urwhatu

Personal DSLAMS as a cheap broadband solution?

I've never owned or operated a DSLAM before, but is there any reason why you couldn't get two of them, two phonelines, and two DSL modems to achieve a 6Mbps backhaul for the price of two phonelines and about $3k worth of equipment? Ebay has a few that are about $1000...

update: you can't buy lines that are direct from point to point.

Ashley, maestro of telecoms and more, says:

Check out the dry pair concept, if you haven't already per your update. Basically it's what you wanted: a point to point pair of cables w/ no dial tone. Alarm companies used to use them all the time to have a direct line to monitor you house with. Cringely wrote about it back in 2001:

http://www.pbs.org/cringely/pulpit/pulpit20010823.html

and google says more, of course:

http://www.google.com/search?q=%22dry+pair%22

Bottom line, I think you're right, unless you know some one the telcos don't like you to roll your own.

Technorati Tags: Networking

Developer programs: The difference between Apple and Sun

At Apple, you can sign up for the developer's program trivially. For an individual, it's about $500 and you get two big things: prerelease software and hardware discounts. The hardware discounts are significant, about 25%, which more than makes up for the cost of the membership. In addition, they don't try to shaft you -- witness the recent upgrade for their Intel transition hardware.

OK, now enter Sun. They also have a developer program, but it offers access only to software. Do you want to release your application on an UltraSparc? Good luck. Want to optimize for the T1? Dream on.

So maybe it's a more involved thing, e.g., the Sun Partner Advantage program for ISV's. Trying to figure out whether or not you're eligible is not so transparent (at Apple, everyone with $500 qualifies -- btw, why the heck does this matter? Why wouldn't you want all comers developing on your hardware?). It's a corporate thing I guess. This looks like it might be useful to help figure out how to get access to some dev hardware - except I have to know what pricelist categories A, B, E, H, and P are. I guess Jonathan Schwartz's ode to the software stack doesn't quite translate like it does over at Apple.

So, right now, there's discussion about what to do about ISV/partnering somewhere off in a corner of the company. It is not clear whether or not we're going to do anything with that and we're going to wait-n-see on Niagara...we bought a cheap sparc box through the online store because the other process was so cumbersome. I was so excited by Niagara a while ago, but now I'm just tired.

Technorati Tags: Programming, Sun

BayLISA talk video...

The kind folks at BayLISA  posted a bunch of videos from the December 15th meeting.  I most enjoyed Steve Hand's Xen presentation (for sheer wizardry) and Ben Rockwood's OpenSolaris presentation (for his raw geek enthusiasm that reminded me of days long past when I coo'd over things like the release of the Mac IIci).  There are also some notably missing videos about DragonFly BSD  ("I saved 20 nanoseconds per context switch, hear me roar!", "Sorry, we haven't worked out the NUMA parts of the OS because we can't afford anything except the cheapest AMD X2 processor" -- dude, someone's got to get this guy a quad processor rig) and Luke Kanies' project Puppet, an alternative to cfengine.  Oh, and there's me.

Is PPC Keyword Squatting the Sincerest Form of Flattery?

I was surprised to see today that we've joined the ranks of Geico and American Blind - another company is paying to be listed when people are searching for Splunk. Cool!

Technorati Tags: Splunk

PATH_MAX

Like many things, PATH_MAX varies from system to system. Since I happen to have access to lots of different OS's here at Splunk, I thought I'd make a quick catalog of them so I don't have to keep looking it up.

Mac/FreeBSD

/usr/include/sys/syslimits.h

Linux (Debian/Fedora Core)

/usr/include/linux/limits.h

Solaris

/usr/include/limits.h

I am, however, appropriately chagrined after reading this diatribe and will be patching my code to remove all traces of PATH_MAX in short order!

Technorati Tags: Programming, Unix

Frink

Though I'm starting to get tired of learning new "trendy" programming languages, Frink (*) looks like it might just be worth learning.

I appreciate that it prevents me, as a former NASA (JPL) employee, from screwing up unit conversions, not that I had anything to do with that particular flub :-). Consider, for example, the statement "keg->case" which, in Frink, means "how many cases are there in a keg". Check out the Frink examples for a dozen more like this including movie magic, company burn rates, and calculations from "The Matrix" ("how many coppertops per person?").

(*) Full disclosure: Curiously, I worked with the real John Frink many years ago at UCLA, hi John and Lori if you're out there!).

Technorati Tags: NASA, Programming

Grub Follows Harvest into the Sunset

Tomorrow, Grub.org is finally heading off to distributed search engine heaven. Most people won't notice, a few will be very happy, and a few will be very sad. Although not surprising at all given the new priorities at Looksmart, it is always a shame to see good ideas flushed (especially the ones you've spent a lot of time on personally).

Miscellaneous:

• At least the GRUB bootloader folks won't have to worry about the name conflict any more
• Matt Cutts says something nice about Grub
• One -Ofun thing we did in one of our releases last year was to set up the Grub screensaver so that it would show an easter egg if you held the left shift key while firing it up :-).

And now for a clustering temporal search engine...

Technorati Tags: Grub, Looksmart

A Zen Approach to Software Robustness

Forwarded internally. Armando Fox and George Candea at Stanford's SWIG propose that the best way to build robust software is to make it crash-only. Looks like the idea's been around for a while, but it's new to me and I find it fascinating...

Technorati Tags: Software Architecture

What Do You Do with Your Logfiles?

Dear Lazyweb,

We've built this Splunk thing, which is essentially a clustered search engine for temporal data, usually logfiles (*). Lots and lots and lots of logfiles. Thing is, everyone manages their technical infrastructure a bit differently and it's hard to figure out what's OK and what's not in terms of how to get at the log contents. Can you clue us in as to how you manage yours at a very high level? E.g.,

• Do you roll your important logfiles to a single location? If so, how are you transporting the logfiles?
• Do you do it in batches or real-time?
• Do you just leave your logfiles on the boxes they were created on?
• How often do you roll them?
• Are there some you care about more than others?
• How many machines do you roll logfiles on?

Thanks...

* - Or other time stamped data...

Update: if you're realizing that your logfile management is more accident than design, some good resources are:

• SAGE has an excellent booklet titled Building a Logging Infrastructure by Abe Singer and Tina Bird that is only available to members.
• Tina Bird's Top Ten syslog Signs that You've Been Hacked
• The Frequently Discussed Topics page at loganalysis.org

Technorati Tags: Splunk

Writing Software Next to SBC Park

A fun thing about working across the street from SBC Park: sometimes, when you're staying late banging out some code that's tricky but rewarding, if you time it just right, about 30,000 people will cheer just as your code makes it through the full set of test cases :-).

Technorati Tags: Splunk