20190416_154017.jpg

Amirhossein Mirhosseini

 

I recently received my Ph.D. in Computer Science and Engineering from the University of Michigan, advised by Professor Thomas Wenisch

 

My primary research interests lie in the broad area of computer systems and architecture, with a primary focus on datacenters and cloud computing.  My research studies have proposed cross-layer datacenter optimizations incorporating contributions in processor microarchitecture and accelerator design, up to datacenter-level scheduling and resource management. My research has encompassed advanced microarchitectural solutions and system software implementations, and benefitted from rigorous mathematical methodologies. I intend to continue my cooperative hardware-software-theory approach to address the growing complexity of emerging cloud services and edge-cloud systems.

 

My dissertation research has focused on “Datacenter Architectures for the Microservices Era”, wherein I holistically tackle the challenges presented by the microservice architecture in cloud applications across the system stack. Microservice architectures have been widely adopted at industry, as they significantly improve programmability, reliability, and scalability of cloud services. Despite widespread adoption of microservice architectures, they present non-trivial challenges to computer system designers and architects. First, due to independent deployment and scaling of individual microservices and their complex interactions, resource allocation and service quality management for the end-to-end system is extremely challenging. Moreover, microservice decomposition results in very short processing times (sometimes as short as a few microseconds), exposing the application’s performance to a slew of subtle μs-scale systems overheads that were negligible before. In my dissertation, I have particularly sought to (1) facilitate cluster-level scaling of complex microservice-based applications to ensure the end-to-end latency target is met at minimal cost, (2) provide better  queueing and scheduling mechanisms to manage the tail latency, particularly in the case of μs-scale microservices where synchronization and preemption costs are non-trivial, and (3) provide microarchitectural solutions and hardware/software co-designs to mitigate μs-scale systems overheads of microservices. My dissertation contributions have been published in top-tier venues across communities.

coe-cse-vert.png