Java interview question

Hello i had an online test in some company and i didn't succeed in doing it(they gave me a timeline) this was the test:

Problem:

1) You have a bunch of log files in:

 C:\CloudShareCodeChallenge\Challenge 

 Sample data with expected results (below) in C:\CloudShareCodeChallenge\Sample

2) Each file may contain two types of records, user records and payment records; you must process all files

3) User records are in the form of UR,<User Identifier>,<First name>,<Last name>

 User records are guaranteed to be unique

4) Payment records are in the form of PR,<Payment Identifier>,<User Identifier>,<Amount>

 There is a bug in the payments processor, and same payments MAY appear more than once

(hence double booking)

You should only take into account one payment per payment id.

5) Important notes:

You may assume all records are correct (parsing wise): only one comma per field, no decimal 
 point in amount and no missing fields, so you may use split etc. for that.

Files are not ordered, so depending on how you read the files you may get a payment record
 before the corresponding user record 

Each file name contains a two letter field records_M2_2TFA7.log only payment records starting
 with those two characters will be in the file

Payments with same prefix may span on more than one file.

All log files must be processed

Keep it clean and simple

If you feel you must add comment, rewrite the code to be self-explanatory.

All the clues/notes above are important :)

6) Your tasks:

Provide the list of the top 10 paying users: Full name and amount

Provide the most common and least common first names with count

Note: There may be more than one name

Provide the number of double booking payments:

   Either in thousandth of percent from the total number of payments 3 decimal points

   For 5% you should show 50.000

   For 0.2345% you will show 2.345

   Or a fraction 12,3435 / 100,000,000

Bonus Questions: How would you write the generator for the above problem

What would you take under consideration, what would be the challenges

Sample data processing results: Payments Ratio = 45.898 (235 / 5120)

Most common name(s): Hugh: 2

Most common name(s): Julio: 2

Most common name(s): Nelson: 2

Most common name(s): Sofia: 2

Most common name(s): Zelma: 2

Least common name(s): Annabelle: 1

Least common name(s): Chandra: 1

Least common name(s): Clayton: 1

Least common name(s): Cody: 1

Least common name(s): Gay: 1

Least common name(s): Guy: 1

Least common name(s): Jerri: 1

Least common name(s): Jessie: 1

Least common name(s): Julianne: 1

Least common name(s): Serena: 1

Top paying users:

Nelson Sensabaugh : 42669

Hugh Maginnis : 39375

Annabelle Glade : 30252

Jerri Bartee : 26393

Chandra Bottorff : 21221

Julianne Deller : 19624

Zelma Lubinsky : 17051

Nelson Caicedo : 16191

Sofia Lucena : 13779

Guy Gratton : 12150

There were 545 files and in each one there were approximitally 260,000 lines Where each line can be a payment record and the line is looking like this one: PR,U3VT2406XY5AC2,I8MKX,53 (of course Payment Identifier,User Identifier,Amount are not the same for each line)

And there can be lines which represent user record and those lines are like this one: UR,B7R1W,Nelson,Hornyak (of course User Identifier,First name,Last name are not the same for each line)

They gave an eclipse environment and said to do it in Java(Regular java programming),but I didn't know how and what things to use? Do I make an HashMap of more than a 100 million entries? Do I use different things? What is the algorithms that I should use and how? I just graduated so unfortunately I didn't succeed in doing this test but I really want to know how to do it?