veyepar: Hardware and Software Architecture of The Machine

Author(s):	Keith Packard
Location	Costa Hall
Date	feb Thu 04	Days Raw Files
Start	11:35	First Raw Start	11:34
Duration	0:45:00	Offset	0:00:18
End	12:20	Last Raw End	12:20
Chapters	00:00
Total cuts_time	45 min.

https://linux.conf.au/schedule/30060/view_talk
raw-playlist raw-mp4-playlist encoded-files-playlist host archive tweet mp4 svg png

assets release.pdf Hardware_and_Software_Architecture_of_The_Machine.json
logs
Admin: episode episode list cut list raw files day marks day marks day image_files

State:

Locked:

clear this to unlock

Locked by:

user/process that locked.

Start:

initially scheduled time from master, adjusted to match reality

Duration:

length in hh:mm:ss

Name:

Video Title (shows in video search results)

Emails:

email(s) of the presenter(s)

Released:

has someone authorised pubication

Normalise:

Channelcopy:

m=mono, 01=copy left to right, 10=right to left, 00=ignore.

Thumbnail:

filename.png

Description:

markdown

Innovations in optical interconnect and storage technologies have
prompted HP to undertake a large-scale development effort called "The
Machine". This radical departure from traditional computer
architectures which we call 'Memory-Driven Computing', where large
scales of heterogenous computation can be brought to bear on a data
set which is accessed not over a traditional network or storage
system, but rather directly from the computational elements using
memory load/store instructions.

This presentation will describe the first implementation of this
architecture, both hardware and software. On the hardware side, this
will include:

• Heterogenous computation node hardware
 • Memory interconnection fabric
 • Hundreds of terabytes of memory
 • Network interconnect
 • Top of Rack Management Server

The system software for The Machine is based on Debian GNU/Linux.  The
presentation will focus on new and updated software elements while
mentioning some of the existing code that has been particularly useful:

• Two-level storage allocation system
 • DAX changes needed to support large storage systems
 • Concurrent file system designed for large shared memories
 • Provisioning services
 • The Machine Environment Simulator (TMES)

As with any research project, The Machine asks more questions than it
answers.

• How should we build data structures for storage where the
   addressable persistent unit is a single byte?

• What changes are needed in the design of large-data software
   systems to best take advantage of this shift from network-driven
   computing to memory-driven computing?

• The Linux operating system model drives a huge range of hardware
   these days. Is The Machine different enough to encourage a
   different model?

Comment:

production notes