Julia : 如何进一步改进操作Redis的效率？

最近在用Julia脚本操作Redis内存数据库，但是感觉Julia的脚本效率比较低。还没有找到相关办法。

julia> using Redis;

julia>  conn = RedisConnection( host="127.0.0.1", port=6379, db=0);

julia> trans = open_transaction(conn);

julia> mydat= Dict{Any,Any}();

julia> @time for i =1:10000 setindex!(mydat,"abcdefghijklmn",i) end;
0.011981 seconds (47.48 k allocations: 1.079 MB)

julia> length(mydat)
10000
julia> mydat2 = mydat;

julia> @time hmset(trans,"mydat_1",mydat)
1.958942 seconds (382.23 k allocations: 2.892 GB, 29.62% gc time)
"QUEUED"

julia> @time for i =10001:1000000 setindex!(mydat,"abcdefghijklmn",i) end;
5.918395 seconds (10.73 M allocations: 228.507 MB, 59.50% gc time)

julia> @time hmset(trans,"mydat_2",mydat)
ERROR: InterruptException:
in string at ascii.jl:44
in pack_command at C:\Users\Administrator\.julia\v0.4\Redis\src\parser.jl:74
in execute_command at C:\Users\Administrator\.julia\v0.4\Redis\src\parser.jl:83
in hmset at C:\Users\Administrator\.julia\v0.4\Redis\src\client.jl:76

julia> @time hmset(conn,"mydat_conn",mydat2)
1.901253 seconds (379.91 k allocations: 2.892 GB, 28.90% gc time)
true
julia> @time reData_conn= hgetall(conn,"mydat_conn");
0.346850 seconds (343.77 k allocations: 18.921 MB, 6.67% gc time)

julia> typeof(reData_conn)
Dict{AbstractString,AbstractString}

julia> length(reData_conn)
10000

问题：当KEY-VALUE在1万量级时，基本运行正常；但达到100万量级时，效率极其低下，只有中断执行。这个问题，不管是否使用事务，基本一样。
事务的效率并不明显，不知为什么？使用不正确？