When a software change is deployed — after being designed, coded, tested, packaged, and tested some more — a journey comes to an end. At the same time, a new journey begins: your customer’s relationship with your service. It’s here, in the domain of operations, that abstract risks like launch schedule slippage give way to tangible risks like lost revenue, degraded trust, and tarnished reputation. Only when it’s available to users can software contribute to (or threaten!) the success of your organization. And so, throughout the past several years, the DevOps Research and Assessment (DORA) project has incrementally deepened our research into the reliability of services, through and beyond deployment, into ongoing operation.

Reliability is a broadly defined term, which refers to a team’s ability to meet their users’ expectations — for software services, it may encompass aspects of availability, latency, correctness, or other characteristics that influence the consistency and quality of user experience. Google’s practice of Site Reliability Engineering (SRE), which has been embraced and extended by a global community of reliability engineering practitioners, is an approach to operations that prioritizes user-oriented measurement, shared responsibility, and collaborative, blameless learning. Starting with the 2021 Accelerate State of DevOps Report, we began asking survey respondents detailed questions about reliability engineering in their organizations. We continued and expanded our investigation in 2022, and found further evidence that modern reliability engineering is widespread: a majority of respondents report that they employ SRE-style practices. With this extensive body of data to draw from, this year we pushed further into analyses of the impact of reliability and its interaction with other dynamics present in our model of technology’s influence on organizational success.

Reliability matters

When reliability is poor, improvements to software delivery have no effect — or even a negative effect — on organizational outcomes

Reliability is more than beneficial: it’s essential. As in prior studies, we find that software delivery performance (as measured by the “four key metrics” of change lead time, deploy frequency, change failure rate, and failure recovery time) is predictive of organizational performance. However, this year’s analysis revealed a previously unseen nuance: the influence of software delivery on organizational performance is predicated on reliability. When reliability is high, high-performance software delivery predicts better outcomes for the organization. But when reliability is poor, improvements to software delivery have no effect — or even a negative effect — on organizational outcomes. This affirms a long-held belief among reliability engineers: “reliability is the most important feature of any system.” If a service or product doesn’t meet its users’ reliability expectations, it’s counter-productive to rapidly ship flashy new features, because users can’t properly experience them. Software delivery relies on a foundation of reliability to create value.

https://storage.googleapis.com/gweb-cloudblog-publish/images/dora.max-900×900.jpg

Reliability is a journey

Any experienced leader will tell you that progress is rarely linear: even with a discipline like SRE, widely practiced and with demonstrable benefits, the path to success is unlikely to follow a straight line. DORA describes the “J-Curve” of organizational transformation, a phenomenon in which durable success comes only after setbacks and lessons learned. This year, we compared the depth of teams’ reliability engineering practices to their impact on the services they provide: will an investment in SRE produce greater reliability? The answer is yes, but with a significant caveat: not at first. Comparing reliability outcomes across a range of levels of SRE adoption, the J-Curve is plainly visible. A team which practices SRE only lightly — at the beginning of their SRE journey, perhaps — is likely not only to not benefit, but to regress in terms of the reliability experienced by their users. However, after these practices have more deeply permeated, an inflection point is reached and we see strong reliability benefits from continuing to grow the reliability engineering capability.

Knowing that it will likely take time to realize the benefits of adopting SRE, it may be tempting to start the process as soon, and as broadly, as possible. But we offer a note of caution here: organization-wide cultural transformation initiatives typically fail from overreach. We studied this and reported findings in a previous report. And even if you manage to beat the odds and fully adopt SRE across multiple teams simultaneously, the cost may be unacceptable: the setbacks in reliability that you are likely to experience early on, amplified across an entire organization all at once, could have catastrophic consequences. Therefore the SRE principle of gradual change should also be applied to the adoption of SRE itself.

Reliability is about people

Reflecting back on over a decade of SRE practice and theory, the Enterprise Roadmap to SRE underlines the importance of culture, suggesting that Site Reliability Engineering is in fact emergent from culture. Tools and frameworks are important; language is essential. But only a trustful, psychologically safe culture can support the environment of continuous learning which enables SRE to manage today’s complex, dynamic technology environments. DORA’s research in 2022 demonstrates the interplay between culture and reliability: we found that “generative” culture, as defined by the Westrum model, is predictive of higher reliability outcomes. And reliability has benefits not only for a system’s users, but for its makers as well: teams whose services are highly reliable are 1.6 times less likely to suffer from burnout.

Got a story to share about your DevOps journey? Submit it to Google Cloud’s 2022 DevOps Awards by January 31, 2023!

Cloud Functions, Google Cloud’s Function as a Service (FaaS) offering, is a lightweight compute platform for creating single-purpose, standalone functions that respond to events, without needing an administrator to manage a server or runtime environment. 

Over the past year we have shipped many new important capabilities on Cloud Functions: new runtimes (Java, .NET, Ruby, PHP), new regions (now up to 22), an enhanced user and developer experience, fine-grained security, and cost and scaling controls. But as we continue to expand the capabilities of Cloud Functions, the number-one friction point of FaaS is the “startup tax,” a.k.a. cold starts: if your function has been scaled down to zero, it can take a few seconds for it to initialize and start serving requests. 

Today, we’re excited to announce minimum (“min”) instances for Cloud Functions. By specifying a minimum number of instances of your application to keep online during periods of low demand, this new feature can dramatically improve performance for your serverless applications and workflows, minimizing your cold starts.

Min instances in action

Let’s take a deeper look at min instances with a popular, real-world use case: recording, transforming and serving a podcast. When you record a podcast, you need to get the audio in the right format (mp3, wav), and then make the podcast accessible so that users can easily access, download and listen to it. It’s also important to make your podcast accessible to the widest audience possible including those with trouble hearing and those who would prefer to read the transcript of the podcast. 

In this post, we show a demo application that takes a recorded podcast, transcribes the audio, stores the text in Cloud Storage, and then emails an end user with a link to the transcribed file, both with and without min instances. 

Approach 1: Building the application with Cloud Functions and Cloud Workflows

In this approach, we use Cloud Functions and Google Cloud Workflows to chain together three individual cloud functions. The first function (transcribe) transcribes the podcast, the second function (store-transcription) consumes the result of the first function in the workflow and stores it in Cloud Storage, and the third function (send-email) is triggered by Cloud Storage when the transcribed result is stored and sends an email to the user to inform them that the workflow is complete.

Cloud Workflows executes the functions in the right order and can be extended to add additional steps in the workflow in the future. While the architecture in this approach is simple, extensible and easy to understand, the cold start problem remains, impacting end-to-end latency. 

Approach 2: Building the application with Cloud Functions, Cloud Workflows and min instances

In this approach, we follow all the same steps as in Approach 1, with a slightly modified configuration that enables a set of min instances for each of the functions in the given workflow.

This approach presents the best of both worlds. It has the simplicity and elegance of wiring up the application architecture using Cloud Workflows and Cloud Functions. Further, each of the functions in this architecture leverages a set of min instances to mitigate the cold-start problem and time to transcribe the podcast.

Comparison of cold start performance

Now consider executing the Podcast transcription workflow using Approach 1, where no min instances are set on the functions that make up the app. Here is an instance of this run with a snapshot of the log entries. The start and end timestamps are highlighted to show the execution of the run. You can see here that the total runtime in Approach 1 took 17 s. 

Approach 1: Execution Time (without Min Instances)

Now consider executing the podcast transformation workflow using Approach 2, where min instances are set on the functions. Here is an instance of this run with a snapshot of the log entries. The start and end timestamps are highlighted to show the execution of the run, for a total of 6 s. 

Approach 2: Execution Time (with Min Instances)

That’s an 11 second difference between the two approaches. The example set of functions are hardcoded with a 2 to 3 second sleep during function initialization, and when combined with average platform cold-start times, you can clearly see the cost of not using min instances.

You can reproduce the above experiment in your own environment using the tutorial here. 

Check out min instances on Cloud Functions

We are super excited to ship min instances on Cloud Functions, which will allow you to run more latency-sensitive applications such as podcast transcription workflows in the serverless model. You can also learn more about Cloud Functions and Cloud Workflows in the following Quickstarts: Cloud Functions, Cloud Workflows.

Today’s enterprise applications are often assembled across distributed environments. This includes the integration of services across multi-cloud, multi-SaaS and on-premises environments. While the approach has the advantage of enabling enterprises to choose the best service available to support their applications, it adds the complexity of delivering services across heterogeneous environments. To solve this, Cloud Load Balancing supports an open cloud strategy, which includes:

• Supporting universal traffic management policies across heterogeneous environment by leveraging open source and open standards
• Enabling a global front-end so applications can leverage a common set of policies and security postures
• Providing tools that give your users the highest possible performance and reliability

Universal traffic management with open source and open standards

Kubernetes is a great solution for managing containers across environments. We believe that traffic management policies should also be supported across environments. Cloud Load Balancing creates homogeneous traffic policies across highly distributed heterogeneous environments by supporting standard-based traffic management in a fully managed solution, and allowing open source Envoy Proxy sidecars to be used on-premises or in a multi-cloud environment, using the same traffic management as our fully managed Cloud Load Balancers.

As enterprises start modernizing services and refactor monolithic applications, they require solutions that can provide consistent traffic management across distributed systems at scale. But organizations want to invest their time and resources innovating and building new applications — not on the infrastructure and networking required to deploy and manage these services. Envoy is an open-source high-performance proxy that runs alongside the application to deliver common platform-agnostic networking capabilities, including:

• New load balancing algorithms, (e.g. round robin, ring hash, least conns)
• Weighted traffic splitting
• Fault injection
• Request mirroring
• Outlier detection
• Additional header transformation options
• Request retries
• Additional backend session affinity options
• Cross-origin resource sharing (CORS)

Hybrid Load Balancing across multi-cloud and private clouds

Over the years Google has deployed Load Balancers across 173+ Edge Pop locations, delivering customer applications at massive-scale on Google infrastructure. And now Google Cloud has introduced Hybrid Load Balancing, extending our Load Balancing capabilities beyond Google’s network to on-premises private clouds and multi-cloud solutions. This allows our customers to migrate applications to the cloud iteratively, or build hybrid applications that are assembled from services that are running across heterogeneous environments.

Supporting modern application delivery with HTTP3/QUIC

Cloud Load Balancing is a fully distributed load balancing solution that balances user traffic (HTTP(s), HTTPS/2, HTTPS/3 with gRPC, TCP/SSL, UDP, and QUIC) to multiple backends to avoid congestion, reduce latency, increase security, and reduce costs. It is built on the same frontend-serving infrastructure that powers Google services, supporting millions of queries per second with consistent high performance and low latency.

To serve massive amounts of traffic, Google built the first scaled-out software-defined load balancing, Maglev, which has been serving global traffic since 2008. It has sustained the rapid global growth of Google services, and it also provides network load balancing functions for Google Cloud Platform customers. To accommodate ever-increasing traffic, Maglev is specifically optimized for packet processing performance with Linux Kernel Offload. Maglev is also equipped with consistent hashing and connection tracking features, to minimize the negative impact of unforeseen faults and failures on connection-oriented protocols.

Another key enabler to support this global-scale is that our Cloud Load Balancers are built on top of QUIC(RFC9000), a protocol developed from the original Google QUIC) (gQUIC). HTTP/3 is supported between the External HTTP(S) Load Balancer, Cloud CDN, and end clients. And once enabled, customers typically see dramatic improvements in performance and throughput.

Google Cloud already supports HTTP3 in Cloud Load Balancer. To use HTTP/3 for your applications, you can enable it on your external HTTPS Load Balancers via the Google Cloud Console or the gCloud SDK with a single click.

If your service is sensitive to latency, QUIC will make it faster because it establishes connections with reduced handshakes. When a web client uses TCP and TLS, it requires two to three round trips with a server to establish a secure connection before the browser can send a request. With QUIC, if a client has connected with a given server before, it can start sending data without any round trips, so your web pages will load faster.

QUIC has advantages over legacy TCP as follows.

Summary

Since 2008, Google has been an innovator in software-defined networking, supporting applications running at massive scale. Google Cloud Load Balancers support HTTP3 and QUIC as a next generation web transport protocol, which significantly improves customer traffic latency. Google Load Balancers also have incorporated the Envoy proxy as a foundational technology, providing our customers with advanced traffic management that’s compatible with the open source Envoy ecosystem. This allows our users to have the choice to combine Google’s fully-managed Cloud Load Balancers with open source Envoy Proxies, to enable consistent traffic management capabilities across a multi-cloud distributed environment. And with Hybrid Load Balancing, customers can leverage our 173+ world wide PoPs to seamlessly manage traffic across Google Cloud, on-premises and other cloud providers.

Google Cloud Load Balancers include all these capabilities natively. And when used together, they support globally-scaled applications that run seamlessly across the heterogeneous environments many enterprises deploy today.

Editor’s note: Today’s post is written by Hatim M, Chief Commercial Officer at BungkusIT. The on-demand delivery service delivers packages within the hour for one million customers across Malaysia and uses Google Maps Platform to create a seamless end-to-end delivery experience for its customers.

Imagine it’s the end of a long day at work, and you’re heading home for dinner as you look forward to a restful evening. Then just as you arrive at the doorstep, your phone buzzes with a message from a family member, asking if you could grab a carton of milk from the supermarket.

With moments like these in mind, we founded BungkusIT to alleviate the stress of everyday chores and give people a break from having to run seemingly mundane errands that can often be time-consuming. With so much to do and so few hours in a day, we want to make life simpler by providing our customers with an on-demand delivery service.

As a social enterprise, our mission is to create jobs and increase earning potential for BungkusIT roadies (those who run deliveries), while also increasing online visibility for  local small and medium enterprises through an e-commerce platform. As such, we don’t charge merchants a fee for being featured on our app. 

Our roadies help connect users with products and services that they need. Whether it’s picking up the keys you forgot, or buying a birthday cake for a loved one, our on-demand roadies can help accomplish your task quickly and efficiently. In fact, we’ve helped many seniors—a large and often underserved population—to run errands and provide last mile support.

Providing reliable service, 24/7

From the day we launched, our promise has always been to complete every task given to us within one hour. Google Maps Platform empowers us to fulfil this promise. We understand the importance of selecting a reliable mapping service from the get-go. After exploring different map products, we made the decision to go with Google Maps Platform due to its detailed and accurate location mapping system. With the support of our partner Searce, we’ve been able to quickly integrate all the relevant products on our platform and optimize our API calls.

On a day-to-day basis, our roadies rely on the Places API to navigate their journeys. It pinpoints the exact location set by customers so that our roadies can be sure they arrive at the correct destination. With so many small roads and alleys in Malaysia, it can be challenging to find the exact place based on an address. Dropping a pin on a specific location removes any confusion.

Because we price our services based on distance, it’s business critical to determine the exact distance of a location and the most efficient route to reach it. By automating that process with the Distance Matrix API, we can guarantee that customers don’t get overcharged, while making sure that we’re adequately compensated for our services.

Improving the user experience

At the end of the day, we want to make people’s lives easier. That’s why having an app that is intuitive is key. More than that, we want customers to be able to create a request as quickly as possible. By automatically completing the location a customer is typing after they’ve keyed in the first few characters, Place Autocomplete does that job for us. 

As a fairly young company, we’ve got a long way to go. I believe that we’ll continue evolving alongside Google Maps Platform and exploring the new features it has to offer. We’re already talking to global brands to expand our services and I’m confident that with Google Maps Platform as our maps partner, we will go a long way.

For more information on Google Maps Platform, visit our website.