NotEnoughMemoryException when sending a large size message [14820]

[Nir-Az]

Hi,

I am trying to send an image throw a FastDDS writer and I get this error:

RTPSWriter:Serialization returns false

When I debug it I see that FastDDS throws NotEnoughMemoryException exception
My IDL contains sequence<octet> raw_data; and I am filling it with a ~2.5 [MB] data

I guess it's got to do with the MemoryManagementPolicy but it is not inforced.
If the message is unbounded, how can FastDDS accept PREALLOCATED_MEMORY_MODE?

Can someone please explain better the difference between PREALLOCATED_WITH_REALLOC_MEMORY_MODE and DYNAMIC_REUSABLE_MEMORY_MODE?

What is more efficient for streaming?

My use case is supporting streaming of different size images (user control) but once the size is chosen , it will send the same message size every time.

Thanks,
Nir

[JLBuenoLopez]

Hi @Nir-Az,

It is responsibility of the user to configure correctly its Fast DDS application. Even if the message is unbounded, Fast DDS allows using PREALLOCATED_MEMORY_MODE because internally Fast DDS allocates a estimated sample size. If at runtime the sample size is bigger that the estimated one, then the exception you mention is thrown.

Regarding the difference between PREALLOCATED_WITH_REALLOC_MEMORY_MODE and DYNAMIC_REUSABLE_MEMORY_MODE, I encourage you to read Fast DDS MemoryManagementPolicy documentation section. The main difference is that PREALLOCATED_WITH_REALLOC_MEMORY_MODE preallocates a certain sample size and if that size is not enough, reallocation is allowed. DYNAMIC_REUSABLE_MEMORY_MODE does not preallocate any memory when Fast DDS is initialized. All memory needed is allocated dynamically. Also, when the memory is released, instead of releasing it to the system, it is saved in a pool so it can be reused. This is the main difference with DYNAMIC_RESERVE_MEMORY_MODE which does not reuse the released memory. DYNAMIC_REUSABLE_MEMORY_MODE memory consumption may be lower than the PREALLOCATED_WITH_REALLOC_MEMORY_MODE one depending on other QoS configuration such as HistoryQosPolicy, ResourceLimitsQosPolicy, etc.

I am afraid I cannot help you with your last question. You should probably test your application using both modes and see what works best for you. Probably, in a steady state, it would not matter.