MongoDB Sharding

Hey Anubhav,

I am continuing to run tests on the Fireworks framework as a part of the CASCADE project. Many of my tests are at large scales and I have been trying to evaluate the bottlenecks on the system.

One bottleneck I have noticed is when multiple workers attempt to access the same, single mongoDB. Seemingly, the workers, in competing to access the centralized DB, are left waiting to access the resources (e.g. Fireworks).

That said, I have considered testing MongoDB’s sharding capabilities as a way to horizontally scale the database, reducing the load on any single DB instance. I was curious if you had any familiarity regarding sharding on the MongoDB and how this would integrate with Fireworks. I will be working from scratch to set up a test DB on the NERSC systems, but if you have any thoughts or experience on the subject, I would greatly appreciate any feedback.

Thanks,

Steve

Hi Steve,

We have not done too much large-scale testing. The initial testing we’ve done has shown bottlenecks when you are trying to achieve a throughput of more 300-400 jobs per minute:

http://onlinelibrary.wiley.com/doi/10.1002/cpe.3505/abstract

Unfortunately I haven’t explored sharding as a solution so don’t have much guidance, but would be very interested in the result. It might be something we could get NERSC to help with.

Best,

Anubhav

···

On Wed, Jul 22, 2015 at 11:43 AM, [email protected] wrote:

Hey Anubhav,

I am continuing to run tests on the Fireworks framework as a part of the CASCADE project. Many of my tests are at large scales and I have been trying to evaluate the bottlenecks on the system.

One bottleneck I have noticed is when multiple workers attempt to access the same, single mongoDB. Seemingly, the workers, in competing to access the centralized DB, are left waiting to access the resources (e.g. Fireworks).

That said, I have considered testing MongoDB’s sharding capabilities as a way to horizontally scale the database, reducing the load on any single DB instance. I was curious if you had any familiarity regarding sharding on the MongoDB and how this would integrate with Fireworks. I will be working from scratch to set up a test DB on the NERSC systems, but if you have any thoughts or experience on the subject, I would greatly appreciate any feedback.

Thanks,

Steve

You received this message because you are subscribed to the Google Groups “fireworkflows” group.

To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].

To view this discussion on the web visit https://groups.google.com/d/msgid/fireworkflows/93e01d97-7342-43b5-9570-e5e7a765c683%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.