top command showing too much swap usage by several processes but the system hasn't such swap memory installed in it. So why it is showing such huge usage in top?

Evidence.
Then show your analysis - maybe someone can point out the error.

depends on your definition of "too much".

Page fault handling ALWAYS goes on. When memory pressure gets high enough code pages will get dropped, and reloaded as neccessary. The code pages come from the binary executable file but are nearly always loaded via page faults, and this can/will show up as "swapping" activity.

It is possible that you have some (maybe, relatively inactive ...) processes which have memory leaks. Since all user-land memory is virtual, this becomes occupied space in swap. The application is no longer referencing the "leaked" memory, so it eventually gets swapped out, but the application's total virtual-memory allocation continues to grow.

If an application's total VM-space continues to grow, but its so-called "working-set size" (the number of pages that must be in-RAM for the process to run efficiently) generally does not, "you probably have a 'leaker' on your hands."

This is a bug in the program. Or, in the case of a server, it might be a bug in one of the client programs.

Since most programs these days are written using languages which provide a "mark-and-sweep garbage collector," the most common cause of leaks is "circular storage references." A block of memory won't be deallocated until there are no remaining references to it. But, consider the case where there are two objects, A and B, and each contains a ("strong ...") reference to the other. Even if no one else is still referring to either object, the two objects continue to refer to each other and therefore keep either of them from being reaped.

And to complete the thought: what you can do about it, in designing a program, is to use "weakened" references, which tells (the software's, not Linux's ...) memory manager that "you can break this link, if you need to." If, say, the reference from A to B is designated as "weak," the garbage collector might decide to break it. Now, B (say ...) has no remaining "strong" references to it, so B can be reaped, and since that (say) means that A now has no remaining "strong" reference to it, A is reaped also, and the leak does not occur.

Mind you, these are programming language issues ... they are bugs. Linux isn't doing the wrong thing. The programs need to be fixed.

Just to add something that may help.

Remember Linux is not Windows. It normally loads much into memory, using that resource to its advantage.
It will then quickly swap out or just clear space if there is no place to swap the memory resident code,
on demand, when another process needs that space.

If there is no swap configured as you seem to indicate, no wonder,right?
Why run without any swap? Swap does not need to be a dedicated partition. You can have a swap file.

Hope that helps,
Tim

It is showing kilobytes, right?

We might know if the OP ever responds with the data to support their analysis/conclusion.

So far, All I hear is crickets !?!
(From the OP)

Well, Linux is very aggressive in its use of buffers, but the accumulation of data in swap implies, to me, the presence of memory that has been allocated by an application program, but then never again referenced. Eventually, the data gets pushed out to the swap file, never to be called-back again because it was "caused by a leak."

The "amount of virtual memory used" by the suspect process will also be "steadily increasing," although the working-set size may not be large at all.

Not exactly.

Linux uses swap to load executable code. If the code is sleeping, it will release those pages for other use, and when they DO become active, the pages are recalled via swap activity.

1 members found this post helpful.

The OP stated this " but the system hasn't such swap memory installed in it".

I'd start by finding out if indeed some swap file or partition is really being used and how much of it is available.

Various distro's use different tools to manage. One example. https://access.redhat.com/documentat...and_Usage.html

@jpollard,

Thanks for the correction/clarification