Monday, September 30, 2013

Third Annual Solaris Family Reunion (Part 2)

Deirdre at Joyent helped to organize The Third Annual Solaris Family Reunion sponsored by Joyent on September 23, 2013.Attendees ranged from former employees at Sun to current employees at Oracle. Each discoursed a short 10 minute presentation as to what they have been doing over the past 3 years. This includes sessions 7 through 11.

Session 7: History at Kevin Zimmerman
Worked at University California at Berkeley.
Deployment of workstations, servers, and system administration teams.
Federated Naming, Storage on Suns, instructional mission on Suns.
Costs of implementing Windows NT was extremely high.
Exchange, Active Directory implementing. Unlimited licensing.
Core ran under Solaris, Linux became more popular.
Oracle was slow to form policy, OpenSolaris was 1/10th the cost of NetApp, Oracle could not provide good support, CIFS code was removed from Solaris 10, Nexenta became central storage partner.
By 2004, undergraduates with mid-range systems was evaporating: web based tools & social networking tools were popular.
Presentations no longer made to university by Oracle, but only to State of California.
Famous quote, "Sun opened computing landscape."
Innovations included: NFS, DTrace, Zones, true client-server, integrated naming, integrated large scale deployment, ZFS, modifications to the network stack..
Since Oracle consumed Sun, there is a much larger community or smaller independent companies.

Session 8: Sunay Tripathi at Pluribus Network
Crossbow and FireEngine
Left on the merger day of Oracle and Sun.
Trying to build server switches on Linux. The calls taken during Beta were just to keep it alive
Famous quote, "Solaris...a rock solid piece of code, it needed to stay alive" and "we started with Linux,  and getting that damn thing to stay alive on a switch, was ***, I mean the cause of the cause... just to keep the damn thing alive."
After moving to OpenSolaris, the OS now just stays alive, the issues are somewhere else.

Session 9: Keith Wesolowski who moved from Sun, to Riverbed, to Joyent
Feb 17 2011, 1 year after change and control, Oracle ejected Keith.
Worked at Riverbed, spent 6 months on WhiteWater product
All Linux and all C++ environment.
Spend first few weeks learning the ropes, when "I realized I don't have the tooling" and the platform "had a lot of performance problems."
Started porting Riverbed software to OpenSolaris to figure out problems with the Riverbed software, but unfortunately, the problems did not exist under OpenSolaris. Could not get Linux SystemTap to even work, except to panic the box, and that did not even create a crash dump under GNU Linux.
The RAID subsystem was unsafe, with the RAID 5 write hole. Customers sent systems back with rampant data corruption.
Famous quote, "after 2 days of basically accomplishing absolutely *** nothing... I sent the suicide note to Brendan saying, man is there anybody hiring someone for something that does not suck?"
Now integrating where hardware and software meet.
(not mentioned: Riverbed and Joyent partnership)

Session 10: Max Bruning now at Joyent
Trained staff all over the world in Sun.
Started porting Linux KVM to SmartOS, Bryan and Robert continued the effort.
Provides training: DTrace, MDB, NodeJS, ZFS Internals.

Session 11: Robert Mustacchi from Sun to Joyent.
Worked on the Fishworks and KT Processor.
Completed the work at Joyent with the Linux KVM Port to SmartOS.
Works all over the stack, enhanced MDB, random hardware work, DTrace enhancements, working networking for virtual machines, and revamped libumem (enhanced small repetitive allocations & frees in the same thread.)

No comments:

Post a Comment