0
I recently started running memcache-tool-ng to analyze my slabs etc. The results that I get back make me think our implementation is horrible, at best. I'm looking for some pointers as to how we should best configure our memcache instances. I think one thing that may be a bit different than the 'norm' is that we store bigger sized keys so I've tried to gear our setup towards that. We have a max key size of 100MB, a chunk growth factor of 1.4 and a 1M max slab chunk size. Here is the output of the memcache-tool-ng
# Item_Size Max_age Pages Count Used Mem To Store Efficiency Evicted Evict_Time OOM
1 96B 2179s 1 1 100.00 MB 76 b 0.00 % 0 0 0
2 136B 959s 1 256 100.00 MB 32.38 KB 0.03 % 0 0 0
3 192B 751s 1 38 100.00 MB 6.88 KB 0.01 % 0 0 0
4 272B 62489s 1 545 100.00 MB 126.75 KB 0.12 % 0 0 0
5 384B 799s 1 100 100.00 MB 31.21 KB 0.03 % 0 0 0
6 544B 3210s 1 75 100.00 MB 33.13 KB 0.03 % 0 0 0
7 768B 885s 1 139 100.00 MB 88.05 KB 0.09 % 0 0 0
8 1.1K 9110s 1 105 100.00 MB 94.24 KB 0.09 % 0 0 0
9 1.5K 25931s 2 736 200.00 MB 960.59 KB 0.47 % 0 0 0
10 2.1K 7041s 1 155 100.00 MB 272.41 KB 0.27 % 0 0 0
11 2.9K 944s 1 102 100.00 MB 254.42 KB 0.25 % 0 0 0
12 4.1K 878s 1 116 100.00 MB 392.27 KB 0.38 % 0 0 0
13 5.7K 885s 1 41 100.00 MB 192.21 KB 0.19 % 0 0 0
14 8.0K 1708s 1 50 100.00 MB 319.13 KB 0.31 % 0 0 0
15 11.2K 722s 1 15 100.00 MB 145.02 KB 0.14 % 0 0 0
16 15.6K 575s 1 7 100.00 MB 86.89 KB 0.08 % 0 0 0
17 21.9K 741s 1 7 100.00 MB 136.52 KB 0.13 % 0 0 0
18 30.6K 881s 2 30 200.00 MB 781.33 KB 0.38 % 0 0 0
19 42.9K 792s 5 3 500.00 MB 110.87 KB 0.02 % 0 0 0
20 60.0K 576s 2 4 200.00 MB 201.09 KB 0.10 % 0 0 0
21 84.0K 945s 24 251 2.34 GB 16.87 MB 0.70 % 0 0 0
22 117.7K 344s 1 1 100.00 MB 111.31 KB 0.11 % 0 0 0
23 164.7K 861s 19 47 1.86 GB 6.66 MB 0.35 % 0 0 0
24 230.6K 691s 5 6 500.00 MB 1.16 MB 0.23 % 0 0 0
25 322.9K 857s 16 15 1.56 GB 4.35 MB 0.27 % 0 0 0
26 452.1K 844s 23 19 2.25 GB 6.57 MB 0.29 % 0 0 0
27 632.9K 144s 17 1 1.66 GB 453.34 KB 0.03 % 0 0 0
28 1024.0K 315s 139 4 13.57 GB 13.07 MB 0.09 % 0 0 0