A company culture of growth

Learn how your company's culture can either advance or totally sabotage your company goals.

Domino Data Lab announced new integrations with Nvidia this week to make it easier to adopt AI infrastructure, scale GPU clusters, run more virtual workloads on high-end GPUs, and package AI apps into container infrastructure.

Domino’s tools streamline the grunt work associated with building out AI and ML applications. Domino automatically spins up workspaces or models on shared infrastructure so many people can share the same infrastructure. When someone is finished with a workload, Domino spins down that workspace to free up the resources for someone else. Domino also tracks usage, letting IT administrators see consumption and make informed decisions about when to increase computing power.

Gartner considers AI orchestration tooling that includes MLOps to be a key trend in 2021.

Easier GPU clustering

Domino currently supports ephemeral clusters built on Apache Spark and Ray, and the company plans to add support for Dask this fall. Domino strategic partnerships VP Thomas Robinson told VentureBeat that Spark has traditionally excelled at large-scale data processing and transformations. Ray has simplified distributed training and hyperparameter optimizations, and Dask has excellent integration with commonly used Pandas and NumPy libraries.

Domino also improved the ability to provision GPU clusters required to run AI training jobs that require more than one Nvidia GPU. Traditionally, it could be difficult and time-consuming to set up machines, ensure network connectivity, and install proper libraries. In addition, it is uncommon for enterprises to give data scientists access and permission to manipulate infrastructure directly. As a result, teams often leave clusters idle between larger projects, rather than reallocate the individual machines for smaller projects.

To improve utilization rates, Domino makes it possible to spin up and spin down interactive sessions, batch jobs, or models hosted on Nvidia DGX infrastructure to allow multiple concurrent and consecutive sessions. Previously users depended on email and spreadsheets to coordinate workloads, which was inefficient.

Domino will add support for Nvidia’s multi-instance GPU technology in September. MIG allows a single GPU to be sliced up into smaller portions (7 slices per GPU for each of the 8 GPUs in a DGX A100 — a total of 56 slices). This will make it possible to divide the capacity of a larger GPU server or cluster into multiple instances or partitions to host many more predictive models on smaller GPU instances. While many deep learning training workloads require a whole machine or multiple machines in a cluster, research, or inference (prediction), workloads are much less GPU-intensive.

“By allowing the GPU to be portioned into pieces, you can have more researchers doing discovery work in notebooks on smaller GPU slices,” Robinson said.

Added container support

Domino also announced immediate support for Nvidia’s new NGC container registry service. This makes it easier to package vetted application and configuration settings into container instances that bake in best practices. This means a data scientist doesn’t have to spend time figuring out how to set up and install all the drivers and tools they need. It also allows organizations to standardize on these containers.

NGC currently supports RAPIDS, TensorFlow, PyTorch, and CUDA. Domino additionally supports containers for SAS, MATLAB, Amazon SageMaker, and private container repositories.

Finally, Domino worked with Nvidia and NetApp to develop a preconfigured hardware/software package called the ONTAP AI Integration Solution. “This is a specced, tested, and verified packaging of everything you need to accelerate your data science work — so there’s no guesswork and no setup needed for an IT department,” Robinson said.

A company culture of growth

Learn how your company's culture can either advance or totally sabotage your company goals.

This article is part of a Gaming Insights series paid for by Facebook.

We recently spoke with Katerina Dudinskaya, VP of Performance Marketing at Gismart, to discuss how the award-winning European publisher is adapting its approach to face new challenges brought about by upcoming iOS 14 changes, and how the team is closing the gap between acquisition and monetization with campaign-level Return On Ad Spend (ROAS).

Facebook: Let’s start off with a macro view of the industry: The pandemic is in full swing, and iOS 14 changes are coming up. How are these massive shifts impacting your business?

Dudinskaya: The pandemic has driven more traffic to almost all of our entertainment products and much of the 2020 uplift has continued through 2021, although we are seeing more competition.

And IDFA deprecation will force us to adapt our approach to marketing and business. For us, the best strategy is just to accept this new reality and try to be creative about how to adapt.

Facebook: As limitations on targeting are becoming top-of-mind for publishers and developers, we’re seeing an evolution in the ads ecosystem. Can you talk about some of the new challenges you’re facing this year?

Dudinskaya: At a high-level, our current challenges are firstly, how to deliver personalized ads and reach our most valuable customers and, secondly, how to provide our monetization partners with reliable user data to predict user-level value more accurately. Latency data and fraudulent data can affect our predictions, and using incorrect data can lead us to the wrong decisions.

We’re also developing updates for our user acquisition structure, analytics, and creative production to help us deal with these challenges.

Facebook: Let’s dive into what it means to build resilience in 2021. What approach are you taking when it comes to app monetization? 

Dudinskaya: Like most other publishers, we will be using a combination of SKAdNetwork and our own analytics systems to track all possible marketing funnel events (downloads, click-throughs, etc). This data will become the base for the decisions and predictions we make.

We’re also working more closely with all our partners: publishers, traffic vendors, and monetization networks. In this new reality, more than ever, you need to work as a team.

Facebook: User acquisition has been a key challenge for publishers and developers. How will you ensure that Gismart will continue to reach and acquire high-value users?

Dudinskaya: We are using all open solutions that are currently available, including AppsFlyer’s proprietary user attribution solution and Unity’s machine learning solution. We also work closely with Facebook Audience Network and Facebook Instant Games teams who help us with user acquisition and monetization. We also continuously improve non-technical ways of personalization, such as growth activities on landing pages, creative production, and UA channel diversification.

Facebook: Many publishers are concerned with revenue loss and the impact of ad targeting limitations and fluctuating CPMs. How do you plan to sustain revenue in this environment? 

Dudinskaya: We plan to apply a holistic approach, working on UA strategies and targeting creative production. We also use some CPM/bidding add-ons produced internally to be able to perform fast testing and to get a snapshot of user funnel metrics from the first impression to the deepest product metric. For games, this could be clickthrough rate, cost per install, cost per app event/acquisition, or average revenue per user.

Facebook: How are you currently measuring return on ad spend (ROAS) at a campaign level? How do you maintain oversight of what’s working on which channels?

Dudinskaya: Facebook Audience Network’s new campaign-level IAA ROAS has helped us to better understand our ROAS, offering accurate insights to make more profitable decisions.

It enables us to acquire quality users, understand the ideal user experience, and ensure long-term engagement and revenue. The synergy of data we receive from different analytics systems, mediation, and other tools give us deeper insights into the channel performance. The more data we can bring into day-to-day operations, the better results we see in our marketing channels.

Facebook: What advice would you share with other publishers and developers who are also preparing to face the headwinds of the upcoming industry changes?

Dudinskaya: It’s worth noting that each advertising platform is quite unique; some creative approaches that work well on one platform, might not perform that well on another. So publishers need to have individual marketing strategies for each platform, taking into consideration their audience and content consumption habits, and the platforms’ technical specifics.

Also, we’re always open to new solutions and we are constantly testing new Facebook features and tools available in an alpha/beta version. Testing new solutions is not only a great way to adapt your marketing strategy but also helps to identify errors and gaps in your current solution set.

Anastasia Petrova is Strategic Partner Manager at Facebook Audience Network.

VB Lab Insights content is created in collaboration with a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected].