SWAP size on systems now having lots of RAM

Latest response

year = 2018. is the SWAP partition obsolete when your system has more than 500GB of RAM and the system disk is 300GB?

This is not on cluster systems having many nodes which may only have 32GB of RAM per node

I am talking about shared memory types of servers or systems which can have either 2 or 4 sockets and easily upwards of 60 cores where there is >= 24 DIMM slots and with 32gb RDIMM and 64gb LRDIMM now the minimum amount of RAM oftentimes is now 128gb on higher end performance oriented servers.

For example some of the servers we are using have 512GB of RAM using RDIMM's, and we are looking to get some newer servers having more than 1 TB of RAM; yes we need the RAM to run certain engineering codes for design & analysis; bigger models = more RAM

Now sometimes we don't need lots of disk space, and I am still making use of or want to use up some 300 GB 2.5" SAS hard drives; with the linux OS installed and all software needed the drive is less than 20gb used and software creates maybe 1-5gb of worth of output data, so there is nothing requiring lots of disk space.

In such a case where a system has more RAM than disk space, what purpose if any does it server to create a SWAP partition on the disk? There is always that warning when building a system stating you have no swap partition and it implies there might be a performance or some other problem.

if there is 500+gb of ram and the system runs out of ram and locks or crashes, then at that point fine. And my experience has been any time the OS starts needing swap that the system basically locks up anyway. Is there any other purpose of the SWAP partition on disk, assuming my understanding of it correct that it's only purpose is a fallback storage resource when RAM was full?

My preferred partition layout is 2 partitions: (1) efi boot partition, and (2) root partition as either ext4 or xfs. No swap.

Responses