Making Open Work
May 8–9, 2017: Training & Tutorials
May 10–11, 2017: Conference
Austin, TX

Doubling OpenStack performance with no code changes by optimizing the Python runtime

Peter Wang (Intel)
5:05pm5:45pm Wednesday, May 10, 2017
Location: Meeting Room 9 A/B
Level: Intermediate
Average rating: ****.
(4.00, 3 ratings)

Who is this presentation for?

  • Software designers, architects, developers, planners, and managers

Prerequisite knowledge

  • Experience with software performance optimization, tuning, Python, and/or OpenStack

What you'll learn

  • Understand why optimizing Python is important
  • Explore Python JIT capabilities and potentials, as well as greater PyPy JIT adaption


Python is a popular scripting language because it’s easy to get code created quickly, and it has an increasingly rich set of open source libraries. Yet in some enterprise applications, such as OpenStack, its usage has been hindered due to poor performance. This is because in many of the operation system hosts, the hardware agnostic CPython, the standard Python interpreter, runs in pure interpreted mode, and is slow.

Optimizing the Python core language—the interpreter itself—can benefit any large application implemented in Python. OpenStack, a leading cloud-computing solution, is mostly written in Python. Peter Wang shares the technical insights for achieving the best OpenStack performance using a just-in-time (JIT) Python runtime, the PyPy JIT, including JIT deployment complexity reduction, incompatibility resolution, and performance tuning techniques. Peter explains how he achieved a more than two-fold throughput increase with a 76 percent response latency reduction for Swift, the Object Storage service in OpenStack, and significant performance improvement for other OpenStack services, including Keystone, Neutron, Nova, Cinder, and Glance.

Photo of Peter Wang

Peter Wang


Peter Wang is a software engineer in the Data Center Software Technology group within Intel’s Software and Services group, where he is currently leading a team of engineers optimizing the Python language. He has been working at various organizations within Intel for more than 21 years, involving software product development for clients and servers, manufacturing and marketing communications. For the last 10 years, Peter has been focusing on performance and power optimization.