Google has long looked to Israel for globally impactful technologies including popular Search features, Waze, Live Caption, Duplex and flood forecasting. At our Decode with Google 15RAEL event last week, we celebrated 15 years of Google innovation in Israel and our longstanding support of the country’s vibrant startup ecosystem. 

Over the years, we’ve expanded our enterprise investments in the country, too. In addition to our over a decade long investment in the space, Google has acquired Israeli-based companies like Alooma, Elastifile and Velostrata, and Uri Frank joined Google Cloud last month to lead our server chip design team from our offices in Tel Aviv and Haifa. 

As we continue to meet growing demand for cloud services in Israel, we’re excited to announce that a new Google Cloud region is coming to Israel to make it easier for customers to serve their own users faster, more reliably and securely.

Our global network of Google Cloud regions are the foundation of the cloud infrastructure we’re building to support our customers. With cloud’s 25 regions and 76 zones around the world, we deliver high-performance, low-latency services and products for Google Cloud’s enterprise and public sector customers. With each new Google Cloud region, customers get access to secure infrastructure, smarter analytics tools, an open platform and the cleanest cloud in the industry. 

Having a region in Israel will help accelerate innovation for customers of all sizes, including PayBox, a digital wallet application owned by Discount Bank, one of Israel’s largest banks. “When we acquired PayBox, our goal was to improve the security and the user experience for its products, but we also wanted to keep the startup’s agility and innovation. Google Cloud has enabled us to do just that,” said Sarit Beck-Barkai, Managing Director of PayBox at Discount Bank.

“We are very excited that leading vendors like Google are investing and launching a local cloud region in Israel. This will make a significant change in the technology landscape of the public-sector, enterprise and SMB markets in Israel. Matrix is proud to be a major part of the transition to the cloud,” said Moti Gutman, CEO at Matrix, technology services company and Google Cloud partner. 

“In the last year, Panorays more than tripled its customer base and scaled its infrastructure, practically at the click of a button. Google Cloud made it easy for us to scale without worrying about DevOps, which meant that our engineers could focus on developing new and better features for our customers. The new region launching in Israel will allow us to serve our local customer base even better, as we’ll be able to experience higher availability and deploy resources in specific regions, thus reducing latency.” said Demi Ben-Ari, Co-founder and CTO, Panorays, a third-party security platform and Google Cloud customer.

“This new cloud region will provide even better access and growth potential for our mutual customers with tech hubs in the region. We are serving hyper growth companies who need Google Cloud’s services and will benefit greatly from this regional presence,” said Yoav Toussia-Cohen, CEO of DoiT International.

When it launches, the Israel region will deliver a comprehensive portfolio of Google Cloud products to private and public sector organizations locally. We look forward to welcoming you to the Israel region, and we’re excited to support your growing business on our platform. 

Learn more about our global cloud infrastructure, including new and upcoming regions, here.

The digital media and entertainment industry is experiencing dramatic growth, as audiences migrate to online experiences and content providers seek to deliver new and innovative content. According to The Global Internet Phenomena Report, streaming video accounted for 53.7% of internet bandwidth traffic, up by 4.8% from a year ago. This rapid growth of over-the-top content is straining existing infrastructure, fueling media companies’ shift to the public clouds with their global presence and greater distribution capacities. In addition, other use cases such as gaming, social networks, AR/VR experiences, and education continue to fuel the need for intelligent media services and operations.

Today, at the 2022 NAB Show Streaming Summit, we’re excited to announce the general availability of Media CDN — a modern, extensible platform for delivering immersive experiences with unparalleled scale and intelligence. Media CDN will enable media and entertainment customers to efficiently and intelligently deliver streaming experiences to viewers anywhere in the world. The same infrastructure that Google has built over the last decade to serve YouTube content to over 2 billion users is now being leveraged to deliver media at scale to Google Cloud customers with Media CDN.

Unparalleled planet-scale reach and scale

Media CDN’s foundational advantage is the Google network. We have invested decades of resources to build tremendous capacity and reach in over 200 countries and more than 1,300 cities around the globe. Modern video applications are sensitive to fluctuations in latency, so getting content closer to users enables higher bitrates and reduces rebuffers, resulting in a superior experience for the end user. Media CDN builds on the success of the existing Cloud CDN portfolio for web and API acceleration and complements it by enabling delivery of immersive media experiences.

In addition to running on planet-scale infrastructure, Media CDN tailors delivery protocols to individual users and network conditions. Media CDN includes out-of-the-box support for QUIC (HTTP/3), TLS 1.3, and BBR, optimizing for last-mile delivery . When the Chrome team rolled out widespread support for QUIC, video rebuffer time decreased by more than 9% and mobile throughput increased by over 7%.

Media CDN also achieves industry-leading offload rates. With multiple tiers of caching, we minimize calls to origin — even for infrequently accessed content. This alleviates performance or capacity stress in the content origin and saves costs. These features are built into the product and seamlessly support customer content hosted on Google Cloud, on-premises, or on a third-party cloud.

“We are excited to leverage Media CDN to continue to deliver an exceptional streaming experience for Stan users across Australia. With Google’s massive network, and a deep reach into the ISPs, we are able to deliver the highest quality video for our users, no matter where they are”—John Hogan, Chief Technology Officer, Stan

“Our mission at U-NEXT is to deliver the highest quality and most entertaining content to our users. Google Cloud’s Media CDN helps us efficiently scale our infrastructure, which is challenging with a vast library of content. Media CDN offloaded 98.3% of requests from our origin server while delivering consistent great quality.”—Rutong Li, Chief Technology Officer, U-NEXT

Broader platform for monetization and immersive experiences

While global distribution is critical for a high-quality end-user experience, it’s only one piece of delivering a world-class platform for immersive experiences. Media CDN offers additional capabilities to enable this transformation — ad insertion, ecosystem integrations and platform extensibility, and powerful AI/ML analytics for interactive experiences.

Streaming providers can improve monetization through integrated ad serving via the Video Stitcher API, which allows manipulation of video content to dynamically insert ads.

Through extensible ecosystem integrations, Media CDN connects customers to key capabilities to simplify their operations. For example, the Transcoder API supports custom streaming formats, while the Live Stream API transcodes mezzanine live signals into direct-to-consumer streaming formats, for multiple device platforms.

Media CDN is built with AI/ML that will give viewers more control over how they see, experience, and even interact with content. For example, sports fans watching a game can obtain real-time stats and analytics, viewers can purchase items from virtual billboards, etc.

Cloud-native and developer-friendly operations

Media companies are under pressure to develop and deploy innovative experiences at a furious pace. Media CDN was built by developers, for developers, with automation and observability built in, giving media providers the speed and flexibility they need to integrate delivery provisioning and management into their content release processes.

Media CDN offers comprehensive APIs and automation tools such as Terraform. Detailed, pre-aggregated metrics and playback tracing make it easy to diagnose performance across the entire infrastructure stack. Real-time visibility is provided via Google Cloud’s operations suite, and integrates with tools that developers already use such as Grafana and ElasticSearch.

“Leveraging the same infrastructure as YouTube, Google Cloud’s Media CDN combines geographic reach, API-first architecture and integration with the Cloud operations suite. This is a transformative move that is aligned with the future of the CDN industry.”— Ghassan Abdo, Research Vice President, WW Telecom, Virtualization and CDN, IDC

“Viewers around the world are demanding best-in-class video quality and performance across modes of consumption. A video-first delivery network can be a game changer in this space. We’re excited to partner with Google Cloud and to leverage Media CDN to enable premium video experiences and customer engagements.”—Juan Martin, Founder and CTO, Firstlight Media

Planet-scale advanced security

Media CDN lets streaming media providers take advantage of Google’s decades-long experience delivering video safely, securely, and reliably. The platform includes deep integration with Google Cloud Armor for planet-scale DDoS protection and a rich set of capabilities to detect and mitigate attacks, prevent abuse, manage risk, and comply with regulatory or licensing requirements.

If you want to deliver rich, immersive experiences to global audiences with an extensible, modern delivery platform, we’d love to hear from you. For more information, including technical specifications and platform architecture, please visit cloud.google.com/media-cdn. To get started with Media CDN, contact your sales team.

For far too long, managing employee expenses has been a time-consuming process that requires manual data entry and reconciliation to bridge the gap between business bank accounts and ERP systems. In the absence of an integrated workflow, finance teams use multiple systems to manage credit card and cash payments, and finding receipts. In most cases, they also lack real-time visibility into company spending.

The complexity grows exponentially as businesses expand, especially into new regions. Extra administration required to manage new bank accounts, card issuers, and local accounting systems impedes decision making and negatively impacts revenues and growth. Businesses of all sizes struggle with this, but it can be especially challenging for medium to large enterprises.

Payhawk set out to help businesses overcome these challenges when we founded the company in 2018. We combine VISA company cards, reimbursable expenses, and accounts payable into a single product. Our customers can automate manual processes, maximize efficiency, and accelerate business expansion.

Setting up our first cloud cluster in less than a week

To support growth and attract investment we were keen to launch our solution on a scalable, future-proof IT architecture that didn’t require extensive technical support. This is where Google Cloud made a big impression, especially the user interface and documentation which massively reduces the resources required to set up clusters and put them into production.

I’m a CTO, not a DevOps specialist, but in less than a week I was able to set up a secure, reliable operating infrastructure. This enabled us to fast-track our application development and we were able to issue our first card in just eight months. Our Google Cloud partner, Cloud Office also gave us valuable assistance, guiding us through the deployment process and advising on Google Cloud’s extensive range of solutions.

Google Kubernetes Engine (GKE) played a critical role, accelerating the deployment and management of our cloud native applications. We use Cloud SQL as our database while other important tools include Cloud Memorystore, Vision AI, Cloud Storage and Artifact Registry for our wider data storage and application needs. With Firebase we’ve been able to build a notification system for mobile devices.

Another incentive is that most other cloud solutions require add-on services to build and keep your product live. With Google Cloud, all the services that Payhawk needs including logging, metrics, monitoring of resources, and utilization of CPU memory come as standard.

For instance, I was really impressed by Google Cloud’s operations suite, which includes Cloud Logging and Cloud Monitoring. If there are any anomalies in our cloud architecture, we can track and resolve them with minimal disruption to our operations. This also removes the need to invest in an additional observability solution.

Reliability that builds customer trust

Google Cloud also supports Payhawk’s mission to put customers at the center of our organization. Thanks to Google Cloud error reporting and tracking and Google Cloud single sign on, Payhawk’s engineering team can anticipate customer issues and correct them in less than one hour. Trust is everything, and Google Cloud gives us the tools to boost customer satisfaction and build long-term relationships.

As a young business, managing costs is also a priority. The Google for Startups Cloud Program, which includes credits for Google software and tools, enabled us to push the business forward without having to worry about financing our infrastructure, especially in the first year. This gave us breathing room to work through funding, application development, and the onboarding of our first customers.

In addition, Google Cloud gives us confidence that we can grow the business fast. In most months we have seen more than 10% growth — in some cases it’s been 20%. In the first half of 2022, the business doubled in size, but Google Cloud gave us the flexibility to scale our infrastructure, adding storage, memory, and processing power as we onboarded new customers. The pricing model is also generous so that we can grow our revenues while keeping control of operational expenditure.

Since launch we have acquired a valuable mix of customers from startups to large businesses that want to reduce the costs of their expenses programs and increase employee satisfaction. They include ATU, a German automobile servicing company, which has successfully digitized its entire procurement process, and Discordia, a Bulgarian logistics business with 10,000 trucks, which has issued Payhawk cards to all its drivers.

Looking to the future, it’s no exaggeration to say that Google Cloud is a foundation of our business and has given investors confidence in our operations. From a first seeding round of €3 million, early this year we closed a Series B extension of $100 million. This gives us a valuation of $1bn and makes Payhawk the first ever Bulgarian unicorn.

We now operate in 32 countries in Europe and the US, and plan to double our team by the end of the year. It feels like we’ve come a long way since we first started using Google Cloud, and I’m thrilled that we have Google Cloud as a global technology partner supporting our mission to transform expense management and financial operations worldwide.

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

Weather forecasting and climate modeling are two of the world’s most computationally complex and demanding tasks. Further, they’re extremely time-sensitive and in high demand — everyone from weekend travelers to large-scale industrial farming operators wants up-to-date weather predictions. To provide timely and meaningful predictions, weather forecasters usually rely on high performance computing (HPC) clusters hosted in an on-premises data center. These on-prem HPC systems require significant capital investment and have high long-term operational costs. They consume a lot of electricity, have largely fixed configurations, and the underlying computer hardware is replaced infrequently.

Using the cloud instead offers increased flexibility, constantly refreshed hardware, high reliability, geo-distributed compute and networking, and a “pay for what you use” pricing model. Ultimately, cloud computing allows forecasters and climate modelers to provide timely and accurate results on a flexible platform using the latest hardware and software systems, in a cost effective manner. This is a big shift compared with traditional approaches to weather forecasting, and can appear challenging. To help, weather forecasters can now run the Weather Research and Forecasting (WRF) modeling system easily on Google Cloud using the new WRF VM image from Fluid Numerics, and achieve the performance of an on-premises supercomputer for a fraction of the price. With this solution, weather forecasters can get a WRF simulation up and running on Google Cloud in less than an hour!

A closer look at WRF
Weather Research and Forecasting (WRF) is a popular open-source numerical weather prediction modeling system used by both researchers and operational organizations. While WRF is primarily used for weather and climate simulation, teams have extended it to support interactions with chemistry, forest fire modeling, and other use cases. WRF development began in the late 1990s through a collaboration between the National Center for Atmospheric Research (NCAR), National Oceanic and Atmospheric Administration (NOAA), U.S. Air Force, Naval Research Laboratory, University of Oklahoma, and the Federal Aviation Administration. The WRF community comprises more than 48,000 users spanning over 160 countries, with the shared goal of supporting atmospheric research and operational forecasting.

The Google Cloud WRF image is built using Google’s MPI best practices for HPC, with the exception that hyperthreading is not disabled by default, and is easily integrated with other HPC solutions on Google Cloud, including SchedMD’s Slurm-GCP. Normally, installing WRF and its dependencies is a time consuming process. With these new WRF VM images, deploying a scalable HPC cluster with WRF v4.2 pre-installed is quick and easy with our Codelab. OpenMPI 4.0.2 was used throughout this work. Google has had good success with Intel MPI, and we intend to study whether further performance gains can be achieved in this context.

Optimizing WRF

Determining the optimal architecture and build settings for performance and cost was a key part of the process in developing the WRF images. We evaluated how to select the ideal compiler, right CPU platform, and the best file system for handling file IO, so you don’t have to. As a test case for assessing performance, we used the CONUS 2.5km benchmark.

Below, the CONUS 2.5km runtime and cost figure shows the run time required for simulating WRF over a two-hour forecast using 480 MPI ranks (a way of numbering processes) for different machine types available on Google Cloud. For each machine type, we’re showing the lowest measured run time from a suite of tests that varied compiler, compiler optimizations, and task affinity.

We found that compute-optimized c2 instances provided the shortest run time. The Slurm job scheduler allows you to map the MPI tasks to compute hardware using task affinity flags. When optimizing the runtime and cost for each machine type, we compared using srun –map-by core –bind-to core to launch WRF, which maps each MPI process to a physical core (two vCPU per MPI rank), and srun –map-by thread –bind-to thread, which maps each MPI process to a single vCPU. Mapping by core and binding MPI ranks to cores is akin to disabling hyperthreading.

The ideal simulation cost and runtime for CONUS 2.5km for each platform is found when each MPI rank is subscribed to each vCPU. When binding to vCPUs, half as many compute resources are needed when compared to binding to physical cores lowering the per-second cost for the simulation. For CONUS 2.5km, we also found that although mapping MPI ranks to cores results in reduced runtime for the same number of MPI ranks, the performance gains are not significant enough to outweigh the cost savings. For this reason, the WRF-GCP solution does not disable hyperthreading by default.

Runtime and simulation cost can be further reduced by selecting an ideal compiler: the figure below (CONUS 2.5km Compiler Comparisons) shows the simulation runtime for the WRF CONUS 2.5km benchmark on eight c2-standard-60 instances, using GCC 10.30, GCC 11.2.0 and the Intel® OneAPI® compilers (v2021.2.0). In all cases, WRF is built using level 3 compiler optimizations and Cascade Lake target architecture flags. By compiling WRF with the Intel® OneAPI® compilers, the WRF simulation runs about 47% faster than the GCC builds, and at about 68% of the cost, on the same hardware. We’ve used OpenMPI 4.0.2 with each of the compilers as the MPI implementation in this work. With other applications, Google has seen good performance with Intel MPI 2018, and we intend to investigate performance comparisons with this and other MPI implementations.

File IO in WRF can become a significant bottleneck as the number of MPI ranks increases. Obtaining the optimal file IO performance requires using parallel file IO in WRF and leveraging a parallel file system such as Lustre.

Below, we show the speedup in file IO activities relative to serial IO on an NFS file system. For this example, we are running the CONUS 2.5km benchmark on c2-standard-60 instances with 960 MPI ranks. By changing WRF’s file IO strategy to parallel IO, we accelerate file IO time by a factor of 60.

We further speed up IO and reduce simulation costs by using a Lustre parallel file system deployed from open-source Lustre Terraform infrastructure-as-code from Fluid Numerics. Lustre is also available with support from DDN’s EXAScaler solution in the Google Cloud Marketplace. In this case, we use four n2-standard-16 instances for the Lustre Object Storage Server (OSS) instances, each with 3TB of Local SSD. The Lustre Metadata Server (MDS) is an n2-standard-16 instance with a 1TB PD-SSD disk. After mounting the Lustre file system to the cluster, we set the Lustre stripe count to 4 so that file IO can be distributed across the four OSS instances. By switching to the Lustre file system for IO, we speed up file IO by an additional factor of 193, which is orders of magnitude faster than a single NFS server with serial IO.

Adding compute resources and increasing the number of MPI ranks reduces the simulation run time. Ideally, with perfect linear scaling, doubling the number of MPI ranks would cut the simulation time in half. However, adding MPI ranks also increases communication overhead, which can increase the cost per simulation. The communication overhead is due to the increased amount of communication necessitated by splitting the problem more finely across more machines.

To assess the scalability of WRF for the CONUS 2.5km benchmark, we can execute a series of model forecasts where we successively double the number of MPI ranks. Below, we show two- hour forecasts on the c2-standard-60 instances with the Lustre file system, varying the number of MPI ranks from 480 to 1920. In all of these runs, MPI ranks are bound to vCPUs so that the number of vCPUs dedicated to each simulation increases with the increase in MPI ranks. While many HPC workloads run best with simultaneous multithreading (SMT) disabled, we find the best performance for CONUS 2.5km with SMT enabled. Thus, the number of MPI ranks in our runs equals the total number of vCPUs.

As you can see, the CONUS 2.5km Runtime & Cost Scaling figure shows that the run time (blue bars) decreases as the number of MPI ranks and the amount of compute resources increases, at least up to 1920 ranks. When transitioning from 480 to 960 MPI ranks, the run time drops, yielding a speedup of about 1.8x. Doubling again to 1920 MPI ranks, though, we obtained an additional speedup of just 1.5x. This declining trend in the speedup with increasing MPI ranks is a signature of MPI overhead, which increases with more MPI ranks.

Determining your best fit

Most tightly-coupled MPI applications such as WRF exhibit this kind of scaling behavior, where scaling efficiency decreases with increasing MPI ranks. This makes assessing cost-scaling alongside performance-scaling critical when considering Total Cost of Ownership (TCO). Thankfully, per-second billing on Google Cloud makes this kind of analysis a little bit easier. As shown above, a second doubling of the count from 960 cores to 1920 cores can provide an additional 1.5x speedup, but at a 32% higher cost. In some circumstances, this faster turnaround may be needed and worth the extra cost.

If you want to get started with WRF quickly and experiment with the CONUS 2.5km benchmark, we’ve encapsulated this deployment in Terraform scripts and prepared an accompanying codelab.

You can learn more about Google Cloud’s high performance computing offerings at https://cloud.google.com/hpc, and you can find out more about Google’s partner Fluid Numerics at https://www.fluidnumerics.com.