Challenge for Organizations Developing AI and ML Models
Organizations developing artificial intelligence (AI) or machine learning (ML) models typically work with massive data sets. This means deploying a significant number of servers equipped with powerful CPU processors, GPU graphics cards, or dedicated AI accelerators – resources that are critical to achieving accelerated computing. Vast amounts of disk space for data storage and powerful telecommunications solutions are also required.
Because of the demands and scale of such IT infrastructure, the challenge is to find a data center that has the appropriate colocation space and can provide servers with secure working conditions.
Requirements for Data Center
- The data center must be able to accommodate and maintain a customer’s high-performance IT infrastructure, and be prepared to adapt to changing demand for space or other data center resources.
- Maintenance must include high-efficiency cooling systems and guaranteed availability of server power, as well as 24/7 on-site technical support.
- Adequate telecommunications solutions must be in place: both internal networks with ultra-low latency and reliable broadband connections to the Internet and global cloud platforms.
- It is also important to consider security: physical, power, and telecommunications, as well as the availability of solutions that support data security, disaster recovery, or geo-redundancy.
The Solution from Atman
For the AI solutions sector, Atman offers a colocation space prepared to accommodate a large number of computing resources requiring high power density, hyper-efficient cooling and effective connectivity.
Power and Environmental Parameters
The technologies implemented in the Atman Data Centers allow for powering and cooling servers equipped with powerful CPUs or GPUs. We use a variety of efficient precision air conditioning systems and are ready to provide customers with any solution they require, including liquid cooling, rear door cooling.
Our facilities are capable of supporting high power density infrastructure, up to 50 kW per rack. The 2N redundancy of the primary power supply system combined with an extensive emergency power supply system in N+1 redundancy guarantee full protection of the customer’s IT infrastructure against power outages (blackouts) and voltage fluctuations (brownouts).
In addition, Atman has a significant reserve of dedicated power and HVAC precision air conditioning redundancy. As a result, our data centers are prepared for significant fluctuations in the load on the power and cooling systems during periods of increased computing resource utilization, depending on the training or operational phase of the AI model.
360-degree Communication
As a telecommunications operator and a data center service provider with decades of experience, Atman is able to provide ultra-low latency on internal networks to enable AI to make decisions and deliver results in real time.
In addition, we guarantee the high bandwidth of Atman’s network links to and from the data center, which is essential when training an AI model on a distributed infrastructure. At the same time, we provide access to the services of many telecom operators, which are offered in the Atman Data Centers on an equal basis (carrier neutrality).
Importantly, Atman sets up private data links (Atman Cloud Connect) to any region of the largest global clouds, ensuring fast and secure connections to AWS, Google Cloud Platform, Microsoft Azure and others. This makes it easy for the customer to build a multi-cloud solution.
Agility and Complementarity
Atman is prepared to nimbly implement data center infrastructure changes to keep pace with rapidly evolving AI technology. Our experienced technical teams prioritize maintaining SLA parameters and are ready to perform remote hands in no more than a dozen minutes.
It is worth noting the benefits that complementary services provided by Atman can bring. By storing data in Atman Object Storage, customers can generate savings. Automated backups through Atman Backup will protect the customer’s most important data and applications from loss (e.g. due to disk failure).
With data center facilities in two and soon three locations in Warsaw, Atman enables and supports the creation of distributed environments to implement geo-redundancy and disaster recovery solutions (DRC) to ensure business continuity of customers’ IT systems (BCM).
Benefits
Energy Security
- High availability (SLA 99.999%) of the power supply; all energy comes from RES
- No single point of failure (2N redundancy)
- Significant reserve of allocated power for both IT equipment and cooling system
- Full protection of customer equipment from blackouts and brownouts (UPS systems, gensets)
Telecommunications Security
- Ultra-low latency network connections in the data center
- Free choice of Internet access or data transmission provider
- High data throughput to and from the data center, link redundancy
- Availability of dedicated connections to global clouds (Cloud Connect Service)
- Presence of an Internet exchange access node (peering, IP transit)
Operational Security
- High quality of data center systems, including high efficiency of HVAC system
- Flexibility and agility of technical services
- 24/7 support from experienced technical services (remote hands)
- Ability to create geo-redundant disaster recovery solutions (DRC)
- Optional services: anti-DDoS protection, Backup as a Service, Object Storage