Seamless, scalable and secure internet connectivity between applications and users is top of mind for all businesses as they adopt cloud. Starting October 1st, 2023, all new and existing customers of Google Cloud Networking Standard Tier will be eligible for 200 GB of free internet data transfer every month. This offer is available independently in each region where customers operate. Standard Tier is available in 28+ Google Cloud regions, amounting to roughly 6 TB of potential data transfer at no charge every month. This makes it easier for startups and small businesses as they evaluate Google Cloud.

Understanding Google Cloud internet data transfer

Google Cloud offers customers two Network Service Tiers with which to connect their resources to the internet: Standard and Premium Tier.

Standard Tier is a secure, reliable and performant way to connect VMs to the internet. By primarily leveraging the public internet to route traffic, Standard Tier offers customers the opportunity to transfer data between their VMs and internet end-points, without traversing Google Cloud’s network backbone. 

As shown below, when a VM connects to an internet end-point over Standard Tier, traffic exits Google at the nearest PoP (Point of Presence) to that Google Cloud region i.e., travels the shortest distance within our network. The traffic then routes over the public internet via BGP to reach its destination.

In contrast, Premium Tier networking carries traffic over Google Cloud’s global network backbone to the edge PoP nearest to the end-user. As a result, the traffic benefits from the premium performance, reliability and security offered by Google’s private network; this is especially appealing to customers with applications that have these requirements. 

Comparing Standard vs. Premium Tier

In addition to 200 GB of free monthly egress, we are excited to announce that Standard Tier is now fully integrated with key Google Cloud networking services like Cloud NAT and Regional Application and Network Load Balancing. 

This table compares the two offerings:

Pricing

Each Standard Tier SKU corresponds to a single region, as listed here. This offer is available separately for each region. After exhausting the 200 GB within a month in a given region, regular tiered rates apply to the traffic until the meter resets on the first of the next month. No action is required from customers and the changes will automatically be reflected starting from the October billing statement.

By providing Standard Tier customers with 200 GB of internet data transfer per region, per month, we are enabling more customers to run their apps and transform with  Google Cloud. To learn more about Standard Tier Networking and how it might fit your needs, please refer to this page.

As the CIO for the U.S. Navy Ship Repair Facility (SRF) in Japan, Peter Guo leads a culturally diverse team that’s charged with supporting the success of global fleet operations. His multi-generational, multi-lingual team has a mix of different motivations for joining SRF. During the Google Workspace Summit earlier this year, Guo shared some of the ways he is helping drive cultural transformation with a multi-generational workforce. 

Guo notes that cultural transformation is core to driving innovation in support of the mission. “People are your most valuable asset,” Guo stressed. “We need people, processes, and technology working together, in that order of importance, to drive cultural change. Investing in technology solves today’s problems. Investing in processes solves next week’s problems. But investing in personnel solves your ‘forever’ problems.”

With the use of Google Workspace, Guo has witnessed how technology can break down barriers to improve organizational collaboration. His team uses Google Workspace not just for basic chat and call capabilities, but to help break down language barriers between Japanese and U.S. personnel. Like Google, Guo believes the effective use of technology can help drive cultural change — especially as more millennials join the workforce. When millennials join public sector teams, “they bring experience using the latest and greatest technologies and processes with them,” Guo explained. By being open to new and diverse experiences, the SRF finds new, innovative ways of working together to solve those “forever problems.”

Using Google Workspace in an Emergency Situation

One innovative way the SRF team stays operationally ready is through role-playing how they would respond in an emergency situation — using Google Workspace tools. One recent example the crew chose to simulate was responding to a Navy ship fire.  

The NAVSEA team arrived at the Tokyo office where they were briefed on Google Workspace, the cloud-based collaboration tools they’d be using to support the immersive exercise. Afterward, the team broke into pods to brainstorm solutions for their mock scenario, with their Captain pushing participants to consider solutions that could engage critical groups from ship crews, to local firefighters, to the General in the office. 

The pods generated several innovative ideas using Google Workspace:

One team proposed using AppSheet to perform future safety checks. This could allow leadership to quickly make decisions based on aggregate safety reports. 

Another team suggested using video translation services to provide real-time reporting via Google Meet in non-native languages. 

A different team came up with a solution to utilize Google Forms and Google Sheets to track over 1,000 items of Base Firefighting inventory, along with Google Maps to trace inventory location. 

These proposed solutions offered new ways to solve the problems outlined in their scenario, but the experience also offered insight into the way the team works together. Creating this immersive, real-time mock situation helped the NAVSEA team build their confidence, establish stronger connections, and strengthen readiness in the case of a real emergency situation.

At Google Cloud Next ‘23, the Cloud Networking Load Balancing team announced multiple enhancements that unlock new use cases and increase your value when using Google Cloud Load Balancing. The four of the marquee features we introduced are:

mTLS support adds client-side authentication during TLS negotiation on global external Application Load Balancers. This capability allows the server to verify the client’s identity in the same way that the client verifies the server’s identity during standard TLS authentication. Thus, mTLS provides an additional security layer that helps protect against man-in-the-middle attacks and other threats.

Service Extensions callouts allow users to customize the data-plane processing of select load balancers with user-created or third-party applications. This customization empowers customers to add new capabilities and optimize how traffic is handled for their Google Cloud workloads, which improves the user experience.

Cross-region internal Application Cloud Load Balancer supports global backends! In other words, the internal load balancer can now distribute load to backends spread globally. Further, global access is built-in, that is clients from any region can access the internal Application Load Balancers. This capability introduces a new level of flexibility when it comes to architecting the hosting of backends or the location of clients.

Cross-project service referencing with global external Application Load Balancers allows organizations to configure Application Load Balancers to route traffic to hundreds of services distributed across multiple different projects. Cross-project service referencing relies on Shared VPC, which allows the connecting of resources from multiple projects to a commonVirtual Private Cloud (VPC) network so that they can securely and efficiently communicate with one another.

Additional details regarding each of these exciting new capabilities follow:

mTLS support on Google Cloud external load balancers

Authentication with HTTPS communication typically works only one way: the client verifies the identity of the server. For applications requiring the load balancer to authenticate the identity of clients that connect to it, both the global external Application Load Balancer and the classic Application Load Balancer support mutual TLS (mTLS), now in preview.

With mTLS, the load balancer requests that the client send a certificate to authenticate itself during the TLS handshake with the load balancer. You can configure a trust store that the load balancer uses to validate the client certificate’s chain of trust.

mTLS offers many benefits, including:

Increased security: mTLS adds an additional layer of security by requiring both the client and the server to be authenticated, which makes it more difficult for attackers to impersonate either party and gain access to sensitive data.

Reduced risk of man-in-the-middle attacks: Man-in-the-middle attacks (MITM) are a type of cyberattack in which an attacker intercepts and redirects traffic between two parties. mTLS helps to mitigate the risk of MITM attacks by requiring both parties to be authenticated.

Improved visibility: mTLS can improve network traffic visibility by allowing organizations to track which clients are connecting to their servers. This can help identify potential security threats and improve the overall security posture.

In addition to these benefits, you can use mTLS with a variety of applications including Apigee X. As a result, mTLS is becoming increasingly popular as a way to improve the security of network communications.

Service Extensions callouts on Application Load Balancers

Service Extensions callouts bring a powerful tool to inject custom logic into Google Cloud Application Load Balancers. This custom logic can address unique customer workflow requirements, offer an easy option for partners to integrate their software with Google services, or help organizations implement Cross-Cloud Network. This capability allows Google Cloud customers or partners to write their own code to perform various activities on traffic processed by a load balancer such as header rewrites, incremental security, custom logging, and custom user authentication.

With Service Extensions callouts, you instruct Google Cloud Load Balancing to forward traffic from within the load-balancing data-processing path via an RPC such as gRPC, to a user-managed application or service running anywhere. These applications can apply various policies or functions before returning the traffic to the load balancer for further processing. 

Benefits of Service Extensions callouts include:

Bespoke implementation – Traffic handling is tailored to address unique workflow requirements and can optimize the performance of cloud applications or services. 

Empowers users – Organizations can develop their own applications or purchase programs to change how a service is delivered to support new or custom requirements.

Agile delivery – Users can independently create and deploy service extensions anytime.

Easy partner integration – Partners can quickly and programmatically integrate their software with Google Cloud Application Load Balancer services and deliver new advanced use cases.

Availability: The public preview of Service Extensions callouts for Cloud Load Balancing will start in October 2023. Supported Application Load Balancers include the global external (not including Classic), regional external, and regional internal.

Cross-region internal Application Load Balancer

Internal Application Load Balancers distribute HTTP and HTTPS traffic to backends hosted on Compute Engine, Google Kubernetes Engine (GKE), Cloud Run, on-premises, and other clouds. Until now, Internal Application Load Balancer has been regional in its scope, i.e., the load balancer could connect with backends in the same region only. Cross-region internal Application Load Balancer removes this constraint.

A cross-region Internal Application Load Balancer is a fully Google-managed service and can be deployed as a multi-region load balancer. It provides global access out of the box; that is, clients from any Google Cloud region can access the load balancer. In addition, cross-region mode enables users to load balance traffic to backends that are globally distributed. This load balancer also enables high availability; placing backends in multiple regions helps insulate users from failures in a single region. If one region’s backends are down, traffic can gracefully fail over to another region. Further, given the load balancer can be deployed in multiple regions, it is resilient to outages that can potentially impact the availability of the load balancer service in a region.

It is important to note that the load balancer proxy instances in cross-region mode are always deployed to the specific regions you choose. 

To summarize, benefits of cross-region internal Application Load Balancers include:

Enables global Cloud Load Balancing for internal traffic: This capability brings new flexibility to how traffic can be distributed via an active/active multi-region setup.

Increased high availability: Availability is improved since the impact of any zone- or region-specific issues can be reduced with failover to services in a different region.

Unlocks Private Service Connect (PSC) use cases: Private Service Connect allows consumers to access managed services privately from inside their VPC network.

Support for Google-managed certificates: This support is achieved using Certificate Manager and Certificate Authority Service.

The cross-region internal Application Load Balancer is now in preview. To learn more, please visit the Internal Application Load Balancer page. 

Cross-project service referencing on global external Application Load Balancers

Cross-project service referencing allows organizations to configure one central load balancer and route traffic to hundreds of services distributed across multiple projects. You can centrally manage all traffic routing rules and policies in one URL map. You can also associate the load balancer with a single set of hostnames and SSL certificates. 

In this model, the load balancer’s frontend and URL map are in a host or service project. The load balancer’s backend services and backends can be distributed across projects in the Shared VPC environment. Cross-project backend services can be referenced in a single URL map.

With support for the global External Application Load Balancers, now in preview, cross-project service referencing is now available on all application load balancers including internal Application Load Balancers and regional external Application Load Balancers.

Benefits of cross-project service referencing include:

Efficient deployment: Optimize the number of load balancers needed to deploy your application by connecting multiple services to a single load balancer. With fewer forwarding rules and other load balancing resources, you not only incur lower costs, but also reduce your operational overhead and quota requirements.

Improved administrative control: By having different projects for each functional team, customers can also achieve separation of roles within their organization. So service owners can focus on building services in service projects, while network teams can provision and maintain load balancers in another project, and both can be connected using cross-project service referencing.

To learn more about Cross-project Service Referencing, please visit the external Application Load Balancer page.

Generative AI is driving the next phase of cloud transformation for communications service providers (CSPs). Although we are still in early days, significant progress has been made towards realizing the promise of gen AI in telecommunications. Google Cloud has been broadening and enhancing the capabilities of its foundation models, democratizing gen AI usage by simplifying model development, enhancing the reliability of gen AI responses, and addressing integration with enterprise systems. 

At the same time, CSPs have been creative in developing a broad range of applications — and in some cases, they have started to put these applications into production use. In this blog, we explore some of the key gen AI developments from Google Cloud and how they benefit CSPs, and provide high-level architectures for some common gen AI applications.

A portfolio to support the different needs of gen AI centric enterprise development

Google Cloud offers a range of foundation models for processing and generating text, code, and images:

These models offer control to AI practitioners through APIs or Generative AI Studio, letting them, for example, control the model used or the length of text segment submitted to the model. Vertex AI unlocks gen AI development by automating several integration and configuration tasks, and can be used by developers with no AI background. Google Cloud’s Vertex AI Search can process long documents for developers, meaningfully accelerating the development and deployment of high-value intelligent apps. For all our recent developments, be sure to check out our gen AI sessions from Next ’23.

CSP gen AI use case momentum 

As we partner with our customers on their gen AI journeys, we have seen a large number of CSP-specific applications being pursued in areas including:

Network operations,

Employee experience, 

Customer retention and average revenue per user (ARPU) growth.

CSPs are rapidly deploying gen AI in their daily business to summarize, categorize, or identify the sentiments in texts, generate email or web content, and answer questions by extracting the information from large corpora of documents using foundation models or products like Vertex AI Search and Conversation. Using these solutions, they are increasing employee productivity and creativity while achieving greater customer engagement and satisfaction. 

Let’s take three CSP applications and explore the applicability of their high-level architectures for other applications. 

Network outage root cause analysis using Vertex AI Search

Analyzing the root causes of a network failure requires mining through several logs, vendor documents, past trouble-tickets, and their resolutions. Vertex AI Search enables CSPs to build a chatbot that can extract relevant information from both structured and unstructured data, and significantly shorten the time for a human engineer to identify probable causes.

The implementation architecture above shows the simplicity in deploying this application. Thousands of documents can be loaded and indexed in a few hours with just a few clicks, with a chatbot ready for testing. 

A number of CSPs have used the above architecture to develop chatbots in various domains. For example, Vodafone is experimenting with Vertex AI Search. Vodafone Voice and Roaming Services has over 10,000 contracts with other telecommunications companies worldwide in a variety of formats, and searching this document repository is often a time consuming process for employees. “We are building an intelligent assistant to securely and quickly search contracts,” said Sherif Bakir, CEO of Vodafone Voice and Roaming Services, about Vodafone’s use of Vertex AI Search. “With generative AI, we’re accelerating otherwise protracted processes, increasing productivity and operational efficiency.” 

Empowering employees with information and automation

Google Cloud’s Contact Center AI (CCAI) has helped CSPs globally to enhance customer satisfaction and efficiency, while lowering call center costs. CCAI’s Dialogflow CX is a tool for developing complex interaction flows between a user and a virtual agent. And CCAI Agent Assist can suggest turn-by-turn smart responses to human agents that can be fine-tuned for specific enterprise needs. It also provides customizable summaries for voice and chat sessions, and can determine the emotional intent of the caller through sentiment analysis. 

Organizations can now leverage Vertex AI Search and Conversation, and gen AI capabilities within Dialogflow, making it possible for CSPs to develop flows that include access to unstructured and structured information, as well as trigger actions such as scheduling a maintenance window or a technician visit using webhooks. It is also possible to generate a complex Dialogflow process from a natural language specification.

These growing capabilities are transforming the way CSPs deliver outstanding customer support. For example, “Bell is excited to explore generative AI and its applications across personalization, marketing, and customer service,’’ says Michel Richer, Vice President, Data Engineering and Artificial Intelligence at Bell. “We look forward to building on our work over the past few years using conversational analytics, with CCAI as a key driver for customer experience improvement. Bell strongly believes that the developments in generative AI are going to be transformative and is looking forward to innovating and partnering in this space with Google Cloud.” 

Attract and retain customers and increase ARPU using Google Telecom Subscriber Insights, PaLM, and Imagen

Google Cloud’s Telecom Subscriber Insights extracts customer information from various data sources, and helps CSPs identify products and services that their subscribers can benefit from while increasing their ARPU. We are also integrating Vertex AI to enable a chatbot to improve the sales experience and targeted offers while interacting with customers in a natural dialog. Such targeted offers to customers also increase customer engagement and reduce churn. 

Google Cloud and its partners have been enabling CSPs to reap such benefits from smart customer data platforms. CSPs can now significantly enhance the speed and effectiveness of such offers by generating rich content with text, images and video for their specific sales campaigns (for example, offering 5G-based home internet service) in minutes, using simple and effective prompts. The messages to specific customers could be tailored based on their location, past buying patterns, and any insights into their current sentiments. 

Imagen can create high resolution images from user’s textual prompts in various styles (for example, photo-realistic or watercolor). Furthermore, a new method of tuning for Imagen, called Style Tuning, enables CSPs to create images aligned to their specific brand guidelines or to include their unique products or designs in the generated images with a small amount of reference images. Similar few-shot text model tuning could be used to tailor the text to a CSP’s corporate style. In addition, Google DeepMind SynthID offers a state-of-the art technology that embeds a watermark directly into the image of pixels, making it imperceptible to the human eye and difficult to tamper with. 

Reaping the benefits while managing the risks

We have explored enterprise-ready gen AI capabilities offered by Google Cloud in our current product portfolio, and we continue to be excited by the range of high-value use cases that CSPs are developing with these capabilities. Underpinning all of this, Google Cloud is committed to addressing the accuracy, privacy and security, regulatory compliance, and intellectual property infringement risks of gen AI, implementing strong guard rails as we adopt and implement gen AI technologies. 

We hope this blog has inspired you as you continue on your gen AI journey! Many of the foundation models, Vertex AI Search and Conversation, andTelecom Subscriber Insights are already generally available to all Google Cloud customers. You can also learn more about how Google Cloud is partnering with CSPs to deliver their cloud transformation here.

In today’s business landscape, it’s clear that the distributed workforce is the norm and is here to stay. As employees access company data and applications from different remote locations, safeguarding enterprise security becomes paramount. 

Chrome Enterprise is committed to security at its core, and our secure-by-design approach enables us to protect enterprise users as they work on the web. In the last few years, we have worked with different security solution providers to provide seamless integrations between their solutions and Chrome Enterprise. These integrations further improve safety features in addition to Chrome Enterprise’s built-in security, providing secure enterprise browsing for businesses.

Chrome Enterprise and Ping Identity for context-aware access control

Today, we’re proud to share that the Chrome Enterprise and Ping Identity Device Trust integration is now available to provide enterprises with context-aware access control to determine user access level based on the device security posture. By combining Ping’s user context and risk signals with Chrome Enterprise’s device signals, security teams can build enhanced policies to ensure users are requesting access from a managed browser, are given a step-up for multi-factor authentication (MFA), or denied access. The signals can also be used to trigger certain actions, such as forcing an update of the device or notifying the user about necessary steps to secure their endpoint.

Get the most out of your secure enterprise browser

Chrome Enterprise is investing heavily in integrations with partner solutions to help organizations maximize their security stack. For example, Chrome Enterprise enables a higher level of secure access evaluation with a seamless experience between Ping Identity and CrowdStrike.

If your organization also uses CrowdStrike Falcon Zero Trust Assessment (ZTA), you can combine it with the Device Trust integration to enhance the authorization decisions with the Ping Identity platform. Chrome browser can collect the CrowdStrike Agent ID from managed devices and share it with Ping Identity. If the device score indicates that the device is high risk, Ping Identity will utilize the high risk score for dynamic access evaluation. This trilateral integration protects businesses by automatically blocking devices identified as high risk from accessing businesses applications and resources.

Getting started

Getting started with the integration is easy. The first step is to make sure that the devices used by your organization are enrolled in either Chrome’s cloud management tool or in ChromeOS device management. With these tools, you can manage Chrome browser and/or ChromeOS from a single, cloud-based Admin console. Setting up is simple. Check out this guide for Chrome’s cloud management tool and see this guide on how to enroll ChromeOS devices.

Once you have your machines enrolled, you can then follow the setup guide or watch the setup video below:

Chrome Enterprise is dedicated to helping businesses work more securely on the web. We continuously partner with leading security providers to deliver secure enterprise browsing solutions, enabling organizations to build best-of-breed technology stacks to meet their unique needs. Enterprises can count on Chrome Enterprise Recommended partner solutions to support their workforce, wherever they work.

Learn more about Chrome Enterprise security here.

Helpful links to get you started:

Help Center Article for Device Trust Connectors Get Started Guide for Chrome Device ManagementSetting up Chrome’s cloud management toolHelp Center Article for Chrome Enterprise Connectors Framework

We are pleased to announce that Assured Workloads on Google Cloud now supports International Traffic in Arms Regulation (ITAR) capabilities in nine U.S. regions. As part of this announcement, SAP National Security Services (SAP NS2®), a secure cloud services provider for government and defense agencies, has achieved ITAR compliance for SAP Ariba, their end-to-end procurement solution, on Google Cloud with Assured Workloads.

ITAR is a set of U.S. government regulations administered by the U.S. Department of State’s Directorate of Defense Trade Controls (DDTC). These regulations control the export of defense articles and defense services and restrict the handling of software and technical data controlled on the United States Munitions List (USML). 

At times, ITAR compliance can be challenging to achieve, as it requires a deep understanding of the business processes that classify ITAR relevant data and applying data residency and access controls to said data. With today’s achievement, Google Cloud supports customers with ITAR-controlled software or technical data through security controls provided through Assured Workloads.

Assured Workloads helps customers achieve a seamless DevOps experience across multiple compliance programs by providing granular control over deployments on Google Cloud. Features such as data residency, cryptographic controls, Assured Support and Assured Workloads Monitoring provide a full suite of capabilities to manage the lifecycle of deploying a workload to meet compliance requirements. Additionally, Assured Workloads can be deployed as part of a customer’s existing Google Cloud environment, as customers do not need to operate two different consoles in order to deploy ITAR workloads.

SAP Ariba connects millions of trading partners worldwide. The platform supports multi-trillion dollar procurement activities, and it is used by many government and defense agencies. SAP Ariba provides real business process integration. It implements a smooth source-to-pay process that integrates with your pay processes, back-end systems, and trading partners to deliver optimal value. SAP NS2 customers that adopt the SAP Ariba network to manage procurement processes for defense contractors must comply with ITAR to meet DDTC’s requirements. 

The combination of Assured Workloads on Google Cloud, International Traffic in Arms Regulation (ITAR) capabilities, and SAP Ariba provides government and defense agencies a secure pathway to the cloud.

“SAP Ariba on Google Cloud provides a foundation for us to continue to grow our secure cloud portfolio to our Public Sector customers. Google Cloud allows us to achieve high execution velocity for compliance, and as a result SAP customers will soon be able to deploy SAP S/4HANA Cloud, private edition on Google Cloud with ITAR compliance,” said Lillian Chang, SVP, Product Strategy, SAP NS2. “This is a testament to Assured Workloads ability to offer a cost-effective way to achieve compliance, its operational simplicity, and the seamless DevOps experience between varying compliance programs.”

Today, our customers can achieve compliance requirements for more than 10 international compliance programs that are already available through Assured Workloads from Google Cloud. For more information, visit: https://cloud.google.com/assured-workloads or sign up for a Free Trial today.

Educators, researchers, IT professionals, student developers, and C-suite leaders showed up in full force in August for Google Cloud Next, our global showcase for the latest cloud technologies and success stories. They attended breakout sessions, product demos, and keynotes to explore the potential of new tools like generative AI and gain insights into how it is already transforming teaching, learning, and advanced research across the country. 

Gen AI offers particular benefits for the education community, which faces constant demands for speed and scale in their key missions of driving scientific breakthroughs and preparing the next generation for the workforce. It can generate cost efficiencies–the time to value for insights– for both research and operations. It can improve user experience for students, staff, and faculty. It can accelerate content development and customize large models to analyze the ever-growing mountains of data generated on campuses everywhere.

Here are some of the top takeaways from the education sessions and demos:

At Secure Your Organization with Zero Trust Controls and Google AI: IT leaders at the University of Notre Dame demonstrated how they used AI to gamify cybersecurity training for over a thousand students. By designing an interactive AI-based Google Wizarding College and a Cybersecurity Carnival they made security training fun and effective. “Security is everyone’s responsibility,” Chas Grundy, Director of IT Strategy and Transformation, says. “Google tools help us set defaults and automate controls to make that burden easier on students and admins.”

The panelists at Responsibly Driving Student Success with AI began by pointing to some of the challenges facing higher education today: the learning losses in math and reading since the pandemic and the overwhelming need to retrain tomorrow’s workforce in new skills. Lev Gonick, Enterprise Chief Information Officer at Arizona State University, believes AI has the potential to recruit and retain students who haven’t been served by traditional education. Virtual tutors and personalized learning can help the 40 million students with Some College, No Credential (SCNC) reach their career goals. COO Miguel Amigot II of IBL Education described how their gen AI-based mentors can provide a one-to-one learning experience to students and professionals looking for tailored skills development. In collaboration with professors, IBL Education builds AI-based modules that can help students with everything from time management skills to social interactions. Anant Agawral, founder of edX, already uses AI to translate edX’s 4,000 online courses into over a dozen languages. “That used to be expensive,” he says, “but now the translations are so fast and so good that we can just focus on distributing them efficiently.”

Google Customers Form Bio and Collegis Education are using AI to make research and student services more efficient. By analyzing how and when students, faculty, and staff interact with online services, Collegis can determine the best times and methods to reach each cohort. Form Bio builds AI into their scientific data platform to help researchers focus on their results. Making computational workflows user-friendly can save time and resources–and accelerate discoveries.

Framing the three-day conference were two special events for higher educational professionals: the Higher Education Innovation Board on August 28 and the San Francisco AI & Research Day event on August 31. These events brought together researchers, CIOs, and IT leaders from campuses across the country to discuss how they are serving their communities, modernizing their infrastructure, and transforming their organizations with the latest cloud tools.

Here are some of the key lessons learned at the higher ed events:

At the Higher Education Innovation Board event, Matt Gunkel, CIO of the University of California, Riverside (UCR), shared the story of how UCR is transforming their operations and services through a pioneering fixed-cost agreement with Google Cloud. Now in its second year, Gunkel could point to significant impact already: researchers are moving workflows to the cloud with HPC Toolkit and some are accelerating their data analysis from 24 hours to 15 minutes. Soon the team intends to deploy gen AI and BigQuery to analyze Student Information System, Learning Management System, and financial data for qualitative insights into student satisfaction.

At San Francisco’s Google Cloud AI & Research Day participants heard from Google experts and C-suite leaders about how Google Cloud technologies can facilitate collaboration, automate security and data governance, and make research and operations more efficient. They watched demos of how gen AI can sift through scientific literature to answer complex queries and generate laboratory protocol–or “ground” research citations in a customer-provided database. Stay tuned for upcoming AI & Research Days hosted at Google offices in other cities.

All the Google Cloud Next ‘23 recorded sessions are now available on-demand. Click here to start exploring the best of Next. To find out how you can get started with gen AI for higher education, sign up for an interactive half-day workshop with Google Cloud and partners Nuvalence and Carahsoft. Participants will work with experts in small groups to design a gen AI strategy package customized for their needs. To learn more about funding opportunities, check out the eligibility for cloud training and academic research credits.

Editor’s note: Today’s article comes from Rishabh Sood, Chief Technology Officer, BluSmart. He shares how BluSmart, India’s first all-electric cab-hailing startup, transformed the ride-hailing experience with Google Maps Platform.

Hailing a cab from our phones used to feel magical. But over time, the service we experienced in India began to degrade to a point that it made users anxious. Customers could no longer rely on estimated arrival times, or use their preferred payment mode. When we set out to build BluSmart, our goal was simple: to restore the magic into the on-demand cab-hailing experience, with a commitment to sustainability.

To address these needs, we returned to the drawing board and adopted a unique, full-stack playbook: BluSmart’s all-electric cabs are housed at strategically-located charging hubs so customers don’t pay surge fees irrespective of weather, and don’t have to wait beyond the promised ETAs for their rides. We lease our own vehicles and offer driver partners regular weekly earnings, so they have a more predictable income and do not need to maintain their cars. More than seven million rides and 15,000 tons of CO2 savings later, we now plan to expand to several more cities, and triple our fleet’s size, with Google Maps Platform playing a key role in our journey.

Optimizing driver ETAs

Prior to implementing Google Maps Platform, we focused on optimizing the ‘dry run’ — the time a driver spends traveling to the customers and looking for their pickup location instead of driving them to their destinations. A long and unpredictable dry run hurts everyone: the rider waits longer, the driver earns less, and BluSmart has fewer free cabs on the road. After implementing Google Maps Platform’s Mobility Services, we could plan our demand more effectively, and reduce our dry run time by virtue of having more accurate ETAs and thus having to keep lower time buffers.

We tap into Google Maps Platform’s historical and geospatial insights to understand the traffic and weather conditions of our most frequented routes. This allows us to proactively map our network of driver partners, so when there’s a new request, we can dispatch the closest one, and arrive at the rider’s doorstep within the promised estimated time of arrival.

Cultivating a safer cab ride

Google Maps Platform has made the BluSmart experience smoother and slightly safer for our driver partners. Previously, they had to juggle between the BluSmart and Google Maps apps during trips, which was unsafe and slowed down our drivers mobile devices. Together with Lepton Software, we integrated Google Maps Platform turn-by-turn navigation right inside the BluSmart app with the Navigation SDK, allowing driver partners to follow directions and keep tabs on their ride’s context from one app experience.

Though we already transport thousands of passengers daily, we’re still a startup, and each day is a new challenge and milestone, from onboarding numerous new electric vehicles to expanding overseas. There’s a lot for us to learn but it’s reassuring, for me especially as a CTO, to have the support of a solid technology like Google Maps Platform in this journey. It’s helped us better deliver on our promise of a superior and frictionless cab-hailing experience by solving key problems around ETAs and locations. Moving forward, we’ll continue to innovate, and come up with new ways we can empower our customers and driver partners. 

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

Google Cloud’s AlloyDB is the next-generation managed PostgreSQL service that is designed to handle cloud-scale operational and analytical workloads. It has several autopilot features that allow it to self-update and self-tune, including automatic memory management and adaptive autovacuum. This blog focuses on how AlloyDB’s adaptive autovacuum feature mitigates PostgreSQL VACUUM challenges and runs the process in the most efficient and uninterrupted manner.

PostgreSQL’s Multi Version Concurrency Control (MVCC) enables multiple transactions to occur simultaneously without blocking each other by creating multiple versions of each row (tuples). It achieves this by creating multiple versions of each row, where each version corresponds to a different point in time. In MVCC, each transaction is assigned a unique Transaction ID (XID), which represents the order of the transaction’s execution. PostgreSQL’s MVCC causes two issues.

1. Transaction ID wraparound: PostgreSQL transaction IDs or XIDs are 32-bit unsigned integers that are assigned to each transaction and also get incremented. When they reach their maximum value, it would wrap around to zero (similar to a ring buffer) and can lead to data corruption. 

2. Bloat: The large accumulation of obsolete data in tables, indexes, and system catalogs leads to bloat. This over time reduces database performance as the query planner’s accuracy is impacted and also the read operations have to go through more pages. 

To address these issues, PostgreSQL has a VACUUM process. When invoked manually, it scans the tables and eliminates inactive tuples and updates table statistics. Users usually configure AUTOVACUUM, an automated background process in PostgreSQL that triggers the VACUUM process based on parameters such as autovacuum_vacuum_cost_limit, autovacuum_vacuum_cost_delay, vacuum_cost_page_hit, vacuum_cost_page_miss, vacuum_cost_page_dirty.

However, there are some of the challenges of PostgreSQL AUTOVACUUM:

The default autovacuum settings may not be sufficient for all workloads. It can be challenging to determine the optimal settings for a particular database and even specific tables. 

The autovacuum process can be a source of contention, especially on busy systems. If not properly configured or managed, the autovacuum process can have a negative impact on performance. 

The autovacuum has fixed resource budgets (cost limit, number of vacuum worker processes, and memory). It cannot automatically adjust vacuum workload based on the customer workload and available resources. 

For databases with very high transaction rates where the VACUUM process is not able to keep up, that can lead to increase in open XIDs, which will eventually lead to a XID wrap-around situation and cause long system downtime. 

VACUUM that lags behind can also cause table space bloat with a lot of dead tuples and index entries. This increases unnecessary storage usage, which in turn impacts backup and restore times. This also affects query performance.

Autovacuum settings need to be adjusted carefully to suit the workload to avoid availability and performance issues. But it can be difficult to adjust vacuum settings, if the workload changes all the time.

AlloyDB’s adaptive autovacuum

AlloyDB for PostgreSQL is designed to handle mission-critical operational and analytical workloads. Large and dynamic workloads make it difficult to manually tune the autovacuum settings. AlloyDB’s adaptive autovacuum is a feature that automatically adjusts the frequency of vacuuming and analyzes operations based on the workload of the database. This helps to ensure that the database is always running at peak performance, even as the workload changes, without any interruption from the vacuum process. AlloyDB’s adaptive autovacuum goals are to:

Ensure reliable and consistent application transactional performance

Maintain high availability of the system by avoiding XID wraparound problems

Enable a hands-off approach to vacuum tuning, freeing the DBAs from having to manually tune the settings for each workload

Honor any autovacuum settings updated/tuned by users and adjust adaptive settings accordingly

AlloyDB’s adaptive autovacuum process monitors and updates the autovacuum-related PostgreSQL parameter values in real time. For example, multiple autovacuum workers can run concurrently on different tables, controlled by the autovacuum_max_workers parameter. The maintenance_work_mem parameter value defines the work memory used by each autovacuum worker. AlloyDB adjusts these parameters dynamically.

How does adaptive autovacuum work?

AlloyDB Adaptive autovacuum uses a number of factors to determine the frequency of vacuuming and analyze operations, including:

The size of the database

The number of dead tuples in the database

The age of the data in the database

The number of transactions per second vs estimated vacuum speed  (for XID throttling, see #2 below)

The following are the adaptive autovacuum improvements and automatically adjusted settings in AlloyDB:

1. Dynamic vacuum resource management: Instead of using a fixed cost limit, AlloyDB uses real-time resource statistics to adjust the vacuum workers. When the system is busy, the vacuum process and resources are throttled. If enough memory is available, additional memory is allocated for vacuum workers to accelerate index vacuum.

2. Dynamic XID Throttling: AlloyDB automatically and continuously monitors the progress of vacuuming and the speed of XID consumption. If a risk of XID wraparound is detected, AlloyDB will gently begin to throttle XID consumption by slowing down transactions. It also allocates more resources to vacuuming so that vacuuming can catch up and return to the safe zone. During this process, the overall transactions per second will be reduced until the XIDs are in the safe zone. When the XID age increases, the vacuum workers are dynamically increased.

3. Efficient vacuuming for larger tables: The default vacuum is based on table-specific statistics stored in pg_stat_all_tables which has the dead tuple ratio. That works for small tables and may not work efficiently for larger, frequently updated tables. AlloyDB has an updated scan mechanism that helps trigger the autovacuum more often that scans chunks of large tables and helps remove dead tuples more efficiently. 

4. Log warning messages: In AlloyDB, the vacuum blockers such as long-running transactions, orphaned prepared transactions, orphaned replication slots etc., are detected and warnings are registered in the PostgreSQL logs so that users can handle the cases in a timely manner.

“Found a backend process XXX with a long running transaction whose transaction id age XXX is larger than or equal to the transaction age threshold XXX.””Found an old prepared transaction XXX whose transaction id age XXX is larger than or equal to the transaction age threshold XXX, database oid: XXX, owner oid: XXX””Found a replication slot XXX whose min transaction id age XXX is larger than or equal to the transaction id age threshold XXX.”

Enabling AlloyDB adaptive autovacuum

Adaptive Autovacuum is enabled by default, but can be disabled (or enabled again in the future) using the flag enable_google_adaptive_autovacuum.  

Benefits of AlloyDB adaptive autovacuum

AlloyDB adaptive autovacuum is designed to adjust the vacuum process based on the workload’s real-time resource utilization and without requiring users to tune the vacuum parameters. However, users can still tune the autovacuum-related parameters, and will AlloyDB honor the their settings. 

There are a number of benefits to using adaptive autovacuum, including:

Increased performance: Adaptive autovacuum helps to ensure that the database is always running at peak performance by removing the bloat, even as the workload changes.

Reduced maintenance: Adaptive autovacuum automatically adjusts the frequency of vacuuming and analyze operations, so you don’t have to worry about it.

Improved availability: Adaptive autovacuum helps to prevent XID wraparound, and thus avoids database outages and increases availability.

In our internal AlloyDB benchmarking tests with 100% cached TPC-C runs on 64vCPU SKU, by enabling Adaptive Autovacuum, we observed on average the Frozen XID age reduced from ~1.5B to ~1B and dead tuples count dropped from ~1.2B to 0.7B, allowing the database to perform optimally over a longer period of time.

Conclusion

Adaptive autovacuum is a powerful autopilot feature that manages the vacuum process efficiently without manual intervention and can help to improve the performance and availability of your AlloyDB database. If you’re not already using adaptive autovacuum, we encourage you to enable it today. 

To learn more about Adaptive Autovacuum, read Configure Adaptive Autovacuum documentation.

To learn about AlloyDB, read AlloyDB for PostgreSQL intelligent scalable storage | Google Cloud Blog

Start building on Google Cloud with $300 in free credits and 20+ always free products. https://cloud.google.com/free

Web3 is meant to usher in a new era for the internet. More accountable and open, it promises to make our networked world more trustworthy and inclusive. But in recent years, the technology that powers Web3 has become increasingly complex. This is especially true in the case of blockchain, the ledger that records transactions for participants in the decentralized finance (DeFi) ecosystem, and the transactions finished by complex interactions of smart contracts, the “lego” of DeFi.

All this complexity, in turn, makes it harder to discover DeFi trading strategies, as  exemplified by how difficult it is to decode and calculate Maximal Extractable Value (MEV), a measure of the value that can be extracted from a blockchain network. For DeFi and Web3 to flourish, we must tame the intrinsic complexity of smart contract trades and make transactions like MEV more transparent for stakeholders, including analysts, institutions, and traders.  

This was the motivation that led me to found Eigenphi. In the past two years, we have developed market-leading transaction analytical tools detailing MEV and liquidity on-chain data, as well as bespoke research that enables DeFi stakeholders to understand blockchain behaviors, structures, and affiliations that affect their interests. For example, we are currently providing MEV evaluation to one of the biggest crypto wallets, empowering their MEV-aware design of order-flow processing. Our approach enhances execution qualities for order flow auction providers, and we’re excited to share our appreciation for DeFi to crypto communities.

A compact team with global reach

We’re a small startup with 20 people. Most are based in Beijing, but we have small teams in the U.S., Canada and Singapore serving audiences in Asia, Europe, and North America. Our IT infrastructure must be equally global, which is one of the reasons that we decided to migrate our systems to Google Cloud, which is available in most regions and delivers industry-leading availability and security.

Another reason for migrating to Google Cloud is that we process massive volumes of trading data for real-time tables and charts on our website. Our algorithms untangle dozens, sometimes hundreds, of transfers in one transaction. This process requires fast and reliable data crunching abilities such as those provided by Google Cloud’s data management and analytics tools: BigQuery,Bigtable andCloud Spanner for data storage, andDataFlow andPub/Sub for data streaming and processing. We also useFirebase for application development andCloud Functions to run our code.

The Google for Startups Cloud Program also gave our business a boost. We got firsthand guidance from Google Cloud experts who helped us migrate to the cloud and select the tools that best matched our data processing needs. They also connected us with the wider Google Cloud community so that we could speak at conferences, boost our marketing, and get advance notice of new Google Cloud features and releases.

Participants in the Google for Startups Cloud Program can also access Google Cloud credits for many products and tools. This helped reduce financial pressure at a time when we needed to grow fast and demonstrate success to our investors. Another financial advantage is that we didn’t need to invest in an infrastructure team, keeping down our staffing costs. Google Cloud provides the infrastructure and support we need.

Although it’s still early days, the results are beyond our greatest expectations. Because we can focus on business and innovation, we have reduced our costs by 70 percent and cut deployment time by 90 percent. Now we see potential forVertex AI and other Google Cloud machine learning tools to uncover more insights in our blockchain data. In addition, we saved 60% of cost and achieved 2x efficiency moving from a data pipeline based on virtual machines (VMs) to Compute Engine. We’re thrilled with the benefits of a modern cloud-native architecture vs. a traditional implementation. 

Advice for aspiring entrepreneurs

We are a highly specialized organization, but there are two important lessons that other entrepreneurs can take from our experience. The first is that no matter how attractive your business model is, it means nothing if you can’t find or afford the technology to make it a reality. We had a clear vision for analyzing transaction data generated by blockchain to reveal the value inside, but this required vast amounts of data storage and processing, mostly in real time. Google Cloud provided the best, most affordable tools for the job.

The second is to stay focused. When we first started, we were slightly distracted by all the business opportunities that blockchain data offers. We had to make a firm decision to focus on analytical tools that measure MEV as the entry point of liquidity trading. This is also where Google Cloud played a critical role. By taking care of the infrastructure and reducing costs, we were able to concentrate on innovation and shortened our time to market.

Eigenphi now enjoys a reputation as an extremely high-quality blockchain transaction analysis tool for the Web3 community, helping DeFi stakeholders such as order flow originators and order flow auction providers  protect and grow their revenues. This is no small achievement, but Google Cloud helped make it possible. From its technology to its people, it offers an amazing ecosystem of resources for any startup to reduce costs and gain a competitive edge, whatever their area of business expertise. We’re excited to continue the partnership to innovate and develop new ways to help stakeholders expand their revenue goals.

If you want to learn more about how Google Cloud can help your startup, visit our pagehere to get more information about our program, andsign up for our communications to get a look at our community activities, digital events, special offers, and more.