How to get high performance on repair huge MYISAM tables?

[Jerry Tian]

Hardware: AMD X2 5000+/8G DDRII 800/RAID0(2X120G normal SATA disks)
Software: ubuntu 8.0.4 x86_64 server edition/MySQL GA 5.1.30 binary install

my.cnf:
////////////////////////////////////
key_buffer = 4096M
max_allowed_packet = 16M
table_cache = 1024
sort_buffer_size = 256M
read_buffer_size = 64M
read_rnd_buffer_size = 1M
myisam_sort_buffer_size = 256M
thread_cache_size = 8
thread_stack = 128K
#thread_concurrency = 4
tmp_table_size = 2048M

#
# * Query Cache Configuration
#
query_cache_limit = 1048576
query_cache_size = 16777216
query_cache_type = 1

////////////////////////////////////

This server is acting as a slave to the remote production master via ADSL link(not that very stable), mainly for BI report generation. The biggest MyISAM table is about the size of 100G, having about a billion rows of records in it, it is also partitioned. The second largest one is about the size of 50G.

Recently, I have to repair this tables for some times, or the server can not continue its replication. But the process is really miserable, it almost takes a day to finish. The command I used is:

mysqlcheck --auto-repair --fast --all-databases

I am not using "--quick" option since according to the help manual, it will only check index file of tables, which is somewhat too risky for me.

I wonder if I could speed this process by tuning some server settings or use other tools? I have used "iostat" and "top" to monitor the server status during the process, the CPU have little chance to reach over 90% usage; and the RAID0 combination's read and write performance just struggle below 5MB/s. Very pool, isn't it?

Any suggestion is appreciated. ;-)