That seems pretty bug to me. Mine are all under a meg.
When you run 'rrdtool info somebigfile.rrd' does it have a lot of different data-sources ("ds")? Mine tend to directly correlate.
Here's some output from looking at a couple average rrd files for me.
$ ls -l la.rrd vmstat.rrd -rw-r--r-- 1 xymon xymon 19640 Jan 29 14:09 la.rrd -rw-r--r-- 1 xymon xymon 323384 Jan 29 14:24 vmstat.rrd $ rrdtool info vmstat.rrd | less $ rrdtool info la.rrd | less $ rrdtool info la.rrd | fgrep 'index = ' ds[la].index = 0 $ rrdtool info vmstat.rrd | fgrep 'index = ' ds[cpu_r].index = 0 ds[cpu_b].index = 1 ds[cpu_w].index = 2 ds[mem_swpd].index = 3 ds[mem_free].index = 4 ds[mem_buff].index = 5 ds[mem_cach].index = 6 ds[mem_si].index = 7 ds[mem_so].index = 8 ds[dsk_bi].index = 9 ds[dsk_bo].index = 10 ds[cpu_int].index = 11 ds[cpu_csw].index = 12 ds[cpu_usr].index = 13 ds[cpu_sys].index = 14 ds[cpu_idl].index = 15 ds[cpu_wait].index = 16 $ rrdtool dump la.rrd | wc 2399 18567 195686 $ rrdtool dump vmstat.rrd | wc 2991 19271 1030655