Media & Gaming
When a title or stream goes down, does your monitoring go too? We keep monitoring on separate kit, and supply and integrate all of it.
Request a quotePublic record, not our clients
The monitoring went down with the game.¹⁵
Roblox’s post-mortem on its 73-hour outage, October 2021: the monitoring that would have shown the cause relied on the affected systems.¹⁵
- 01
In your requirement
Monitoring keeps working when the game’s own backend has failed
- 02
We check
What your monitoring depends on, tested with that dependency off
- 03
First step
Send what you need to buy and by when. You get a reply the same day, then a quote.¹⁴
Distribution, billing and vendor payments, all disrupted.¹⁶
Publisher Lee Enterprises’ SEC filing on a February 3, 2025 attack. We write into your requirement: what publishes, bills and pays has a stated restore order and time. We check when it last ran.
What we design.
Render nodes and storage for your next title? TrendForce put server CPU and board lead times near a year in April.¹⁷
- A studio contract requirement, assessed through TPN.
- Version in force. v5.3.2.
- Our coverage. Lateral movement blocked, shown at acceptance.⁶
- Between sites. Content is encrypted at line rate: AES-256-GCM, with NIST post-quantum algorithms.⁶⁻⁷
- Covered by. Security · Networking
- 48 GPUs for render and upscaling, for a broadcaster.⁸
- Six nodes. Eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs in each.⁸
- To the media store. 200 Gb/s per node.⁸
- 40 racks into a colocation hall, with no maintenance window.⁹
- One rack pair at a time. Twenty pairs; each old pair is drained by the load balancer before anything is unplugged.⁹
- Burned in first. Each new pair runs 48 hours under full load before any service moves onto it.⁹
- AWS to Google Cloud for a game studio's live telemetry: at least 6 Gbps at launch-weekend peak.¹⁰
- Encrypted. MACsec on every physical link between the providers' routers.¹⁰
- Fallback. Telemetry continues through the colocation edge at 2 Gbps or more while the link is down.¹⁰
- Six backend services, three cloud regions, one launch.¹¹
- Cases tested. At least 12, each with a stimulus and a measurable response.¹¹
- Report. Issued within 15 working days of the workshop.¹¹
- 40 of 40 services send traces, metrics and logs over OTLP.¹²
- One trace per session. Login, entitlements, matchmaking, allocation, match start.¹²
- Paged fast. Within 5 minutes of the burn-rate threshold.¹²
- AWS and Azure costs in one FOCUS dataset; shared clusters split by rule.¹³
- Split rule. By requested CPU and memory per namespace.¹³
- Allocated. At least 95% of monthly spend, to a named team or the platform owner.¹³
- A studio contract requirement, assessed through TPN.
- Version in force. v5.3.2.
- Our coverage. Lateral movement blocked, shown at acceptance.⁶
- Between sites. Content is encrypted at line rate: AES-256-GCM, with NIST post-quantum algorithms.⁶⁻⁷
- Covered by. Security · Networking
- 48 GPUs for render and upscaling, for a broadcaster.⁸
- Six nodes. Eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs in each.⁸
- To the media store. 200 Gb/s per node.⁸
- 40 racks into a colocation hall, with no maintenance window.⁹
- One rack pair at a time. Twenty pairs; each old pair is drained by the load balancer before anything is unplugged.⁹
- Burned in first. Each new pair runs 48 hours under full load before any service moves onto it.⁹
- AWS to Google Cloud for a game studio's live telemetry: at least 6 Gbps at launch-weekend peak.¹⁰
- Encrypted. MACsec on every physical link between the providers' routers.¹⁰
- Fallback. Telemetry continues through the colocation edge at 2 Gbps or more while the link is down.¹⁰
- Six backend services, three cloud regions, one launch.¹¹
- Cases tested. At least 12, each with a stimulus and a measurable response.¹¹
- Report. Issued within 15 working days of the workshop.¹¹
- 40 of 40 services send traces, metrics and logs over OTLP.¹²
- One trace per session. Login, entitlements, matchmaking, allocation, match start.¹²
- Paged fast. Within 5 minutes of the burn-rate threshold.¹²
- AWS and Azure costs in one FOCUS dataset; shared clusters split by rule.¹³
- Split rule. By requested CPU and memory per namespace.¹³
- Allocated. At least 95% of monthly spend, to a named team or the platform owner.¹³
Written by named engineers.
MPA best practices
The Content Security Best Practices version in force.⁶
AES-256-GCM
Content between sites, encrypted at line rate.⁷
Post-quantum ready
With NIST post-quantum algorithms.⁷
48 hours burn-in
Each new rack pair, under full load, before any service moves.⁹
No maintenance window
40 racks moved into a colocation hall, one pair at a time.⁹
Splunk or ThousandEyes for monitoring. Arista or Cisco for the fabric.
NVIDIA GPUs for render. Akamai for delivery. We supply them all.
Our Networking Blueprint is written by Diksha Shah, Steve Phillips, David Morgan, Avinash Dhinakaran and Mark Winston.
Read the blueprintDesign studies
Design studies in this sector
Talk to an expert.
Send what you need to buy and by when. You get a reply the same day, then a quote.¹⁴
Talk to an expert.