Grouper Users - Open Discussion List

[Tim Darby <>]

I agree with Chris. We found that the latency between AWS and on-prem was killing Grouper performance. It worked well when we put it all in the cloud. Also, I'll put in a plug for MariaDB RDS, which works great with Grouper.

Tim Darby

UA Identity Services | The University of Arizona |  | +1 520.626.3799

On Tue, Apr 17, 2018 at 10:01 PM Hyzer, Chris <> wrote:

I would run the JVM next to the database (either both on prem or both in cloud)

 

Thanks

Chris

 

 

From: [mailto:] On Behalf Of Greg Haverkamp
Sent: Wednesday, April 18, 2018 12:47 AM
To:
Subject: [grouper-users] Anyone using cloud RDBMS with on-prem Grouper?

 

My deployment progress has been literally slowed, and I'm trying to figure out if there's something I'm just doing wrong.

 

I've generally had pretty good luck with AWS RDS (though with Postgres), and we run a couple of services off of it.  Our plan was to deploy the Grouper DB to AWS RDS (MySQL) while running Grouper on-prem, at least until we work out our desired strategy for pushing everything out.  

 

However, we've found Grouper to be ridiculously slow in this model.  Slow enough that it makes it difficult to try to get anything done with it.

 

Today I spent the day poking, prodding, resizing, creating new resources, etc., all which seem to indicate that perhaps the Grouper DB is not well-suited to even minimal latency.  I'm wondering if anyone else has attempted this and found otherwise?

 

At this point, I've tried AWS RDS MySQL (5.7).  With my load, gsh takes about a minute to display its prompt.  This seems to be the case regardless of RDS instance size (bumped to get "Network Performance" from "Medium" to "High", General SSD storage size (bumped it up to 100 GB to try to get the performance wins), etc.  I also converted to Amazon Aurora, in an attempt to see if it could be that.  However, performance remained basically the across all attempts.  Around 1 minute to get to the gsh prompt and around 1-2 seconds to findSubject().

 

Then, I decided to try a quick benchmark.  I cranked up a MySQL 5.7 Docker on my desktop, loaded up my data, and pointed my Grouper instance at it.  gsh appeared in around 10 seconds, and the findSubject() returned almost immediately.

 

Deciding I'd give one last shot at not having to run my own database, I cranked up MySQL at Google and loaded the data.  It returns faster, but only about 20% faster than AWS.  (It's perhaps not coincidental that my ICMP RTTs around about 20% faster to Google than to AWS.)

 

I'm guessing this is likely in the nature of Grouper, and that perhaps it's just running so many queries that 20 ms per query starts to add up.  Does that seem about right?

 

Greg