John R. Galloway, Jr.
795 Beaver Creek Way
San Jose, CA. 95133
(408) 250-1038
2010
Ingres,
Performance Architect 2/07 –
Ingres
Corporation was spun out of Computer Associates in March of 2005 (after being
bought by ASK years earlier). The
technology has its origins from the University of California at Berkeley as the
first (and open source!) relational database released some 30 years ago. A dedicated performance group was
formed to determine the productÕs competitiveness and help guide its
improvement in all areas of performance.
Projects are under way in TPC benchmarking, testing infrastructure, and
subsystem prototypes. Since 2008
most of the work has centered on VectorWise a new columnar based vector-processing
engine that has been integrated into the product. I have also been the chair of the TPC-H technical
subcommittee since 2/10.
PalmSource,
Performance Lead 1/06 - 1/07
Palmsource
is the software division responsible for the PalmOS spun out of Palm, then
bought by Access (Japan) and now with groups (from acquisitions) in France and
China. As announced at 3GSM in
March, Access is building a new system for handheld devices and phones
providing a PalmOS like user experience supporting PDAs and smartphones but
using a common Linux/Open-source base plus custom middleware. I was hired as Performance Lead, but
changing company plans required my assistance on more general development and I
ended up designing and implementing volume management (bridging the PalmOS
world with Linux HAL/udev) as well as contributing to global settings mgt,
power mgt, and a host of other odds and ends, while also keeping an eye on
performance.
Toasterfish,
Founder 11/03 – 11/05
Toasterfish
was a self-funded garage startup that I did not manage to really get
launched. The idea was to bring a
set of high level robotic modules and infrastructure components (power mgt,
drives, processors, etc) to allow developers to focus on the software that is
so badly needed to achieve the robotics revolution. Robotics is really in the same state as the early 80s computer
industry offering only parts and vertical systems. What is needed is the same explosion of software development
that open systems and low cost hardware made possible with mini-computers &
unix and PCs & windows. Alas a lack of funding has allowed this opportunity
to slip by. (http://www.toasterfish.com).
Zambeel,
Senior Principal Engineer, Fremont CA 9/01 – 4/03 - then as Storad – 703
Zambeel
(now Agami) developed a fully autonomous policy based cluster NAS built from
commodity components and Linux. This system offered the first provisionable (ops/s, MB/s, and
capacity) NAS product using a highly scalable fully fault tolerant
architecture. I lead the group
that designed and implemented the Network Data Management Protocol
(backup/recovery), which had to not only achieve high performance but
adjustable performance in keeping with the policy based management of the
overall product. Before during and
after NDMP I also made major contributions to network time protocol support,
cluster coordination, messaging, disk failure analysis, and customer
support. I walked away when the
company was intentionally imploded to shed debt and reincarnated as
Storad->Agami.
Independent Consultant – San Jose,
CA 11/87 – 9/99
A contract to rework the bottom half of the
Informix database engine to exploit the Sequent SMP architecture started a
unique relationship as a consultant for Informix core development. During this period major work was done
on many versions of single and multi node Informix with a focus on storage
management, logging, performance, scalability, and concurrency. The core I/O, scheduling, and OS layers
were the primary focus, but significant work was also done on utility
applications and their interaction with the database. Work for other groups was done as well including porting,
benchmarks (TPC-B/C/D and customer specific) QA, and on-site support (for the
Municipal Court of Seattle, the largest Informix installation of its day).
The contract nature of the work with
Informix also allowed OEM relationships to form and contracts were done for
MIPS, Siemens, Sequent, Data General and others typically focusing on
performance investigations and enhancements usually to exploit the features of
a newly released model.
During the last four years of this period
I formed a two-person company (Database Engines) to continue essentially the
same work I had been doing but with a somewhat broader scope. Work with the
major DMBS vendors and their OEMs continued including Informix, Oracle,
Sequent, Sun, and Tandem. Contract
content was expanded to include CORBA infrastructure and application server
technology (at Oracle).
Opus Systems - Cupertino, CA 9/86 - 10/87
Opus
built Unix-coprocessor boards that plug into PCs, turning the PC into a Unix
workstation. I was responsible for
the graphics package which involved porting new versions of the X Window system
to the Opus dual processor architecture (PC plus coprocessor). The system was successful in small
niche markets requiring the (then) higher performance CPUs, but lost favor once
PCs caught up with the workstation vendors.
AIT – Santa Clara, CA 10/85 - 4/86
American
Information Technology was founded to build a computer family based on a
proprietary 64 bit microprocessor.
I was part of a team that designed a micro-kernel dealing with memory,
processor, and process management as a service layer under several operating
systems (initially Unix). Several developers, including myself, left less than
a year after joining when we realized that the founders were not capable of
creating a corporate structure to match the technical one we were chartered
with delivering
Independent Consultant & NASA
SBIR - San Jose, CA 8/83 - 9/85
During this period I did a major project
for GE-Calma in addition to a small project for Synapse and took an extended
vacation in Nepal. The Calma project was to redefine and then implement (a port
of) their proprietary user level network package on Data General systems. Binary compatibility was maintained
with both old (16 bit) and new (32 bit) users concurrently on the same
system. The system worked well and
was implemented on several platforms.
In 1984 I received a $50k NASA SBIR grant
for research into a new file system architecture (what today would be termed a
disk level object store). The prototype work went well and verified excellent
performance and reliability gains, but required too much memory for current
DRAM prices and required a significant departure from SCSI, which did not seem
achievable in 1984.
Synapse Computer – Milpitas,
CA 7/81 - 7/83
Synapse Computer was
founded to build and market systems developed specifically for the OLTP
market. The system pioneered a
microprocessor based cache coherent symmetric shared memory design very much
like today's Sequent Symmetry series.
I was a major contributor to the overall architecture and designed and
implemented the kernel file system and associated memory mapped file I/O
system. The system was particularly innovative in its SMP hardware and layered
software. The later which placed a
RDBMS below the user visible OS
layer. Thus the kernel file system
was essentially the I/O subsystem of the database. An initial revision of the system was completed and performed
extremely well but targeting mission critical OLTP as our first market proved
to be a poor choice and the company never gained sufficient traction to sustain
itself.
Data General – Westboro MA 11/75 - 6/81
At DG I worked on 3
operating systems and a new architecture. I started by writing device drivers
but quickly became one of the senior members of a group developing the Advanced
Operating System. I was primarily responsible for all aspects of I/O
(controller interface definition, file systems, user libraries, etc.).. After AOS I joined a team developing a
new 32 bit architecture & OS (a precursor to the eventual MV line) and
completed my five years at DGC as the technical lead of a group doing a new
realtime OS for a new line of microprocessors.
Case Institute of Technology –
Cleveland OH 9/71 - 8/75
At Case I declined to
follow the conventional Computer Engineering curriculum. Instead I entered the
Undergraduate Scholars program which essentially allowed me to complete a
masters level of courses and projects in about a 60/40 HW/SW mix while getting
only a BS in return (it seemed like a good idea at the time!). Then of course I went into software not
hardware.