Applying The Universal Scalability Law to Distributed Systems
Neil J. Gunther
Performance Dynamics, Castro Valley, California, USA
Abstract
When I originally developed the
Universal Scalability Law (USL),
it was in the context of tightly-coupled Unix multiprocessors, which
led to an inherent dependency between the serial contention term and the data consistency term in the
USL, i.e., no contention, no coherency penalty [1,2].
A decade later, I realized that the USL would have broader applicability to distributed clusters
if this dependency was removed [3].
In this talk I will show examples of how the
most recent version of the
USL (with three parameters α, β, γ) can be applied as a statistical regression
model to a variety of large-scale distributed systems, such as
Hadoop [4], Zookeeper, Sirius, AWS cloud, and Avalanche DLT,
in order to quantify their scalability in terms of numerical concurrency, contention,
and coherency values.
References
- [1]
-
N. J. Gunther, "A Simple Capacity Model of Massively Parallel Transaction Systems,"
CMG Conference, San Diego (1993)
- [2]
-
N. J. Gunther, The Practical Performance Analyst, McGraw-Hill (1998)
- [3]
-
N. J. Gunther, Guerrilla Capacity Planning, Springer (2007)
- [4]
-
N. J. Gunther, P. Puglia and K. Tomasette,
"Hadoop Superlinear Scalability: The perpetual motion of parallel performance,"
Communications of the ACM,
Vol. 58 No. 4, Pages 46-55 (2015)
DSConf 2019
Distributed Systems Conference
Pune, India
11:00 am IST, Saturday February 16, 2019
09:30 pm PST, Friday February 15, 2019
05:30 am UTC, Saturday February 16, 2019
File translated from
TEX
by
TTH,
version 3.81.
On 31 Dec 2018, 11:51.