Letters to the Editor

Readers Speak Out: Letters to the Editor

This month: One reader needs all the information he can get on xcalls; another suggests Bill Rosenblatt re-think MUDs; meanwhile, Peter Galvin adds security audit systems to his list of topics to cover in '98

SunWorld
January  1998
[Next story]
[Table of Contents]
[Search]
Sun's Site

Mail this
article to
a friend

Performance Q & A with Adrian Cockcroft

[Read Me]http://www.sunworld.com/sunworldonline/swol-01-1998/swol-01-perf.html

Optimizing with xcalls

Adrian,

I am using mpstat to check the performance of a multiprocessor system and have found that among several fields, xcall is very important.

There are no comments on xcall (inter processor cross call) in the Solaris man page. Can you tell me about it in more detail?

As far as I know, it is a mechanism to guarantee TLB consistency among several processors and processes. To ensure this consistency, should I test TLB every time something is changed in TLB on every processor?

If so, I fear overhead will be very high and system performance not so good in the multiprocessor environment.

Tae-sung Kim

Tae-sung,

xcall occurs when one CPU interrupts another one for whatever reason. One usage is to keep TLBs in sync. Since the TLB load is done by fast trap software on Ultra, and Ultra has very fast multilevel interrupt handling, it is still efficient to have large numbers of xcalls.

Note that UltraSPARC external caches use physical memory addressing -- not virtual as in most competitors. This means that the cache contents are independent of the MMU, and there is a lot less cache invalidation. This is one reason why Sun can build a 64-CPU SMP that works efficiently, and it has a much bigger impact than the TLB xcall issue.

UltraSPARC has four (I think) nested trap levels, all with their own registers, so even if an interrupt or low-level trap is happening, an additional MMU trap can be taken to reload the TLB -- all without having to save any registers. This is another feature that is much more sophisticated than other competing designs, and makes xcalls very fast.

Adrian

Bill's Bookshelf: "Why did we build cyberspace like that?" by Bill Rosenblatt

[Read Me]http://www.sunworld.com/sunworldonline/swol-12-1997/swol-12-bookshelf.html

MUDs: not just for the socially inept

Bill,

Regarding your comment December's Bookshelf column:

"...I don't see the point of MUDs, except as escapist entertainment for the socially inept..."
I guess that you haven't had much experience with MUDs or their physical world precursors, play-by-mail gaming.

One important aspect of both of these is that they allow real-time social interaction for people who are physically challenged, either temporarily or permanently. A few years ago I had a medical condition that left me bedridden at home for 60 days. I could have spent the time watching a lot of TV or playing single-player computer games, but instead I obtained access to a student account at a nearby university and spent much of my time online, often in MUDs which were just getting started at that time. I found the social interaction very helpful in keeping my spirits up during an otherwise depressing time.

Christopher Reeve's and Stephen Hawking's lives are exceptional in that they have a lot of equipment designed to allow them to lead lives which are as normal as possible, given the nature of their disabilities. There are a large number of people who will likely never attain that lifestyle, yet are as intelligent as you or I. I know through personal experience that many of these people have found social interaction in play-by-mail gaming, and I strongly suspect that many of them are migrating to the online world.

A now-famous cartoon shows a canine sitting at a computer and telling a friend, "On the Internet, no one knows that you're a dog." In real life, on the Internet, no one knows that you're in an iron lung.

Sam Denton

Sam,

I have had some experience with ASCII-based MUDs (back when I was a Unix sysadmin and really bored). After a little while, I despaired of them and moved on to my still-current downtime mental occupation: cryptic crossword puzzles. Nevertheless, I acknowledge and respect your point -- it is a good one.

Bill

Security column: "Pete's Wicked World" by Peter Galvin

[Read Me]http://www.sunworld.com/sunworldonline/swol-01-1998/swol-01-security.html

Audits, audits, audits...how?

Peter,

I suggest that you do a column on how to set up and manage "BSM auditing" on your Solaris 2.x OS.

I've spent the better part of a day trying to learn more about this topic. CERT and other security-minded people extoll the wonders of examining audit logs, but no one tells you how to set it up.

docs.sun.com has the core answerbooks on auditing, but they warn that an improperly set up audit system can fill up a hard drive in minutes. Amazon.com has nothing.

Auditing seems to be a very powerful tool for tracking the paths of black-hat types. Yet, in my book, the subject is a blank page.

Perhaps everyone wants it to be enabled, but no one actually uses it.

Bill Burns

Bill,

Good idea. I'll add it to my list of topics to cover.

Peter

What did you think of this article?
-Very worth reading
-Worth reading
-Not worth reading
-Too long
-Just right
-Too short
-Too technical
-Just right
-Not technical enough
 
 
 
    

SunWorld
[Table of Contents]
Sun's Site
[Search]
Feedback
[Next story]
Sun's Site

[(c) Copyright  Web Publishing Inc., and IDG Communication company]

If you have technical problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-01-1998/swol-01-letters.html
Last modified: