is happily being the wheel rather than a rusty old spoke
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
I haven't dealt with a memory management program since the old QEMM (can't believe I remember that name!!!) and briefly a scam one put out by Synchronous Softcorp.
Why on earth would you want a memory manager in this day/age of Pentiums and tons of cheap memory?
I personally don't use ANYTHING that speeds up my computer, especially anything that works with the browser. Those tend to work by pre-fetching pages, which often is not desirable.
Especially in the case of an admin or moderator who has "delete" links available to them. Or worse.
I tried one briefly years ago when I was working as the Site Admin at SI, and poor Brad had to spend a lot of time undeleting posts my "pre-fetcher" had deleted.
Was the 19th the last hop? If not, what was the IP address of it?
Instead of doing Start, then Run, then the tracert command, type CMD in Start.Run box and it'll open a DOS window.
19 hops for me and there were some slow ones and some outright bad ones along the way. That was on the first try. Second try was 18 hops and the slowest one was 67ms.
Anyone who's seeing slowness and knows how to do this, could you do me a favor and type the following at a DOS prompt and PM me the output?
TRACERT 207.90.219.90
Check it now. You're back on the list! <g>
The current day's posting volume is a BIG part of the formula. How come you guys aren't posting much today? My watch list looks like an explosion in a green factory!
I agree. It's been absolutely screaming for me all day. Though I've still been doing some tweaking intra-day along with rewriting all the posting and reading routines. It's really time to redo them from scratch instead of patching the old stuff. I expect to have the new ones ready to rock the week after next.
I did find a frequently-used index/query pair that weren't quite right, fixed it, and the db CPU's look like they're getting bored!
Fixing the index and query spiked the CPU's for about a minute, so there were probably some timeouts and slow responses about half an hour ago, but it was well worth it. This is the smoothest histogram I can EVER remember seeing on the db. Previously, it was constantly jumping back and forth between about 35% and 80% utilization, but now it's staying in the 25-40% range for the most part and hasn't exceeded 50% even once since the index/query change.
Looks like the change made life a tiny bit easier for the webserver, too. It's nearly a straight line at 27%.
If people are seeing slowness at this point, it's got to be a bandwidth issue. The CPU's are as happy as can be!
test
What Matt means to say is that the message_id is forced into uniqueness because it's assigned by the db itself, but the board/intra-board message number combo isn't. It *should* end up unique, because the system grabs the last message number for that board and increments the current message number by 1, but apparently when the db gets bogged down, it's possible for a lot of messages for the same board to hit the db at the exact millisecond and all get assigned the same intra-board message number.
Edit: Yikes!
The message-posting stuff is legacy code and the sql string is built in ASP rather than using a stored procedure like I do on SI (where I don't think I've ever had this problem).
The line that actually inserts the message is about 20 lines down from where it determines what the new intra-board message should be. So it's easy to see how a busy db box could cause multiple posts to have the same intra-board number.
This one's going to be tough to fix because everything having to do with posting just is. But I'm going to take a shot at it and see if I can make this a stored proc so it'll run a LOT faster and have far less opportunity to get dupe numbers.
First, I need to make that stop happening. Even when the machine gets swamped like it was this morning.
I scheduled the "catch up" population of Private Message search indexing to fire off at 4 this morning. And every morning. It was still running. I stopped it and would guess the site's running better, since the CPU utilization has come back down to an acceptable level.
I'm leaving it off until I get a handle on what's going on with the index-induced loading. I suspect that's not even a minor inconvenience since I know I personally don't use PM-Search unless I'm looking for a fairly old message.
I'll look around to see if I've already written a renumbering routine. Would think I'd want to walk through the system in an ASP and do lots of pausing rather than writing the SQL Query From Hell to do it and swamp the machine for hours in the process.
Once that's found or written, I'll just have it process the day's posts every night, while I work on making it not happen in the first place.
Edit: I rescheduled the PM-Search incremental build to 1:00 AM and will let it fire off tonight since I can stop it in the morning if need be. And will get up earlier tomorrow as well as showing Matt and Dave how to stop it if they need to.
Wow! Site's screaming right now!
I just changed the board read tracking to count reads done by any logged-in account rather than only premium members.
I backed it off to premium members earlier to save some system loading and to circumvent some extremely rampant manipulation of the numbers.
Looks like we can handle the extra system loading now, and since reads that're tracked for Hot Boards lists now have a userid involved, the manipulation should stop. And/or we'll be able to easily identify and deal with the culprits.
I'm having sporadic slowness problems but have just noticed that it's a local bandwidth issue.
Okay, I'm scratching my head a bit now. Histograms look really good on both boxes! Performance should be blistering, to say the least!
It's not. From what I'm seeing.
Edit: That particular post went through amazingly fast, though.
Edit 2: As did the edit.
How's it doing now?
test
test
test
test again
And one more
and an udder
and anudder
anudder test
Looks like we're back on the air and the histograms are looking good for now.
Gonna restart the full-text search indexing and it'll probably take it all night to get caught back up.
Still? I could see it having the potential to do that for a while this morning, but not right now.
There's a sawtooth pattern to the CPU utilization histograms that I don't particularly like seeing, but they're not spiking high enough that they'd typically result in page requests hanging.
We're planning a reboot a while after the close, though, which should at least level things out. It's been quite a while since both of these machines have been booted.
No ideas occur to me right now. That board is loading fine for me.
<groan>
I'm guessing that'll be a VERY expensive car.
I'd love to have an all-electric car that's good enough (power, handling, braking) to silently peel paint at the track. This one's probably not it and prohibitively expensive to boot.
What I want probably won't exist for quite a while.
I'm looking into home-brews for increasing the Camry's economy, but I'm not keen on voiding the warranty on it or spending $10k while doing so.
Not terribly.
Database size is a relatively small portion of the overall performance equation. And PM's are a relatively small part of that portion.
The issue is the number of transactions we're trying to accomplish with a finite amount of processor horsepower and even hard-drive bandwidth/response time.
And now that I've finally gotten the machine to stop trying to full-text index for now and seen the CPU utilization plummet to a far more relaxed level (as in 25-40% vs 85-100%), the main culprit seems to be pretty clearly identified.
While it was trying to catch up on 2 days worth of backlog yesterday, I also noticed a kind of randomness to it. It definitely wasn't walking through the posts in chronological order.
The inner workings of Full-Text Search in SQL Server 2000 are not only a major mystery, they're also not very good and have been dramatically improved in SQL Server 2005, which we're planning to upgrade to in the next 30-60 days. Along with a more powerful box. And even once the box arrives, it'll take a while to move over to it. That's a pretty huge project. Just moving the data over will take a while and we'll have to shut down the site at some point to finish the data move. Of course, on a weekend.
Looks like we were getting some timeouts for a few minutes when submitting posts. The server is so busy it gets extra-swamped with seemingly innocuous instructions like "Stop populating the search index for now."
It looks like it's still trying to update search and hasn't yet seen/obeyed the order to stop. But the processor utilization has dropped enough to make it less likely to throw errors when posting.
I'll hold off until after the close before telling it again to get the search index caught up.
Edit: If you get one of those timeouts, you should be able to hit the Back button on your browser, and hit Submit again. People doing that (and failing a time or two) is why some messages are showing reply counts greater than the real number of replies. The reply-incrementing part is apparently happening despite the reply not "taking".
Working on it. It should be caught up tomorrow. The current setup will rarely be realtime (hasn't been for months) but should get/stay fairly close after tomorrow.
I think the lag has typically been about half an hour at worst when it's trying to stay caught up realtime. It'll get about half an hour behind, then suddenly get caught up, then the lag starts growing again.
I'm making private message search non-realtime (not needed, I'd think) in the hopes that'll free up enough processor power to keep realtime servicing of full-text indexing available without putting too much load on the database server.
Probably was a provider to one of our lower-tier ad networks.
If someone is on your Ignore list or you're on theirs, they can't send you a PM, regardless of the state of the on/off toggle.
Hey, my auto refresh isn't working..... Do I need to go out and get another computer? :)
Yep.
Make it a Dell 2850 with a pair of 3.6Ghz Xeon's, a decent-sized array of 15k-rpm SCSI drives, 4 gig of RAM, and W2K3 Server and SQL Server 2K5 pre-installed.
Need my shipping address? :)
Edit: Message 12212111, eh? Keep trying to find a grub in there, but not seeing it.
I rode my daughter's 1975 Suzuki GT185 to her softball game last night when I got back from the track, and couldn't help but think about how you describe their sound. It's particularly true of that bike, especially since it's got a baffle in need of repacking, but I've as yet been completely unable to remove it. It's particularly pathetic because I'm so heavy for such a small bike and it's not making the kind of compression it should be, so it has to work extremely hard to build up any kind of speed.
But a big two-stroke like the H2 or the Water Buffalo has a substantially different sound. Brash and abrasive, but with the bass undertones that you can only get from displacement.
I didn't much care for the aesthetics when I first saw it in the driveway, but it's since gotten more attractive.
Drove it to the office and back last night, which I think is about a 12-mile loop, most of it highway.
Averaged 82.5 mpg on that trip. It was quite easy to figure out the methodology on this car (thanks to the graphic display of the engine/motor-genny/battery/wheels interaction) and not only did I squeeze amazing economy out of it, the display showed just as much battery charge when I got home as it did when I left.
What I think I figured out pretty quickly is that the "Do I need to start the gas engine?" question is both load and throttle-position sensitive, with throttle-position being more heavily weighted. I think. It wasn't a very thorough drive.
I was able to get it up to pretty decent speed without the gas engine starting and will experiment to see what the max is.
A little odd not hearing the gas engine and the car being completely silent at stops. When rolling, the only way I know the gas engine is running is the mpg gauge and the Navi display.
It's also a VERY cute trick what this car does about keys. There isn't an ignition key and I don't know yet if there's one for the doors or trunk. Didn't think to look. There wasn't a key on the fob my wife gave me last night.
There's simply a fob. If it's in your posession, the doors will unlock when you lift the handle. And the "Power" button in the car will work. It's really neat how you can leave the car and it's quite secure, but it "recognizes" you and enables itself when you get near it with the fob. Oh, and apparently it's supposed to be quite difficult/impossible to lock yourself out of the car. According to the wife. I'll need to read up on the details.
A surprisingly decent sound-system, too!
What'd be a pretty cool addition would be a LOT more battery power, more reliance on the electric motor when there's a lot of juice in the batteries, and the ability to plug it in to recharge the batteries. Despite solar panels still costing too much to be financially feasible, it'd be pretty cool to have a few hundred watts worth of them and know that x number of miles in the car were totally free and green.
Oh, and before I reset the mpg display, I floored it and was VERY pleasantly surprised. This car will definitely get out of its own way! The electric motor is very torquey! Very unlike my daughter's Civic.
It's back and I've changed the default and shortest allowable interval to 60 seconds.
Public message-searching should be up-to-date shortly (sometime today) and getting itself caught up on an hourly basis from here on out. I'll do the same with Private messages once this one's done running.
Also, I admit that I don't know why CHDT isn't on the most read list. It's on one (home page), but not the official 'Most Read' list. A question is in to Bob (geek) about this.
My guess would be because of coding changes I made Thursday to reduce the workload on the database server and to circumvent some pretty egregious manipulation that was being observed in the log files. No idea which boards were involved. But one example was someone running a script that did over 100k reads of two messages from two different boards over and over again in rapid succession.
One of the changes I made was having the system only count reads from premium members. Another change is that every night at midnight, the boards with reads recorded but zero posts for the previous day get the previous day's tracking record removed. This is actually several thousand boards.
I'll consider having the system count free-member reads once I have some more horsepower to work with, but for now it's too high cost to track all of the millions of daily reads.
Best lap time today 1:53:047 in the new STi with the crappy Potenza's and OEM pads on it.
I typically did high 1:56's in the WRX.
A turn I used to consistently exit at 95 mph right at the cone in the old WRX, I was immediately doing 100, then eventually 105 today. In some places, the gearing in the car is just right, in other places, unfortunate. In the turn where I'm now doing 105 (T4 at MAM), I'm just starting to bounce the rev limiter right at the cone. If I get any faster (and I plan to), I'll have to take that turn in 5th.
Tires and brakes should definitely get me into upper 1:51's (faster than the Mustang), and I don't think it's going to be very difficult to make this about a 1:46 car.
For the most part, it handles as one would expect. Take the WRX's bad traits and throw a lot more speed at them. Except this one alternates between extreme understeer and extreme oversteer mid-turn. I need to find a way to disable that computer control of the center diff. I prefer to know which extreme to expect at any time. This random stuff is pretty tricky.
Overall, I'm VERY happy with this car and Sheridan just might have to step his game up a notch once I've got tires and brakes.
Wife called and she picked up her Camry Hybrid yesterday and really loves it. No idea what kind of fuel economy she's getting yet. I'm looking forward to seeing what's the best I can do.
Best lap time today 1:53:047 in the new STi with the crappy Potenza's and OEM pads on it.
I typically did high 1:56's in the WRX.
A turn I used to consistently exit at 95 mph right at the cone in the old WRX, I was immediately doing 100, then eventually 105 today. In some places, the gearing in the car is just right, in other places, unfortunate. In the turn where I'm now doing 105 (T4 at MAM), I'm just starting to bounce the rev limiter right at the cone. If I get any faster (and I plan to), I'll have to take that turn in 5th.
Tires and brakes should definitely get me into upper 1:51's (faster than the Mustang), and I don't think it's going to be very difficult to make this about a 1:46 car.
For the most part, it handles as one would expect. Take the WRX's bad traits and throw a lot more speed at them. Except this one alternates between extreme understeer and extreme oversteer mid-turn. I need to find a way to disable that computer control of the center diff. I prefer to know which extreme to expect at any time. This random stuff is pretty tricky.