Diana the Valkyrie's Newsletter - July 2004
A hard man is good to beat
Galleries added this month.
Stories added this month.
Nothing new
Movies added this month.
A new record; May 2 brought in 164297 pictures in 19.8 gigabytes.
"At Her Feet", a new Kasie Cavaough video, for foot lovers.
As I feared - June 8 was nice and warm, so the air conditioning had to work harder, which overloaded the power supply to the Data Shed, and everything cut out. Fortunately, the UPSes took most of the strain, but I still had to do a bit of running around to bring up some of the computers.
If only, I thought, if only I'd known that the power had cut out. Then I'd have had 15 minutes at least to sort it out before the UPSes ran out of power. You see, when this happens, the UPSes start beeping. But the Valkyrie Data Shed is a long way from where I usually am (50 feet, maybe), and there's no chance I'll hear their pathetic little beeps.
So, I thought, let's make it louder. I thought up all sorts of elaborate ideas, involving light bulbs and photocells, but what I wound up doing was this.
I thought, I'll put up a computer in there, that doesn't run off a UPS. And I'll check it periodically; if it's down, that means that power has failed. And I thought about that a bit more, and I thought that computers go down for other reasons (see about Gerda below, for example). And then I thought, I don't actually need a computer there, I just need something that I can ping. So I took one of the old Sonicwall firewalls, which I haven't used for years, (I made my own firewall, and it's working great) and I used that. There's no moving parts, you see, so it's less likely to stop working.
Karen (who also runs the Robot Arm and the Valk-signal) pings the Sonicwall once per minute. If there's no response, then there's something badly amiss somewhere. So Karen triggers the Screamer.
The Screamer is something I got for a burglar-alarm system, but I got one more of them than I actually used. You give it five volts, and it gives you 110 decibels of sound. Lovely. Enought to waken the dead, or at least, enough to waken a fast-asleep Valkyrie.
I said I'd tell you about Gerda. Gerda used to have a few functions, including the Robot Arm before I moved it. Now, all that Gerda does is the error processing for all the other servers. When you try to access a page that isn't on one of my servers, the error page that you see, is coming to you from Gerda.
Gerda has an 800 MHz Duron, and the CPU fan stopped working. Mechanical, you see. Mechanical things always stop working after a few years. Electronics last a lot longer. This particular fan was a weird design that I'd bought just one of. I can't remember why I never bought any more, but I'm glad I didn't.
When the CPU fan stopped working, the CPU just got hotter and hotter, until Linux went into a "kernel panic". That means "I'm so confused that I don't even know how to tell you that I'm confused". And the computer stopped.
Easily fixed - I installed a new CPU fan. But when I did, I realised that I have this sensibly-sized computer doing almost nothing. So, I moved the error-handling task over to a Pentium-133 that's been doing nothing except attract spam for the last few years. Gerda will be recycled as another newsthumbs storage machine.
The two new DSL lines arrived on schedule, June 6. So now my capacity is 100 megabits in Watford (that's where all the bulk transfer happens), a 2 megabit line here (used for old newsthumbs, secure server and some other stuff), and the six half-megabit DSL lines here, which are mostly used for backup of data from Watford to here. It means that I don't have to hit the road every couple of months, to change the servers.
Those six DSL lines give me 3 megabits of download at a fraction of the cost of the 2 megabit line (which is a professional-quality leased line). The downside is that they're not 100% reliable. They're pretty good, just not as good as the 2 megabit leased line.
So, just after I did the power-outage monitor, I did a DSL-line monitor. If one of the lines goes down now, I'll get an email to tell me so (plus there's some stuff that will try to bring it back online). It's not a big deal if one of the DSLs goes out. A power outage in the Valkyrie Data Shed is a big deal, and one that I'd want to know about ASAP. I'm even willing to be wakened in the middle of the night. The Screamer will do that.
The server that holds Boomerflex, Femflex and about 50 other DtV Family sites, was about 80% full. I don't like leaving a server until it gets 99% full, it means I *have* to do something urgently. I'd rather do things when I don't need to do them urgently.
So I replaced Hanni, a 366 MHz server with 80gb of disk, with a new server that's 2800 MHz and 470 gb of disk. So now it's only 11% filled, and the creation of the thumbnails each time they're redone, is ten times faster. I doubt if surfers will notice any difference. But the folks who are running their web sites there have noticed the speed improvement.
At the same time, I replaced the backup servers for the DtV Family sites. Surfers will never see those unless something goes very badly wrong, but they make me sleep better at night.
June 20, 4am. Norah got her knitting tangled up. The loading went up to 750 (1 is fine, 2 is large, 4 is very big and 8 is more than I'm happy with). And nothing was happening on the server. Norah is the front end for the web site, so it's one of the most important servers. From the point of view of surfers, the site was down.
I woke up several hours later, did a remote reboot of the computer (bless you, Masterswitch) and everything came back up just fine.
Another major upgrade has been in the Newsthumbs area. This hasn't been visible to surfers yet, but here's what's happened so far.
The volume of newsthumbs has been steadily growing. A couple of years ago, it was 30-40 thousand pictures/day, now it's 120-150 thousand. This means it takes longer to compute the thumbnails each day, and longer to download all those gigabytes from Watford to Chesham. That's why I've been adding DSL lines, it's a cheap bandwidth to do the download over.
The reason I download to Chesham, is that at Watford, I have cheaper bandwidth and expensive space (you wouldn't believe what it costs to rent a 19 inch wide, 36 inch deep, 6 foot high cabinet unless you're already doing this sort of thing). At Chesham, I have cheaper space (the Valkyrie Data Shed) and expensive bandwidth (my megastream leased line). So, I keep the current newsthumbs at Watford (they get accessed the most) and I keep the older newsthumbs at Chesham (they take up the most space for storage).
The volume has gotten to the point now where my standard 1-terabyte (1000 gigabytes) server is only lasting for less than two months. That means that every two months, there's another "older server", and I have to do a whole bunch of complicated things to make the changeover happen. But soon it'll be every six weeks, then once per month, and so on. I can see the writing on the wall.
So I've redesigned the system.
First of all, I try not to store the same file two or more times. It's in the nature of Usenet that the same picture (or other file) gets posted again and again. So, I'm identifying these files by computing an MD5 checksum on each one, and when I find two checksums the same, I delete one of the files, so it's only stored once (although it still looks like there's two files there, it's all done with mirrors [links, actually]). I've been doing that for some time.
The first change I've made, is in the downloading. Now, before I download a file, I check to see if it's already at Chesham. If it is, I don't download it, I just create a link to the file that's already there. How do I know if it's already there? I've created a database of the MD5 checksums of all the files on the server, and I check that database. This is saving me maybe 20-40% of my download, and it means that I don't actually need the two extra DSL lines that I just got put in. Oh well, I'll need them eventually.
The main change will be a lot more far-reaching. I currently compute the thumbnails on one server in Watford (and also on another one, which is my backup in case the main one crashes), and it takes 18 or 20 hours per day. You can see that I don't have much headroom for growth there, these computers can only work 24 hours per day. I'm planning to divide that work over four servers, so I'm expecting that each one will be able to finish the job in 5-6 hours. So, I need four servers where I had one, and I want a backup in case of failure, so I need a second set of four servers to duplicate those. And if a computer crashes, I don't want to have to make a trip to Watford to replace it, so I want a couple of spare servers sitting there doing nothing. So that I'll need two major and permanent crashes before I need to make the trip again. So, I need ten rather hefty computers.
I already have five in place (I had six, but one went pear-shaped). So I'm planning to take out from service, eight servers in Watford (two that aren't working, three that are so old, they're less than 500 MHz, and three others that are sort of OK but not high spec). And I'll put in eight servers, each of them running a 3.2 GHz Pentium (Athlons generate too much heat for the 1 3/4 inch high computer cases that I use to save space), 2 gb of memory, and six 400 gb drives. The main cost will be those 48 hard disks, which retail at about $500 - $600, but I plan to get them wholesale.
I plan to change the "older server" scheme. I'll leave the ones that I already have, but my idea is that there will be Older Servers 3 to 13 (as currently), and then one more older server which will last for years and years instead of months. The way it will do that, is that Older Server 15 will be a database. It won't actually store the pictures, it will store a database of where the pictures are. So, when you access alt.great.pictures/0320/myfavouritepicture.jpg, the Older Server 15 will get that request, and instead of just accessing the file with that name, it will feed the request to a program. The program will look up alt.great.pictures/0320/myfavouritepicture.jpg in a database and discover that the MD5 checksum of that file is 2382ef71032c7e731928ed62b19027ec. Then it will consult another database, and discover that the file with that checksum is stored on server 26, with the filename alt.prettygood.pictures/1264/anotherpic.jpg, so it will got and get that file, and present it to your browser.
That means that the program that does all this (which I'm imaginatively calling "get") will be doing a lot of work, and will be run every time anyone access the older newsthumbs. I might use mod_perl to do it, which is something else I need to look into, because I've never used it before, but apparently, it makes things faster when you're using the same perl program again and again. And please don't anyone suggest I write this in C, I'd rather learn Ancient Greek.
Searching wil change, too. Instead of you needing to search each Older Server separately, I'm planning to do a search engine that lets you search them all with one search.
All this sounds far too complicated, of course, so it'll probably give me lots of grief before I get it working. I'll make lots of mistakes in doing it, and I'll have to find and correct these mistakes. But it'll mean that you'll be able to access all the older stuff from that point on without having to go to a different server every couple of months. I might also be able to make searching work the same way.
For me, it means that I won't have the bimonthly stress of swapping servers; it'll happen pretty much continuously and automatically. All that I'll need to do, is keep building these 1-terabyte servers at Chesham, which isn't difficult to do, they take maybe 30 minutes to make, and another 15 minutes to install Linux and the newsthumbs software. And when one gets filled, I'll just tell the downloaders to start filling the next one.
The database I'm using is the Berkeley DB, which costs me $0 (it's Open Source) and is well spoken of by the people who speak of such things. I've also thought of using MySQL, but that's not really aimed at doing the sort of think I'm planning. I think the Berkeley DB will be faster and lighter. The big drawback of the Berkeley DB, is that if I'm writing to it, I can't also read from it. So I might have to use MySQL. I'm planning to make a trip down to London, to raid Foyles from some books on this. In this area, I really don't know how to proceed, but I've never let that stop me before, all the information you need is in books.
So, that's the plan. Of course, no plan survives contact with reality, so what I wind up doing is likely to be a bit different. But it's good to have a plan, it means I know which way I'm heading. Yes I do expect there's fifteen better ways to do this using something I've never even heard of, and that the way I plan to do it is A) unoriginal and B) suboptimal, but it's the best that I could come up with, and it looks to me like it should work. This isn't some dead important government-contracted hospital-records-taxation-vehicle-licence-and-passport system that have been so brilliantly successfluly written by teams of hotshot accountant/system-engineers. So I don't have the billion-dollar software budgets that those folks have. Just one Valkyrie.
There's about 100 million pictures on Newsthumbs. This plan should be able to handle four billion. At current rates of picture flow, that takes me 70 years forward, but the rate is speeding up, so it'll probably only last 25 years, maybe less. And then I'll have to think of a new design.
Nothing new.
I don't make these up. These are actual spams sent to me, which just
strike me as funny. I don't include their contact details - go find your own spammers!
We hsaJve be8en notified that your mpwortg?agte r#aGtpde is fixed aHt a
very hhigh interest rpazte. Tbyhevrefore you are cuUqrre*nt
ovFlerpapying, wvhicrh sums-uMp to thoZRusan(dNs of dollarbs annu^aloly
Sorry, could you say that again without the chewing gum in your mouth?
We htmave been ntotiQfiseZId th+at youger mo7rtga*ge rJkate is fixed at
a very high inUt~erest r]a/te.
No, I stll can't understand you. Do you speak Spanish? Latin? Greek?
It's to good to pass up!
No, it isn't.
This letter is not an advertisement. It's a business proposal.
This letter is not a business proposal. It's an advertisement.
We've sponsored lots of the women; Nicole Bass, Andrulla Blanchette, Sheila Burgess, Christine Envall, Marilyn Perret, Peggy Schoolcraft, Larisa Hakobyan, Steph Parks.
We're also sponsoring individual events such as the Femsport Valkyrie Festival, and the New York Muscle Club, and funding athletes to go to events with grant dollars.
We're also doing free hosting and free bandwidth for many of our sponsored women. Bandwidth can mount up to a large bill when you're running a popular web site.
And we've sponsored Heather Foster, Kara Bohigian, Priscilla Ribic, KerryAnn Allen, Linda Cusmano and Jodi Miller. Also Anita Ramsey and Rhonda Dethlefs.
Member | Posts |
| zig563 | 5059 |
| madman3579 | 4905 |
| pangel004 | 3847 |
| tre1313 | 3491 |
| maeven032 | 3203 |
| ens1961 | 3062 |
| tkokidd0 | 3013 |
| boomerflex | 2965 |
| buffy18976 | 2917 |
| TomNine | 2879 |
| hiram2000 | 2872 |
| rainer0000 | 1725 |
| lpdorman69 | 1697 |
| mit19237 | 1612 |
| jcc115 | 1497 |
| ginny2442 | 1412 |
| micha74 | 1323 |
| Jabber | 1304 |
| gaily304 | 1291 |
| guest | 1154 |
Zig, by a short nose from Madman, with Pam not too far behind.
I've put up a new style of board; it includes threading, the possibility of moderation, and lots more. You can see what Tom Nine's Tussling Tenement looks like, we gave it a new coat of paint, and it looks great. Mixed wrestling sessions is a popular topic!
This month we had 3718 posts to the boards.
Most posted Board of the month | Poster of the month | ||||||||||||||||||||||
|
|||||||||||||||||||||||
| Lots of great discussions going on the Politics board; outsourcing and upcoming elections. But the board named for William Wickenstein is still frisky. | HomoAncient is again this month's outstanding poster. |
Mavis is counting the number of times the message list is checked for each board. This gives a very different picture from the one above.
Most listed Board of the month | Most read Board of the month | ||||||||||||||||||||||||
|
|
||||||||||||||||||||||||
| With TomNine going to the new board style, he's not included in the stats, but his new board has been very successful. With the absence of Tom, Fistman is in the lead now. | The Grinch got the stats. |
Thatcherist entrepreneurialism is still alive and kicking (and selling used panties) in the Conservative Party
So there I was, quietly reading the Saturday newspaper in the smallest room in the house, when my Screamer went off. Yes - power cut. But I'm glad to say, everything functioned as it should. It gave me quie a shock, though.
I checked the site statistics that Sandra counts up each night.
At the end of June 2004, there were about 715,000 pictures (45 gigabytes), 153 gigabytes of video, 8400 text files (mostly stories) and a total of about 198 gigabytes. The Current Newsthumbs has 6.8 million pictures; there's about 96 million pictures altogether in Newsthumbs. We'll hit 100 million within a couple of months.