Number of close oligoes for 25-mers.
#MaxDiff	closeCount
0	1
1	76
2	5476
3	378076

Phase 1  - using normal hash function on 25-mers testing on chromosome 22 only.

5	203M	28.9
10	295M    32.2
20 	479M	44.0
30	663M	56.9

18 Meg per oligo
~ 1.2 S/oligo + 23S  Nonlinear


Phase 2 - using custom hash on chr22 only.
#oligos	memory	time	hashHitRatio timekkr1u01
5	190M	25.4	0.056
10	204M	29.7	     	20.6
20	234M	33.2
30      263M    36.9
40      292M    41.3
80      410M    56.3	0.710
100	468M    63.2	0.892

3 Meg per oligo + 175M
~ 0.35 S/oligo + 24S

on chr21 + chr22
#oligos	time
10	 56.1
100	101.5

on chr5 (about 180Meg)
#oligos	time
10	117.0
100	195.0

Extrapolating to genome-wide looks like if you
just did masked, it might take 20s/oligo.
On chromosome 22 have 15 million oligos.  
300 million seconds
300,000 cluster seconds
100 cluster hours

Looks like could go 70x faster just looking for up
to 2 mismatches.  This would take it down to 1.5 cluster
hours for 22, and 150 for the genome.

on chr5 with maxDiff=2
#oligos	memory	time	maxDiff
100	302M	120.5	1
100	306M 	121.8	2