Skip to main content

High Scalability:Denomalizing data for Billions of files for scaling snapshot times

We store metadata about billions of files in Mysql shards and each shard has Folder,File and version table. The schema looked like  this.

Now each file can have n versions and some customers want infinite versions. Problem with scaling snapshot query is that the file system snapshot required information about latest version only. To get latest version you need to join folder, file, version table and discard older rows. I had described the consistent scaling challenge and the improvements we had done to improve snapshot times for our cloud file system in

The latest improvement we did is to denormalize the information about latest entry on to file table itself.

Now to generate a snapshot we just need to join Folder and File table. And the improvements are huge. We use box ananometer to log slow queries over all databases and the second query is normalized query and third query is denormalized query. Across all databases we are tracking the time has reduced 3 times  (486 sec to 159 sec), also no of rows examined and rows sent has reduced by half.

Here is the normalized query graph

Here is the denormalized query graph
Now denormalizing billions of rows have unique challenges and you need to do it without impacting customers. We spread billions of rows across 28 master and there are 28 slaves for these masters. 

To do performance testing we took the biggest database with customers having 27M rows and imported it in a test environment and migrated it.

27M versions snapshot times before denormalization = 1.2 hours
27M versions snapshot times from denormalized tables = 6.5 minutes constant with or without caching

Off course as we are doubling the data we need to optimize the database tables after denormalization else we were running into row chaining problem.

For production go live as usual we started with feature flags and we added 2 flags latest_entry_migrated and latest_entry_active field on customer model. Then we wrote code that on basis of latest_entry_active flag would execute normalized or denormalized query. Once the code was live in all services then we began migration for few workgroups to test for any bugs. We migrated each data centre every weekend and within a month we had all databases upgraded with denormalized rows and snapshot queries have even gone from slow query logs from many databases.

One sideeffect of this denormalization as this opens up gateway for us to implement infinite versions because now we can sub shard versions on different tables and even in different databases.


Popular posts from this blog

RabbitMQ java clients for beginners

Here is a sample of a consumer and producer example for RabbitMQ. The steps are
Download ErlangDownload Rabbit MQ ServerDownload Rabbit MQ Java client jarsCompile and run the below two class and you are done.
This sample create a Durable Exchange, Queue and a Message. You will have to start the consumer first before you start the for the first time.

For more information on AMQP, Exchanges, Queues, read this excellent tutorial
import com.rabbitmq.client.Connection; import com.rabbitmq.client.Channel; import com.rabbitmq.client.*; public class RabbitMQProducer { public static void main(String []args) throws Exception { ConnectionFactory factory = new ConnectionFactory(); factory.setUsername("guest"); factory.setPassword("guest"); factory.setVirtualHost("/"); factory.setHost(""); factory.setPort(5672); Conne…

What a rocky start to labor day weekend

Woke up by earthquake at 7:00 AM in morning and then couldn't get to sleep. I took a bath, made my tea and started checking emails and saw that after last night deployment three storage node out of 100s of nodes were running into Full GC. What was special about the 3 nodes was that each one was in a different Data centre but it was named same app02.  This got me curious I asked the node to be taken out of rotation and take a heap dump.  Yesterday night a new release has happened and I had upgraded spymemcached library version as new relic now natively supports instrumentation on it so it was a suspect. And the hunch was a bullseye, the heap dump clearly showed it taking 1.3G and full GCs were taking 6 sec but not claiming anything.

I have a quartz job in each jvm that takes a thread dump every 5 minutes and saves last 300 of them, checking few of them quickly showed a common thread among all 3 data centres. It seems there was a long running job that was trying to replicate pending…

Email slavery

It seems I have become an EmailSlave. The first half of the day is spent in just answering to emails. There are so many emails where I am copied but I need not be. There are many emails  where its a 1-2 page email and somewhere down someone says @KP please answer this.  So it seems daily my work schedule is:
Signin to newrelic and check anomalies for 15 min. Check emails related production exception report and yes there are a ton of these report daily. Need a better tool here as this model is not scalable. I need to reduce the incoming data at me to only see relevant data like what newrelic does. May be I need to create a webapp out of these emails.Check emails for next few minutes before team callsDo team callsThen again back to checking emails until a I have taken a best shot at answering everyone waiting for my reply.Attend team meetings on Tue/Thu
Being an architect and coder at heart I don't feel satisfied at end of the day if there is nothing tangible getting done at the end.…