you have servers in the US, and servers in Singapore, and that they need to communicate.

So you create a VPC (Virtual Private Cloud) network in the US data center, another one in the Singapore data center, and then you will connect them by setting up inter-region VPC peering or a VPN (Virtual Private Network) or a transit VPC or other routing magic.

Lots of work, right? And many moving parts, so lots of opportunities for things to break.

With GCP, however, what makes my clients go “aha!” is when they realize that in GCP a single VPC network covers the entire planet. Only subnets are attached to a geographic location, and virtual machines communicate between subnets on private IPs (good old RFC1918 addresses) — no extra routing needed.

So, to make your server communicate across continents on GCP, here are the steps:

• create a VPC network
• create a subnet in the US, put your US servers in it
• create a subnet in Singapore, put your Singapore servers in it

That’s all there is to it. Your VPC network spanning 2 continents is ready to use. Below is a screenshot of how it looks on my account, for a VPC network called ‘my-global-network’ with 2 subnets. The first column (“us-central1” and “asia-southeast1”) contains the name of the GCP regions (read: data centers). The second column is the subnet name that I picked when I created them.

A machine in the US (on the “us-central” subnet) with IP 10.0.0.5 can communicate directly with a machine in Singapore (on the “singapore”) subnet with IP 10.10.0.8.

Nothing else to set up.

And thanks to the way these networks work, the Google Cloud Load Balancer can present a single IP to the world, and forward traffic to the instances that are the closest to you geographically without having to setup a tedious DNS-based load balancing. But that’s worth an entire blog. I’ll save it for another day.

Firewalls with Tags and (Almost) No IP Addresses

There is no network security without a firewall so unsurprisingly GCP comes with one built-in.

Now, I don’t know about you, but nothing makes my brain hurt like a list of firewall rules displaying IP ranges and addresses and ‘Allow/Deny’ directives. It looks a bit like this:

If you imagine a normal network with a few dozen (hundred?) servers, you can quickly see how this can get out of control. You’d better have a solid printout of your network layout to refer to when you start adding and changing rules. And good luck debugging things!

Wouldn’t it be nice if, instead, you could just tell the firewall: “the HTTP traffic from outside can only reach the HTTP servers and the MySQL database is only reachable by the HTTP server(s) on the same network?”

Turns out it’s pretty simple on GCP by using a little thing called network tags. As the documentation says:

“Network tags are text attributes you can add to Compute Engine virtual machine (VM) instances. Tags allow you to make firewall rules and routes applicable to specific VM instances.”

So let’s see how it works. Firewall rules in GCP are defined in terms of source and target (the traffic flows from the source to the target). You can define filtering rules that apply to the source or the target, and in both cases you can use tags.

This is simpler shown with an example. The rule below states that on the default network, the traffic to the VMs with the tag mysql-server can come from the VMs with the tag http-appserver. Any other traffic is “Deny”-ed by default.

All you have to do is to tag your machines properly, and they will automatically be covered by the rule. You don’t need to enter their IP range.

That’s neat if you ask me. It makes it a lot simpler to grasp what’s happening.

Of course, there’s a TON more to firewalls in GCP. Tags also apply to routes and you can mix and match IP-based rules with tag-based rules. Not to mention that thing called service accounts, but I’ll leave those for another day.

The bottom line is that you can create most rules by just expressing a business need and not having to remember complicated network layouts. I have no hard stats, but I’m pretty sure this has saved me hours of work.

Console Access to VMs from the Browser

Easily access virtual machines (VMs) from the Google Cloud console was one of my first “aha!” moments when I started using GCP.

This is a screen capture of my Google Cloud console, with a virtual machine and its internal IP.

The last column has a header that says “Connect” and when you click on the word “SSH” a separate windows pops up. You wait for a few seconds, and… this is what you get. Your personal shell access — in a browser popup no less.

You are connected through ssh to the virtual machine of your choice. You did not have to download ssh keys and put them in the ~/.ssh directory, do the correct chmod command and run a long-winded ssh -i ~/.ssh/somekey me@<it-took-me-forever-to-copy-paste-the-address-here>

In addition, you have access to a few nifty features such as uploading and downloading files, changing the user etc. Just use the menu behind the cog icon at the top right.

In truth, you should not need to connect directly that often, but when you have to, this is a godsend.

Your Personal Jumphost from the Google Cloud Console

The Google Cloud console has a cool trick: you can actually connect to a virtual environment that is managed by the Google Cloud console itself. It serves a bit as a jump host. You can access most resources from the projects from it, and you can activate it directly from the top menu with, no particular setup on your side. It’s called the Cloud Shell.

This is how it looks at the top right of the console:

When you activate the Cloud Shell, the session opens at the bottom of the console. You get a command line prompt and it’s fully configured with the gcloud command line tool (the jack-of-all-trades of Google Cloud scripting).

You can do a great many things from there, and this even includes uploading and downloading files, editing code or deploying it, a web preview for your AppEngine application, and more.

So you can get access to a fully configured shell environment in your project from any laptop where you can connect with your credentials. On top of this, it persists between connections so you can fine-tune it to your needs and have these changes available the next time you re-connect.

This has saved me many times during my previous life as a traveling consultant!

Live migration

Did I say 5 “Aha!” moments ? Well, you’ve been patient reading all the way to here, so here’s one more for free.

Google Cloud has an amazing way to literally “teleport” a running virtual machine between physical hosts without stopping it. It’s called Live Migration. It allows Google to move your virtual machine away from a defective host, or a host that needs a patch or an upgrade, or for any other infrastructure related reason.

It’s all done in the background, and is totally transparent, so you never really see it happening. Unless you look VERY closely. I once did a demo to a client, where a machine was live migrated while he was simulating a solid network load — and we did not lose a single packet, with no noticeable degradation in latency.

And that’s a wrap!

So there you go. These are 5+1 things that made me go “Aha!” when I became more familiar with the Google Cloud Platform, and that still make my clients do the same.

There is a lot of depth to the platform, and my examples above only scratch the surface of our features. I encourage you to try it yourself. There is a generous free tier, and when you are ready to take the plunge and create that new company, please contact us at Google Cloud for Startups. We’ll get you up and running in no time.

Jerome is a Startup Architect at Google Cloud. Based in Singapore, he helps startups make the most of the Google Cloud Platform.

Scale-out workloads demand the best combination of performance and price to bring down the cost of delivering applications, all while providing an excellent user experience. We are excited to announce a new virtual machine (VM) family, Tau VMs, coming to Google Cloud. Tau VMs extend Compute Engine’s VM offerings with a new option optimized for cost-effective performance of scale-out workloads. 

T2D, the first instance type in the Tau VM family, is based on 3rd Gen AMD EPYC^TM processors and leapfrogs the VMs for scale-out workloads of any leading public cloud provider available today, both in terms of performance and workload total cost of ownership (TCO). The x86 compatibility provided by these AMD EPYC processor-based VMs gives you market-leading performance improvements and cost savings, without having to port your applications to a new processor architecture. 

As illustrated below, Tau VMs offer 56% higher absolute performance and 42% higher price-performance (est. SPECrate2017_int_base) compared to general-purpose VMs from any of the leading public cloud vendors.

What our customers and partners are saying

Snap
“At Snap, it is critical for our business to continue improving our scale-out compute infrastructure for key Snapchat capabilities like AR, Lenses, Spotlight and Maps,” said Cody Powell, Senior Engineering Manager, Snap Inc. “We were impressed when we tested Google Cloud’s new Tau VMs with Google Kubernetes Engine. While it’s early days, we believe we can gain double digits in infrastructure performance improvements for key workloads—enabling us to do more with less and invest even more in new features for our amazing Snapchat community.”

Twitter
“High performance at the right price point is a critical consideration as we work to serve the global public conversation,” said Nick Tornow, Platform Lead, Twitter. “We are excited by initial tests that show potential for double digit performance improvement. We are collaborating with Google Cloud to more deeply evaluate benefits on price and performance for specific compute workloads that we can realize through use of the new Tau VM family.”

DoiT
“DoiT partners with leading cloud vendors who are focused on growth and cost optimization,” said Yoav Toussia-Cohen, CEO, DoiT International. “In our preliminary testing of Google’s new Tau VMs with the Coremark benchmark, we were thrilled to see the incredible performance at 50% better than a comparable ARM-based offering from another leading public cloud. With Tau VMs, Google Cloud has set a new bar for price-performance, making the cloud even more accessible to digital-native companies. We are excited to bring Google’s Tau VMs to our joint customers.”

Designed for demanding scale-out workloads 

Tau VMs bring the benefit of Google’s long-standing experience engineering platforms for scale-out workloads to our customers. They come in multiple predefined VM shapes, with up to 60vCPUs per VM, and 4GB of memory per vCPU. They offer up to 32 Gbps networking bandwidth and a wide range of network attached storage options, making Tau VMs ideal for scale-out workloads including web servers, containerized microservices, data-logging processing, media transcoding, and large-scale Java applications. 

Google Kubernetes Engine support

Google Kubernetes Engine (GKE) is the de facto standard for organizations looking for advanced container orchestration, delivering the highest levels of reliability, security, and scalability. GKE supports Tau VMs on day 1, helping you optimize price-performance for your containerized workloads. You can add Tau VMs to your GKE clusters by specifying the T2D machine type in your GKE node-pools. 

Pricing

Tau VMs will be priced to support significant TCO and price-performance improvements for your cloud applications. A 32vCPU VM with 128GB RAM will be priced at $1.3520 per hour for on-demand usage in us-central1. 

Coming soon to a Google Cloud region near you

If you are interested in trying out T2D VMs when they become available in Q3 2021 please sign-up here.