A real-time operating system (RTOS), or similarly, a general-purpose operating system (GPOS) with real-time extensions, has design attributes to improve the determinism for each of the application states required for operation. The GPOS system is typically the type of system you use every day to check email, watch videos, etc. 

     The RTOS systems may also be used every day, but it performs different services such as, deploying airbags or anti-lock brakes during a detected event. Those types of systems are time sensitive as one can imagine. They need to perform life-saving measures on time, every time. A system is said to be deterministic if it's possible to predict the timings of its computation precisely, and the computations are logically correct. There are two key aspects that make an RTOS "real-time"; the first is to make sure that the processor can accept an interrupt at any time or most of the time. There is almost no time where the operating system cannot accept an interrupt. Some applications need only an average response time, while others require that every deadline is met every time. The second key aspect to "real-time" is to ensure very low cycle time for the scheduler to schedule any deferred workload.
     So why not just use Linux? There are two basic time requirement genres called soft real-time and hard real-time computing. Soft real-time computing provides a guarantee of a specific level of CPU bandwidth in a specific unit of time. For example, an application that needs 10 milliseconds of CPU bandwidth and must have that requirement met within 100 milliseconds has a soft real-time requirement. This is often where a GPOS such as Linux is used. Hard real-time computing looks at the response time rather than at a bandwidth guarantee, when an application must respond to an event within a specific time. One example would include responding to a periodic interrupt where the worst-case response time must be less than the interrupt time. This ensures that no interrupt events are lost.

     There are methods of making Linux behave more like a RTOS. One such method is the preempt kernel approach, where the Linux kernel is modified to reduce the amount of time the kernel spends in non-preemptive sections of code. All interrupts are initially handled by the core and are passed to standard Linux only when there are no real-time tasks to run. However, when timing is critical and on the level of millisecond thread capability, a true RTOS will often provide you more deterministic results. If your embedded application requires split second timing, consider starting with one of the many non-Linux, open-source RTOS's found here. 
     Other Reference Links: Walls,  Barry & Crowley,  Koolwal