TL; DR: HOSTKEY, headquartered in Amsterdam, has provided competitively priced equipment leasing, colocation services, and cloud solutions to customers across Europe and North America for more than a decade. Now, the company’s high-performance GPU servers are helping data scientists increase the efficiency of compute-intensive artificial intelligence and machine learning workloads. The company is also offering a new support package designed to lend customers a helping hand through outsourced on-site datacenter support.
Data science jobs are in high demand — and they continuously rank among the best sources of employment in America.
According to Glassdoor, one of the world’s most extensive job and recruitment sites, those pursuing careers in data science enjoy not only ample employment opportunities but also triple-digit median base salaries and favorable job satisfaction ratings.
But life as a data scientist isn’t all roses. In many ways, the emerging industry is outpacing the IT infrastructure needed to support its compute-intensive workloads, leading to high costs and poor efficiency.
Dutch hosting provider HOSTKEY is looking to change this with its GPU servers, which use graphics processing units in tandem with CPUs to accelerate scientific computing at affordable price points. These servers come pre-installed with machine learning and neural network libraries, including PyTorch, Keras, Theano, and TensorFlow.
“It can be difficult for this audience to cover all of their costs using similar GPU platforms from Microsoft Azure, Amazon Web Services (AWS), and Google Cloud, which are quite expensive,” said Dmitry Asafyev, Senior Marketing Communications Manager at HOSTKEY. “They’re always happy to get affordable service and support through us.”
HOSTKEY is also currently receiving applications for its GPU Winners Grant Program, which provides free GPU servers in support of selected scientific projects, promising startups, and winners of big data-related competitions.
In addition, the company recently announced the launch of Smart Hands, a division of the company staffed by industry professionals who offer on-site colocation support in any datacenter in the Netherlands. The service provides options for businesses to offload IT management and maintenance tasks, especially as telework increases as a result of COVID-19.
GPU-Accelerated Computing via Parallel Architecture
HOSTKEY has provided everything from affordable cloud and dedicated servers to GPU and colocation solutions to customers across Europe and North America for more than 10 years.
Dmitry told us that the company has recently seen increased demand for its dedicated and virtual GPU servers, which can be purchased in a ready-to-use format or customized using a powerful HOSTKEY wizard.
“We are very flexible,” he said. “We can build customized servers in approximately two business days and VPS GPU in 4 hours. We can also install AI frameworks and other software on your server or allow you to deploy your own container.”
HOSTKEY’s dedicated and virtual CPU servers feature both NVIDIA GeForce GTX 1080, 1080Ti, and RTX 2080Ti graphics cards designed to accelerate high-compute tasks such as data analysis, training neural networks, rendering, and 3D animation.
In June, HOSTKEY published a comparison of GPU cloud platforms and dedicated servers using NVIDIA graphics cards based on data from Catalyst, a high-level library for in-depth learning research. The authors conclude that it is more affordable to train a machine learning model on a HOSTKEY server with a GeForce GTX 1080 Ti card than it is to use an AWS server with an NVIDIA Tesla K80 Accelerator. Differences in speed, though minimal, were observed favoring the Tesla hardware.
“The general trend is this: cheaper consumer-level graphics cards provide overall better value for the money than Tesla’s expensive GPUs,” the authors stated. “The deficiency in the sheer speed of the GTX 1080 Ti, when compared to Tesla, can be compensated for by an increase in the number of cards, thus eliminating any advantage in using the Tesla.”
According to the article, using expensive resources such as the Tesla V100 GPU for training is, simply put, overkill:
“If the task does not imply some kind of ultra-high load, such as face recognition from a large number of cameras, then just one or two graphics cards of the GeForce GTX 1080 Ti level are enough for most jobs — a key element in managing a project’s budget.”
Not sure if you’re ready to commit? Research teams, startups, and established companies can apply for a free one-week trial of a virtual GPU server by providing HOSTKEY with a valid mobile phone number and country code. After the trial period ends, users can either pay for the server to continue using it or do nothing — in which case HOSTKEY will simply terminate the resource.
Building a Close Relationship with the Data Science Community
HOSTKEY is laser-focused on catering to its GPU server customer base, which includes engineers, architects, designers, and transcoding specialists, among other professionals.
“We serve every industry that employs artificial intelligence and machine learning,” said Nadia Belova, Business Partnerships Manager at HOSTKEY.
The company supports this community via the HOSTKEY GPU Grant Program for scientific research projects, startups, and winners of data science competitions. The program, open to full-time faculty or research fellows at established academic institutions, was created to support research and projects centered on innovative uses of GPU processing.
Applicants must provide information on the research area, how the GPU server will be used, challenges addressed, benefits to society, and relevant industrial applications. Winners will receive computing resources to power the work described in their submission.
In return, HOSTKEY requires winners to post links to the company in publications and presentations, as well as mention the company in related speeches, for one year following the award.
In addition to providing access to free computing power via GPU servers, HOSTKEY is focused on building positive relationships with the data science community via attentive around-the-clock support.
“We are featured on many review platforms, and if you go and read the reviews about us, you’ll see that we rarely receive fewer than four or five stars,” Nadia told us. “People always tell us that the one feature they love the most is our responsive, high-quality customer support.”
Even amid the uncertainty of the coronavirus crisis, HOSTKEY has successfully transitioned to a remote workforce without disrupting daily operations. According to a recent press release, the company’s customer-care reaction times remain at 15 minutes. Its sales, accounting, and tech departments have faced zero interruptions, and the HOSTKEY team continues to operate at standard capacity.
New: HOSTKEY’s Remote Smart Hands Service
To further elevate customer service, HOSTKEY recently launched Smart Hands, a 24/7 on-site support solution that customers can employ in any datacenter in the Netherlands.
The on-call service includes assembly and installation, system and network administration, maintenance, project deactivation, server and network support, troubleshooting, inventory auditing, and storage management. HOSTKEY will even purchase and rent necessary hardware locally to minimize expenses for clients with staff who are not close to the datacenter.
In an increasingly isolated world, Dmitry said Smart Hands was designed to help businesses maintain remote infrastructure and services while protecting employees from unnecessary travel. HOSTKEY staff already service more than 4,000 servers in multiple datacenters in the area, tackling everything from routine maintenance to urgent and complex operations
“If required, we can take photos and video, or even live stream the whole work process and its results while reacting in real time to your input,” Dmitry said. “These services are quite appropriate for our current COVID-19 environment.”