No matter how well-designed your game, player experience will suffer if you don't have the right compute. First off, you want options in terms of machine configuration, including CPU, RAM, storage and bandwidth. Plus, since location has a direct impact on latency (the further your server from a player, the higher their ping time will be), you need data centres in strategic locations around the world.
You also need to make sure your setup is resilient should one centre go down. That means tapping into multiple infrastructure layers to avoid putting all your eggs in one basket, so to speak. Then there's the question of scale: what happens if you get a sudden influx of players you weren't expecting? Having cloud on-demand via a hybrid model means you avoid the dreaded server crashes that too often hijack what should be an epic PR opportunity, without committing to the full-time cost of additional compute.
1. Find the right supplier
Work out the optimum machine and VM configuration across your game server fleets and game types. Then, research best-fit suppliers that pass the server build requirements, have the capability to scale, and operate technically advanced processes when it comes to incidents.
With countless suppliers operating all over the world, deciding which to go with can be a minefield. And with a finite supply of machines, competition has never been tougher. A specialist intermediary can help you decide how much compute you need and make expert recommendations on type and location based on playtesting. As well as make sure the right compute is available in the right place, at the right time, they'll also ensure machines are quality assured, have the right operating system, and run smoothly.
2. Prepare more compute than you think you need
Don't get caught out when your studio is in the spotlight. Mitigate uncertainty around launch events and unexpected PR opportunities (like a streamer going live on your game and causing player numbers to spike) by making sure you have plenty of compute power prepared should you need it.
As a rule of thumb, your 'warchest' should be pre-stocked with around 2% to 3% of the farm for major regions like Europe, North America and Asia, along with roughly 5% to 6% for smaller regions like South America and Australia -- all kept warm and ready to go at a moment's notice. This means you're able to deploy new machines almost immediately, rather than the usual six hours for major regions.
While bare-metal likely forms the backbone of your compute, cloud offers more flexibility for short-term player spikes. So, ideally you also want the ability to scale on demand using hybrid cloud for peak times.
3. Allocate enough time and resource
Despite the automated world we live in, managing a compute supply chain is, at least for the time being, largely a relationships game. And make no mistake: building and maintaining relationships with up to a dozen different suppliers (depending on your game) takes time.
Plus, there's the machine preparation and deployment. With different UI and AI layouts for every supplier, it can be quite the task. Outsourcing your supply chain management means you benefit from specialist expertise, while keeping your team focused and agile.
"Mitigate uncertainty around launch events and unexpected PR opportunities by making sure you have plenty of compute power prepared should you need it"
4. Keep an eye on your costs
Studio budgets are unwieldy at the best of times. If you want to keep the accounts department on side, you'll need a working operating cost model to understand the real-term expense of running your game. Different servers operate on varying monthly or yearly finance arrangements; plus, you need to keep a keen eye on machine purchasement and removals based on different Terms of Service conditions.
To keep your game going, game server orchestration from Improbable Multiplayer Services (IMS) scales automatically into the more-costly cloud only when the initial bare-metal setup is fully booked. To minimise costs, game servers are shifted back to bare-metal the moment more becomes available. Not only that, the expert team is constantly reviewing new hardware trends that could potentially reduce your operating costs.
5. Always have a Plan B
Tech breaks. It's a fact of gaming. As well as keeping track of on-going maintenance work (planned outages) across all your suppliers and putting backups in place, make sure you have mitigation strategies in place for the unexpected outages, too. Unified processes such as fail-over or ticket escalation ensures everyone involved knows exactly how to get your game back online as soon as possible.
6. Look to the future
Games are always evolving, which means your compute needs are constantly changing, too. As well as periodically reviewing your machine utilisation for efficiency, regular testing and forecasting based on your content and release roadmap is crucial to ensure player experience isn't let down by insufficient tech.
Plus, new hardware is available all the time, which could improve player experience and even reduce your costs. To optimise your compute setup, you want to be continually monitoring trends and improving your network and server load to ensure a unified multiplayer experience.
Don't have the expertise or capacity in-house? Working with a direct-line support team gives you the strategic thinking and personalised support of real-life experts, without the added headcount.
How Gamepires managed SCUM's unexpected success with the help of IMS
Within hours of SCUM's release, expected concurrent numbers were surpassed 20 times over. Plus, because the first server build wasn't yet fully optimised, Gamepires had less game servers per machine than planned.
To maintain a smooth multiplayer experience and avoid the dreaded negative Steam reviews, Improbable Multiplayer Services deployed hundreds of additional machines per region within hours. By maintaining multiple suppliers and securing a unified performance through a proper supplier tiering, the team worked with Gamepires to take the launch from potential PR disaster to a total success story.
Want insider insights on live online multiplayer development, straight to your inbox?
Partnering with Improbable Multiplayer Services on your managed infrastructure is like having a compute concierge. We'll work with you to understand your game's needs, then choose the right machines for you and manage all your supplier relationships, so you only have to deal with one central contact. Plus, because we're solely focused on finding the optimum solution via the best suppliers for your game, we can guarantee impartial, tailored service that helps you achieve the best possible player experience.
About the author: Sebastian Reckzeh has over 20 years of experience in game server operations as a Founder of Gportal and a Founder and COO of Zeuz. Since Zeuz became part of Improbable in 2020, Reckzeh has taken on the mantle of VP Technical Operations for Improbable Multiplayer Services.