With all the buzz around modular interoperability and MOSA, designers of U.S. military systems must also recognize the benefits of shifting to distributed computing architectures. Centralized single-box systems like CMOSS, OpenVPX, and other board/backplane chassis promise to ease integration challenges and vendor lock-in but hosting multiple capabilities inside them is the worst kind of “modularity”.

Why military systems need to migrate from centralized to distributed architectures.
Modern computing environments need rapidly customizable and scalable distributed systems.

Centralized computing breeds obsolescence
Centralized architectures like VME and OpenVPX are based on boards connected to a common bus within a chassis (see Figure 1), all following the same standards. SOSA restricts these boards down to the pins on the connectors—a mandate that locks the system to the original configuration at the expense of any evolution at the card, chassis, or top-level vehicle/airframe level.More specifically, these systems pose several challenges for hardware designers and system architects:

• Having all components comply to the same standards can hinder migration to new hardware and interconnect types (think of moving from VME to OpenVPX—the cards are incompatible and hence not able to be migrated directly). This prevents the system from keeping pace with evolving industry products and mission requirements.
  
  
• As all boards are housed within thesame chassis, component layout is restrictedto close proximity with each other and powermust be supplied by a single source. Thiscan be big power, too—up to thousands ofWatts for a full chassis. This locks engineersinto less-than-optimal system designs for theintended vehicle or platform since the powersupply and heat dissipation may “fix” whereand how the chassis is installed.
  
  
• Boards require a common backplane,which is a complex, expensive, andcustomized component, that may degrade high-speed signal integrity and adds to theoverall system cost. The backplane also maynot handle some signals like 1000Base-KRor fiber optics very well, and the signalsinterconnection scheme can’t be changedwithout redesigning the backplane. If evenone signal pair changes, this can shatter thepromise of interoperability between vendors.
  
  
• A sealed, bus-based architecturerequires the chassis to be either air- orconduction-cooled and the boards insidemust have impeccable thermal conductivityto the chassis sidewalls. This box representsa high-heat entity in the vehicle or platform,limiting the choices for location, mounting,and heat dissipation. And it’s an all-or-nothing approach: one chassis that must coolall cards and elements inside.
  
  
• Expanding system capabilities isconstrained by the availability of backplaneslots, power, and cooling resourcesprovisioned during the first release, meaningthese systems have limited flexibility andgrowth potential. Adding more often meansupsizing the whole chassis or adding anotherbox to increase capabilities—essentiallybecoming a distributed architecture withbetter options presenting themselves.

Distributed architectures are true fit-for-purpose

Distributed architectures offer many advantages in terms of flexibility, scalability, and future growth potential. De-centralizing system functions throughout the vehicle or asset allows engineers to place components strategically in different locations to optimize wiring, power distribution, and cooling (see Figure 2). By distributing functions across modules, it becomes easier to upgrade and replace individual components, versus having to replace the system as a whole, improving the system's adaptability to future mission requirements.

Other advantages include:

• Giving engineers more options toaddress heat dissipation challenges. Bydistributing components throughout thevehicle, the heat load can be shared acrossa larger area rather than concentrating ona single location – problematic in smallervehicles and confined environments.

• Greater flexibility in electrical powerand data distribution. With interconnecttechnologies like Thunderbolt™ 4, a singlecable can send power and data to simplifycabling infrastructure and reduce weight.

• More granularity in componentupgrades to reduce the impact ofobsolescence on the entire system.Programs can replace individual modulesor subsystems without requiring whole-chassis swap out or extensive system-wide modifications. As hardware designers and system engineers, it’s crucial to embrace distributed architectures for new projects and modernization programs. Decentralized components and interconnects are the key to building computing systems that are flexible, adaptable, and ready to meet the challenges of modern defense operations.