Ben Laurie <[log in to unmask]> wrote
Tue, 22 Sep 2015 15:26:12 +0000:

| > For a rough idea of how much resources a catlfish log uses, here's the
| > amount of RAM
| 
| 
| Measured in B? kB? MB?

It's kB, as output from ps -ely on a linux system.

I realise it's not such a great measurement. First because it was done
directly after startup and second because it didn't include SZ. I'll try
again (RSS and SZ in kB):

name         RSS    SZ
--------------------------
frontend-1:  147552 442948
frontend-2:  334228 441220
storage-1:    51908 213350
storage-2:    49588 195385
signing-1:    55812 178553
signing-2:    52240 177401

148M seems suspiciously low for 5.7M entries though, since frontend
nodes keep the tree in memory. The machine does use the swap, according
to free(1), so there is a chance frontend-1 has pages swapped out. I
don't know. Let me try some better RAM benchmarking on a less crowded
machine later.

Looking at SZ (virtual size), it seems like a machine running
frontend+storage should be happy enough if given 1GB of RAM for
catlfish. But I should verify this with more than output from ps.