One Minute Evaluations for 14-16 October

CS 438 - Operating Systems Concepts

• Topic: We talked about lots of interesting stuff including how systems communicate with devices.

  Question: Why is Solaris called Slowlaris? Does JavaOS use objects for devices?

  One minute response: Step into my office and I'll show you why it's called Slowlaris. For the most part, JavaOS relies on a host-resident kernel to be the interface to the hardware.

• Topic: Device management, resource management.

  Question: Do we have an assignment due at the end of the month?

  One minute response: I'll let you know.

• Topic: Resource management.

  Question: Will we get more rain?

  One minute response: Undoubtedly.

• Topic: How drivers work and facilitate IO - How drivers receive commands.

• Topic: This week we discussed devices and how they're abstracted and managed by the os. We also discussed deadlock, conditions for deadlock, and deadlock prevention.

• Topic: We talked about device management and the unused ways to avoid deadlock.

  Question: Is there a limit to the number of devices a system can have running at the same time without slowing the system down.

  One minute response: Sure, but my guess is that the physical requirements for connecting a lot of devices to the system would be a problem before the lack of CPU horsepower would.

• Topic: We spoke about deadlock and techniques to avoid them.

  Question: What does linux do since now its being used more for desktop/standalone as well as multipurpose servers.

  One minute response: Most systems deal with deadlock by reserving the last 5-10% of any resource for itself, so that there are resources available for dealing with the deadlock. Other than that, handling deadlock is pretty much ad hoc.

• Topic: Device management; devices in OS.

• Topic: Device management.

  Question: Is deadlocking common? Can we switch classrooms to something with a window?

  One minute response: It's fairly easy for the unskilled, when writing distributed computations, to drive their computations into deadlock. Because these are hard to detect and don't effect anything beyond the deadlocked computation, the os tends to ignore them (unless the deadlocked computation holds popular or important resources, in which case something, somewhere would eventually notice). Massive, system-wide lock-ups are rarer, but still happen, particularly among networked systems.

  You'll want to speak to Debbie Mellish about changing classrooms. I hope you have better luck than I had.

• Topic: Device management.

• Topic: Deadlock, livelock, device managers, resources, device driver.

  Question: Are there any times when OSs want to induce deadlock and/or Livelock?

  One minute response: None that I can think of. It would be hard to induce deadlock, and difficult to undo it once obtained. OSs have more powerful and easier to use techniques of exerting control over processes.

• Topic: Four techniques are used in what kinds of systems (examples).

  Question: Why not implemented in any [?]?

  One minute response: Because they're too expensive.

• Topic: Talked about scheduling (process and threads), devices and I/O.

  Question: Do we have a project? What OS do you use/like?

  One minute response: As soon as I can think of a project, you'll have one. I use Unix knock-offs (Solaris, FreeBSD, Debian); I like TENEX, an obscure, dead OS that ran on semi-obscure, dead hardware (The PDP-10) from an unobscure, dead company (Digital Equipment Corp).

• Question: How much wood would a woodchuck chuck, if a woodchuck could chuck wood?

  One minute response: African or European?

• Topic: Process management, deadlocking.

  Question: Why do the operating systems feel using no anti-deadlocking technique is better?

  One minute response: Anti-deadlocking techniques are a bit like security: everybody agrees it's a good thing to have but nobody wants to pay for it.

This page last modified on 16 July 2003.