ways to improve the performance of this scenario

Question

I have a map with huge amount of data being populated (around 300,000 records in approx)

and iterating it as below ,

    for (Map.Entry&         


        

Answer 1

The for loop is taking time due to two reasons.
1) Individual Email improve it by less Transport connection
2) Individual commits improve it by

So Ideal is to handle both, I would recommend do it for batch of 1000, then play with numbers

Example

int BATCH_SIZE = 1000
conn = DriverManager.getConnection("username","password");
conn.setAutoCommit(false);
Statement stmt = conn.createStatement(
        ResultSet.TYPE_SCROLL_INSENSITIVE,
        ResultSet.CONCUR_UPDATABLE);
int count = 0;

Map<String, String> emails_map = new HashMap(BATCH_SIZE)<String, String>;
for (Map.Entry<String, List<ClassOBj>> entry : testMap
        .entrySet()) {
    String email = get_useremail();
    String const_val = do_magic(); // this is how you are computing some constant
    String body = construct_body();

    count++;
    String SQL = "YOUR UPDATE STATEMENT";
    stmt.executeUpdate(SQL);  
    emails_map.put(email, body); // can create 
    if (count % BATCH_SIZE == 0) {
        //commits all transcations
        conn.commit();
        //bulk send emails sending 
        //http://stackoverflow.com/questions/13287515/how-to-send-bulk-mails-using-javax-mail-api-efficiently-can-we-use-reuse-auth

        bulk_emails_send(emails_map)
    }

}


public void bulk_emails_send(Map<String, String> emails_map) {
    // Get the default Session object through your setting
    Session session = Session.getDefaultInstance(properties);
    Transport t = session.getTransport();
    t.connect();
    try {
        for (String email_id in emails_map) {
            Message m = new MimeMessage(session);
            //add to, from , subject, body
            m.saveChanges();
            t.sendMessage(m, m.getAllRecipients());
        }
    } finally {
        t.close();
    }
}

Answer 2

Instead of updating the database in each loop.Try to update the database after completing the loop.

There are different way to optimize the large database update. Best one of that is

• Insert modified values into a temporary table
• Update the original table from the temporary table

Also use a thread based queuing mechanism to send your emails

Answer 3

You should use jdbc batch update feature.

While iterating over map you add batch to your prepared statement. When you has added (say) 2000 records, you call stmt.batchUpdate() which will update 2000 diferent records in fast way.

Some example you can see here:

http://www.mkyong.com/jdbc/jdbc-preparedstatement-example-batch-update/

Second thing - If you can, make transaction commit after each batchUpdate. Transaction for 300k records may be to much for your database configuration. Splitting this update into many transactions will have better performance - but only if you can not to have transaction on all records.

Answer 4

I'll try to summarize all good points mentioned above.

Your options are,

1. Use multithreading wherever possible but keep it mind that multithreading comes with a cost of additional memory(heapdumps and application will be down).

2. Another good option is use bulk update but again bulk update comes with cost of more lock time at database end. so use it wisely as threads will be waiting till one thread finishes with updating as update will take exclusive lock i.e. no sharing possible.

3. Try threading in along with async jobs (specially for mails) try to use different server/process for mail handling and send mail requests to mail server asynchronously which is quite fast as your mail generation task is handled by other process (In real life mail exchange can accept some delay thus your slow mail server is allowed but application server is not.)

4. If possible move your complete Db update logic in stored procedure on DB server which will save your lot of time (Rule of Thumb : Always let A handle tasks for which A is designed/optimized i.e. DB are designed for faster DB operations than pgming languages.)

Couldn't give you answer specific to your need but I hope this was helpful in terms of improving it :)

Answer 5

My suggestion is you can use Stored procedure. or use can use batch wise update instead of this.

More about sql batch update.

Answer 6

I would make something like this. Prepare the data for the operation.

I suppose you are updating a table like user which should have a unique Id.

Map<String, String> emailIds = new HashMap<String, String>();
Map<String, String> emails = new HashMap<String, String>();
for (Map.Entry<String, List<ClassOBj>> entry : testMap.entrySet()) {
 -- DONOT DO THIS// send email with map keys as email'id
 -- DONOT DO THIS// email content is populated from the list
 -- DONOT DO THIS// Perform a sql update to the column with the dyanamic value generated here with the email'id
emails.put(emailId, content);
emailIds.put(id, emailId);

}

bulkEmailSend(emails);
bulkEmailUpdate(emailIds);

bulkEmailSend and bulkEmailUpdate are the methods which should be written to make the appropriate calls.

So, use the bulk email send and bulk emailId update methods to update the values back to the database.