Oughtness, Isness and the world of systems management

preview
Posted by Charles // Wed, Apr 19, 2006 8:00 PM

"Oughtness"? "Isness"? What?

We recently caught up with Anders Vinberg, Architect and one of the leaders of Microsoft's Management infrastructure team (technologies like MOM, WU, etc) and wordsmith(he makes up words to describe complex systems). Also present was Architect Raymond McCollum, who works on things like the new event log mechanism in Windows Vista. Systems management is full of really challenging problems and Anders' team has done some incredible work in this area.

Show: Going Deep

Tags: ,

Video Length: 00:00:00 Replies: 6 // Views: 6,247
  ali_raza_shaikh
 
 
  Thu, Apr 6 2006 3:43 PM
great, very good insight about systems management


  Zeo
  Channel 9 :)
 
  Thu, Apr 6 2006 4:51 PM
Wow! Going Deep sure rocks! isness and oughtness....

  staceyw
  Bouncin'
 
  Fri, Apr 7 2006 2:58 PM
Nice.  Makes me think this could/should refactor the meaning of what a Service is.  At a level a service is just a program.  If all programs now needed to support the notion or Stop/Pause/Start, then all programs could be Services with no special install or project type.  And management could start/stop/monitor any program on the system without any special handling of Service or normal app.  So I could write a standard console app that does XYZ.  Maybe I just want to use it as a "normal" app.  But maybe I want to use the *same app and have it be a service latter.  Now because I already had to support Stop/Start/Monitor, I can just drop it into a Services Dir on the computer and it just runs as a service.  Already debuged as normal console app, don't need to compile two different versions and figure out how to deploy it, etc.  It just works.  If something behaves badly as a service, maybe the system just stops it and reports the issue and removes it from Services dir until certain quality gate is passed (manually defined or other). 

Maybe also when I log from my app, important logs flow by proxy via MS log hosting to the Developer of app (via subscription).  Developer pays small fee for this hosting proxy.  Encrypt what is secure as needed and allow certain things public that may be helpful to MS such as Exception messages, etc.  This feedback loop can only help improve the platform for all.   Naturally, when this is in place, things like Registration, license fee, etc could just flow via the same method.   Cool stuff.

  MikeSan
 
 
  Fri, Apr 7 2006 11:26 PM
Channel 9 Drinking Game. 1. Shot of tequila every time Charles says "Excellent" 2. Shotgun Can of Beer every time Scoble laughs loudly 3. One Sip of Beer every time a Microsoftie starts a sentence with the word "So"

  Charles
  Welcome Change
 
  Sun, Apr 9 2006 9:22 PM

That sounds fun, MikeSan. I'm in.

Excellent,
C



  deedubb
 
 
  Tue, Apr 11 2006 1:25 AM

First, I've been waiting ever since http://channel9.msdn.com/ShowPost.aspx?PostID=148820 to hear about some of the advancements in system management & I would like to thank Charles for getting this together...

On to the main topic: It sounds as if the windows event log system is going to be an improved version of the syslog - monitoring system, collections, generalization and standardization - will the windows event log system be able to receive syslog messages even tho they're not in XML? Will the encryption mechanism be made available so other organizations can send events? In monitoring hetro-phobic system won't work for a lot of organizations; it's why I think monitoring for a lot of companies is based on UNIX scripts and open source solutions...

I must admit I'm a little surprised by the virtualization movement, I get that resources are not being utilized so putting unalike processes on a system can be a cost savings - and a layer of separation between that is important for a number of reasons but are the hyper-visors really such a small perf hit to make this worth it when we (sysadmins) are paying so much more for the server hardware over a workstation chassis - for the small perf gains of using a special motherboard or 10k drives? Or is that the point - lowering the hardware cost - moving to a distributed virtual customer level hardware? For years I've been a supporter of fewer systems - ever since I was the tape jockey for a small company that had too many servers...