Media & Entertainment
Oleg Yudin, Head of NOC at Gcore, talks about why a geographically distributed network is important for the success of any online business, about competing with American and Asian service providers, and development plans for 2020–2021.
— What does the Gcore network look like today, and what is its place in the company’s business?
— As of the beginning of 2020, Gcore’s own network infrastructure covers all continents (except for Antarctica) and is present in every key market. The network consists of more than 100 points of presence in 65 cities across the world.
“The Gcore network is the company’s asset base, and its biggest competitive advantage, providing uninterrupted operation and immediate availability of all our services.”
Head of NOC
It is not an exaggeration to say that, in the case of Gcore, the network is the company’s asset base and its biggest competitive advantage, providing uninterrupted operation and immediate availability (within 30 ms) of all our services, including the Content Delivery Network, Cloud, Hosting, Media Platform, Storage, and DDoS Protection. In terms of the quality and functionality of our network, we’re among the top 7 world players already, and we compete successfully with such market leaders as Amazon, Microsoft, Akamai, and Cloudflare.
— What are its main advantages?
— First of all, it’s the quality. I think the most important factor in the success of our network is its initial ideology. Right after the founding of our company, we were building our own global content delivery infrastructure in order to solve a very complex issue for our first customers: to deliver the most demanding type of content—high quality online games—over vast distances, literally across borders and oceans.
What is important to understand here is that the gameplay is impossible if users in Chicago, London, and Shanghai don’t have a real-time connection to each other. We’re not talking about a one second delay, but about 20–30 ms!
The architecture we created has even set a few world records (including being recognized by Guinness World Records) for providing a seamless online battle experience for a million users playing simultaneously. As far as I know, no other company in our field can boast the same achievement.
Today, our content delivery network is ready for any peak load. According to Citrix.com (the leader in the field of independent CDN monitoring and performance optimization), our network is one of the best in Europe, in terms of performance by the following parameters: best throughput, best response time, and best file load time.
— Can you give us any real-life examples?
— What this means for users is, for example, you’re able to watch more than 1,500,000 videos in Full HD simultaneously, download more than 5,000,000 songs per second, or listen to more than 20,000,000 tracks.
— I see. And how well does your network perform in North America, specifically in the USA and Canada?
— In the last three years, we have been able to make substantial progress on the American market: while in 2016 the average response time of our network in the US was 65 ms, today we’re talking about 26 ms and even less. It confidently puts us among the top 5 CDN service providers on the local market. We were able to achieve this because of two major factors: an actively expanding network of peering partners (we have more than 10,000 of them today), and of course, the 10 points of presence of our infrastructure throughout the country, from the West to the East coast. Currently, the data centers in the US fully cover our customers’ needs, Canada included. We do intend, however, to install several additional points in Canada as well, to address the growing market needs.
— How do you plan to develop your network in the next two years?
— We’re working on about 100 projects for new points of presence. In the next two years, we should open at least 20 additional locations of our infrastructure. We’ll definitely strengthen the European region further. More specifically, we have plans for Austria, for the Balkans, including Serbia, and for Turkey. Our presence in South America will be significantly expanded as well: besides the existing points in Brazil, we’ll deploy our infrastructure in Colombia, Peru, Chile, and Argentina. In addition, at least two new locations will appear in the USA, and points of presence will be launched in Mexico, India, Thailand, and Taiwan.
“We’re working on about 100 projects for new points of presence. In the next two years, we should open at least 20 additional locations of our infrastructure.”
Head of NOC
In this regard, it’s important to mention that our flexibility and readiness to make customizations are also advantages of Gcore that distinguish us favorably from the global top 5 American and Asian cloud companies. There are a number of cases when we opened CDN and hosting points of presence specifically to suit the needs of particular clients that needed to strengthen themselves in a certain region, as long as such work was at least minimally economically justifiable for us. If you contact an IT giant in the USA or China with such an offer, with a budget of less than 5–10 million dollars, they most likely won’t even hear you out.
— Tell us more about the architecture of your network, the data centers you choose, the equipment and technologies you use.
— Our network architecture is based on four main principles. First, using the edge technology of developing infrastructure with full automation: the new PoPs we open are located as close as possible to the end product users and consumers of our customers’ content.
Second, stable performance, which means that during peak loads, we quickly redistribute traffic between existing nodes to achieve an optimal ratio. Traffic moves to the nearest locations based on geoinformation (city, region, country) about customer queries to the overloaded server.
Third, reliable connectivity. To achieve good connectivity and quality peering, we choose the best Tier III and Tier IV data centers, and the biggest internet exchange points. There we are connected with the networks of the biggest internet providers and our peering partners.
Finally, we keep traffic local. We manage four independent network segments: North America, Asia, and Europe. To achieve optimal traffic routing, we announce regional prefixes only in each point of presence.
“Our network architecture is based on four main principles. First, using the edge technology of developing infrastructure with full automation: the new points of presence we open are located as close as possible to the end product users and consumers of our customers’ content.”
Head of NOC
We use network equipment with 100GE interfaces as edge routers. For our cache servers, we use 100 GB network adapters. We only install Samsung Enterprise series SSDs to ensure the best performance: standard SAS (PM1633) for medium-load regions and NVME (PM1725) in the most loaded points. The majority of the SSDs have a capacity of 3.2 TB. We are also actively implementing new 7.68 TB drives.
To monitor and balance the ever-growing network load more effectively, we have implemented the SD Edge technologies that are being actively developed today. They allow us to automate a whole range of processes that previously required manual maintenance.
It should also be noted that our network, and all the Gcore services based upon it, are protected against cyberattacks, including DDoS attacks, with the help of our own solution which is based on unique smart traffic filtering technology.
It combines the analysis of statistical, signature, technical, and behavioral factors, and allows us to neutralize low-frequency attacks at the application level, and to cut off even singular bot queries.
— What challenges did you have to overcome when expanding your network, and in which countries?
— Probably the main challenge for us when expanding the network has been the slow feedback from our telecom partners and internet providers due to their internal bureaucracy.
For example, in a number of countries in Latin America and the Middle East, we spent 3–4 months to get approval for standard procedures (like connecting to a network), while in the USA and Europe, it only takes a week or two. It’s also important, of course, to take into account the local particularities of importing and deploying infrastructure equipment.
Over the years of working in various corners of the world, however, we have learned to deal expeditiously with the majority of emerging issues.
This article discusses how GPUs are shaping a new reality in the hottest subset of AI training: deep learning. We’ll…
In this article, we explore how Intelligence Processing Units (IPUs) and graphics processing units (GPUs) drive the rapid evolution of…
Navigating the maze of hosting solutions can be a daunting task. Assessing the differences and deciding which best fits your…