tb32012

SafeAssign Maintenance and AWS Migration

Blog Post created by tb32012 on Mar 2, 2018
We recently announced a downtime event for the SafeAssign service on March 8th, and I'd like to take the opportunity to provide more information about the reason for this downtime to provide transparency for the Bb Community:

 

For over a decade, the SafeAssign service has been hosted in Blackboard's Managed Hosting datacenters and for most of those years the performance of the service was stable and able to support the volume of submissions coming from Blackboard Learn customers.  However, over the past several years Learn clients' interest and usage of Safeassign has increased significantly:

 


 

SafeAssign is also a very cyclical product and there are small periods of very high numbers of submissions followed by long periods of low to medium numbers of submissions.  This peak load period is difficult to accommodate without relying on modern cloud computing techniques:

 


 

The graph above represents the number of papers submitted per minute over the past two weeks as of this writing; the spikes are ~midnight EST, and the big spikes are Sunday nights leading into Monday mornings.  You can see that each day has a relative peak but Sunday nights are the highest volume days with volume trailing off towards Saturday each week.

 

To handle both the increases in volume over the years and the peak volume periods, we're moving the SafeAssign infrastructure into the Amazon Web Services platform on March 8.  Actually, we've been moving the service to AWS in pieces over the course of about nine months, and March 8 represents the final migration of the remaining components of the service.

 

Unfortunately, given the scope of the migration it is not possible to prevent some downtime for the service during this transition.  We understand the impact that downtime for the SafeAssign service has on clients and users.  We are planning this downtime on a Thursday to allow enough buffer prior to the expected peak of Sunday night (per the graph above) should any problems arise during the migration.

 

The move to AWS should represent one of the final full-downtime events that SafeAssign should ever experience.  Aside from the entire AWS infrastructure we rely on going down (which can happen, though is quite rare), the risk of SafeAssign outages should be significantly reduced by this transition.  The AWS infrastructure also provides opportunity to implement automated scaling policies which will reduce the possibility of delays during peak load periods.  Finally, our architecture design within AWS will support additional improvements to things like matching algorithms, search data sources, and other areas which will improve the overall quality of the SafeAssign service.

Outcomes