Beefy Boxes and Bandwidth Generously Provided by pair Networks
go ahead... be a heretic

Disappointed with perlmonks' speed (discussion)

by deprecated (Priest)
on Dec 17, 2001 at 20:31 UTC ( #132552=monkdiscuss: print w/replies, xml ) Need Help??

As somebody who is blessed with large bandwidth, I am forced to wonder what sort of connection perlmonks has and what sort of hardware it runs on, as well as the hits and througput to and from the site.

Speakeasy offers 1.1mbit sdsl for $200/mo to where I live (which I get, and its more than adequate for a large napster server). I find that between work and home, I am not able to get to perlmonks without a 10-30 second lag in requests. I find that sometimes stopping the request and re-starting it helps to get through. converter mentioned that he has seen the site get faster since the ad was taken away -- I havent noticed this at all.

I have donated in the past, and plan to donate again after my finances re-settle (this terrorism thing has really affected my employer). What can I do to increase the responsiveness of the site? We're getting more than $200/mo, and an additional 1.1mbit is something like 140kbyte/s. How big is the average perlmonks website?

Can I contribute database code review skills? I mean, I hesitate to suggest something as simple as adding indexes (how does our database look)? I found that for large-scale databases that Postgres was faster for me than mysql, although admittedly, I have tuned postgres about as far as it will go on our ultra 10's.

I'd love to help, as it seems to me the monastery needs some help.

brother dep

Laziness, Impatience, Hubris, and Generosity.

  • Comment on Disappointed with perlmonks' speed (discussion)

Replies are listed 'Best First'.
Re: Disappointed with perlmonks' speed (discussion)
by Masem (Monsignor) on Dec 17, 2001 at 23:11 UTC
    I have a feeling it's something upstream of PM. I've seen the same problems from two different access points with PM, but the symptoms make little sense for it to be a problem with CPU cycles on the local hardware. Instead, I think that a few hops upstream may be flaky, as it has more of a feel to dropped packets than it does to overloading a server.

    With the ads removed, that would help the response time (presume the ads were on the same network segment but different server), but wouldn't completely remove it.

    And the lost packets are not the same for everyone; I've watched CB chatter go by, and when one of the temporary speed slowdowns hit and disappate, the amount of CB chatter feels like there's been a lot more going on while I've missed it. (Also, seeing jcwren's Robomonk in action, there's been points where after Robomonk resees the site that a large influx of messages appear).

    So unless I'm mistaken, I doubt there's little that can be done to fix it at vroom's end, even if more money was through at it. It's probably not even the ISP, most likely it's the next level up. But I haven't sat there with traceroutes or similar tools to isolate it as such.

    Dr. Michael K. Neylon - || "You've left the lens cap of your mind on again, Pinky" - The Brain
    "I can see my house from here!"
    It's not what you know, but knowing how to find it if you don't know that's important

      You're probably right, however one thing that I've noticed is that the big delay usually comes when I log in. After that it's generally useable. I often have to Stop the page during login after it spins for a minute or two, and hit the login button again. Sometimes when I then attempt to login again, it will immediately log in, but then other times it will take several attempts (maybe 5 minutes total) to get in.

      Here is a traceroute taken when I had trouble logging in (seems fine, no hangups):

      $ /usr/sbin/traceroute traceroute to (, 30 hops max, 38 byte pack +ets 1 ( 0.545 ms 0.465 ms 0.436 ms 2 ( 4.841 ms 2.600 ms 2.307 ms 3 ( 24.117 ms + 3.163 ms 5.019 ms 4 ( 3.088 ms 16.287 m +s 30.606 ms 5 ( 3.203 ms + 3.096 ms 11.059 ms 6 ( 3.256 ms + 3.082 ms 3.299 ms 7 ( 3.554 ms 3.477 ms 3. +539 ms 8 ( 3.538 ms 3.404 ms 3. +413 ms 9 ( 8.633 ms 8.595 ms 8.6 +31 ms 10 ( 8.912 ms 16.037 ms 9 +.603 ms 11 ( 89.949 ms 70.057 ms +70.201 ms 12 ( 71.379 ms 71.413 ms +71.394 ms 13 ( 115.884 ms 74.740 ms + 78.527 ms 14 ( 76.103 ms 76.653 ms 76.478 ms 15 ( 78.098 ms 78.423 ms 77.512 ms 16 ( 74.470 ms 93.624 ms 75.2 +80 ms

      On my side, I rarely experience large page load delays. I browser permonks both at work (actually the company -- TELUS --I work for sells ADSL to local population) and from home, where I'm hooked up to a 1.5 Mbps ADSL line (provided by my company).

      Certainly, buying new hardware would never hurt. I'm wondering wherethere there's a mentioning on this site which server it runs on? In any event, more $$ will always equate to better hardware and greater performance... (I'm considering to put some of my paycheque on the offering plate to aide the process).

Re: Disappointed with perlmonks' speed (discussion)
by chromatic (Archbishop) on Dec 17, 2001 at 22:28 UTC
    I'm not seeing the lag you are. Even on a dial-up, it's pretty responsive.

    As for the database, my esteemed colleage Krow and I have discussed some enhancements in the past. He's responsible for most of the brilliant speed improvements in Slash. I suspect his best idea is to gather all long text columns into a separate table, leaving document-nodetype derived metadata in smaller tables. Unfortunately, that's not on either radar at the moment.

Re: Disappointed with perlmonks' speed (troubleshooting info, analysis)
by ybiC (Prior) on Dec 19, 2001 at 00:41 UTC
    In the interest of providing additional info for effective troubleshooting, here's what I've experienced for the past couple weeks:

    At home on 56k dial to Earthlink, Omaha, NE:   frequent intermittant delay of a minute or more in *retrieval* of pages from the Monastery.   It's not a local rendering problem because I can watch the external modem's RD/SD LEDs and see when the data is being received.   This occurs whether or not I block the banner ads with Junkbuster.

    At work through Corporate MS proxy to T1 to AT&T, Omaha, NE:   frequent intermittant timeout error from proxy.

    Whether at home or at work, the problem is not episodic.   In other words, there doesn't appear to be times when it's "broken" and times when it's "working".   At any given time roughly 1/2 of the page loads happen quickly and 1/2 are sloooow.   Is not browser-dependant for me - same with Mozilla (Linux) and IE or Opera (Win2k).

    However, framechat at home works fine, but IIRC, it works from XML feeds, not the regular HTML pages.

    Of these possible problem points, my direct observations would seem to rule out local browser and banner ads as contributing to the problem.   Reports in this thread from fellow monks who *aren't* experiencing the severe lag would seem to rule out PM host (hw | sw | webserver | E2 | bandwidth), leaving Internet connectivity problems as a strong candidate.   Except that I don't see why page retrieval would be slow and XML feed not be.

    • PM host hw
    • PM host OS
    • PM webserver
    • PM E2
    • PM bandwidth
    • Banner ad hosting
    • Internet connectivity at-large
    • local browser

    Anyone else have information that might help track down and hopefully resolve this problem?
        striving toward Perl Adept
        (it's pronounced "why-bick")

    I tried 1000 byte pings to PM and got consistant quick responses while I was experiencing intermittant slow PM page loads.   That and a CB comment by good monk tye led me to this apparant solution to the problem.

      I can just second that.

      When accessing perlmonks from several different places I get half the pages fast and the other half not at all (read timeout).
      And I have never problems with fullpage cb

Re: Disappointed with perlmonks' speed (discussion)
by count0 (Friar) on Dec 17, 2001 at 20:59 UTC
    Can I contribute database code review skills?... how does our database look)?

    Granted vroom and other monks have surely hacked at the base code, this site runs on the Everything engine. The code, database, and even the custom Perlmonks nodes (like the chatterbox, post/preview, comments, etc) can all be downloaded there. =)
(crazyinsomniac) Re: Disappointed with perlmonks' speed (discussion)
by crazyinsomniac (Prior) on Dec 18, 2001 at 07:45 UTC
    I almost always browse on a dialup (56k) and always have at least 5 perlmonks windows which I continually refresh and stuff, and response is pretty dang fast, even on my slow connection. The only thing that's ever slowed down for me are really really big nodes, because I have my node-depth at 99 and node threshold at 100000. The only other thing that's kind of slowed for me is my message inbox, but that's cause I got some 400 messages in there (and yes, I do need them all).

    Disclaimer: Don't blame. It came from inside the void

    perl -e "$q=$_;map({chr unpack qq;H*;,$_}split(q;;,q*H*));print;$q/$q;"

Re: (Page Construction also a culprit) Disappointed with perlmonks' speed (discussion)
by buzzcutbuddha (Chaplain) on Dec 18, 2001 at 22:28 UTC
    Another factor that will affect the appearance of load time is the HTML construction of the page. Depending on the speed of your machine at home vs. your machine at work and the browser you're using, you can see vastly huge differences in load time, even on the same connection.

    As point of reference, the Monastery Gates uses 14 tables, some of them nested, and at least 6 tables on most other pages.

    The more tables you have on a page (especially nested ones), the more work it takes your browser and therefore your PC to render the page. Having to use tables like isn't anyone's fault here, or even at EveryDevel. It's needed to make sure that every user coming to PerlMonks can browse the site, even if they're using Netscape 3.

    Maybe once everyone's browsing with fully CSS compliant browsers (and who knows how long that will take, if it will ever occur) we can move beyond using tables to something more DHTML-ish.
Re: Disappointed with perlmonks' speed (discussion)
by Fastolfe (Vicar) on Dec 20, 2001 at 03:54 UTC

    To add my own anecdote to this thread, I very regularly see periods where the site is completely unresponsive (read timeouts) from both at work (with a nasty HTTP proxy) and at home (straight DSL connection). Generally things work fine for 5-10 minutes, and then there will be a 1-3 minute period where I cannot get a response from the site at all. When it does respond, though, it's generally pretty speedy.

    Though I stopped trying to pursue this when I asked others if they were seeing similar problems. Until this thread, everyone I'd spoken to had stated they saw none of the problems I saw.

    Could this be an issue with an ISP slightly upstream from PM? That might explain a significant number of us seeing frequent problems while others (whose data may be routed separately) see no problems. *shrug*

Log In?

What's my password?
Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: monkdiscuss [id://132552]
Approved by root
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others meditating upon the Monastery: (3)
As of 2023-09-29 00:59 GMT
Find Nodes?
    Voting Booth?

    No recent polls found