[Insert I have no moderator points, but consider this an outstanding posting here]
WTF is up with using an open-caret close-caret sequence? I've seen other people do it. I gave up just now, but does one need to use HTML entities to make that work even with "plain old text" formatting?
Our mail server is currently handling about 1M messages a day. IO became a serious issue. We're still using sendmail, and I'm not going to give it up (we know it, we have a custom builds for strange applications, it works). As others have noted, load average doesn't mean much here - I have some machines with a load average at 4 that are actually idle and fine, and others at.2 that need tuning. Ignore it and concentrate on what matters.
Assuming IO matters, I am putting my full faith (and job) on Mylex controllers. I love them. I only have one in production, but am about to deploy 5 more, and we'll come in at about 600G managed by them. They just work. The DAC960SXi I have in production (for 7months now) has been flawless, delivering wire speed doing RAID 5 without any effort after initial config (which is a bit annoying, to be sure).
My production system using it is doing far too many things - mail, staging server, enterprise backup. This is changing - lack of time and historical accident made it that way. The point is that the Mylex handles it with no grief.
If you're building these, be aware that Mylex external controllers need to be mounted in a box with "internal" style connectors. For good RAID cases, check out http://www.storagepath.com/ - they are what I'm using. They look low rent, but the boxes are nice (if a bit expensive).
Down to specifics. For a mail only machine doing the sort of volume you're talking about, I'd deploy a dual processor box with three SCSI busses (one for spool, two for mbox/system access - system access is pretty cheap in comparison) attached to two harware RAID setups. Granted volume allows, I'd go RAID 5 for spool (with 18G disks, that's ~65G spool) and hot spares. For mboxes, I'd do 0+1, for as much space as needed. Stripe disks on independent controllers, mirrored to each other. Striped mirrors can grow, as you need them to (RAID 5 can't, easily). You don't want to lose anyone's mail. Hot spares for each.
Assuming 100G of mboxes, that's a total of 17 18G disks. Add three Mylex DAC9660SXis and (initially) 3 rack mount cases, and that's something around ~24K.
Availability beyond disk is a different question, that gets platform specific. I do mainly Solaris now, so I can't talk much about Linux for this. Mylex controllers can do dual active/dual host configurations, but things get more complex, and a summary here doesn't make sense.
Other options like A1000s (Sun specific) and Netapps require different approaches - they're very different beasts. We have all of the above, and treat them very differently. We'll buy them all again - they're all decent - but are good at different things.
If you can, buy raw Mylex contollers through a reseller like TechData or similar - you'll save a lot.
Not to quibble, but the internal disk options for a 4500 are not FCAL (One has to purchase disk boards, which use a system slot, are 2x4.2G each, and very expensive. It is a silly option). The 3500 are, but you are limited to 8 9.1G disks.
For the most part, if you're buying x500s, you're probably looking at Ax000 disk arrays, too.
You're better of using a distributed peer-to-peer chat protocol with potential anonymizers in the mix than praying for "something that doesn't expose your IP". Better to protect your box and use a protocol that doesn't depend on a server than to depend on them to protect your IP address. No, there isn't a protocol that does this yet. Keep your eyes open, though. You might see one soon. -j
Banner ads are failing as a product industry wide because (a) they didn't fulfill the promise of giving advertisers microsegmented audiences and (b) because the total number of banners available for sale is rocketing as everyone's hit counts continues to rise. The reason why software specific banners could work much better is simple - the market sample is self selecting to a finer gradient than most websites. If you're looking at an online cook book, you're more likely to be interested in cookware. If you're looking at a JDK reference... You get the idea. This doesn't work as well on web pages because microsegmenting claims can't really be backed up with statistics very well. One wonders why print magazines do so well.
Shared storage is an issue. We bought a Netapp for that... that of course isn't within everyone's budget, nor am I completely happy with it (Short answer - a Netapp is nifty, and mostly functional. Mostly functional is Ok for something that costs 1/10th of the price. Hell, maybe I'm doing something wrong.). For most web related things, carbon-copy machines are fine until you get to the Yahoo range of traffic. But if you're that bleeding edge, you have tons of consultants telling you what to do with your servers, and thus different problems.
Bah. You don't know what toasters is. toasters@mathworks.com. does. Net Apps seem to be about spending a lot of money, fretting over the fruits of that cash (cache?), and endlessly, well, fucking with the parameters for this box that was advertised as simplified storage management*. *I may be poisoned, but I have yet to see a network-beast (aside from Cisco) that actually was what it was advertised to be. Netapp, etc. are not there yet (autonegotiating on standard hubs would be a start). My Filer 720 still spews NFS errors whenever I ask it to copy more than a few K. I suspect that's because I can't make it behave like the other Solaris box accessing it; which does so over a crossover cable. Go figure. I guess my next storage device is a Sun server with RAID. At least that works. Wait, it is cheaper, too. I can hope, tho. Some day, this netapp box will cease to be a lame ass bottleneck. I just hope I'm not dead then. Oh, yeah - slashdot nanotech will be there by then. Sorry. -j
Indeed, superbad is the shit. Those who use the web should hang heads in shame if superbad doesn't get beyond a nomination for the Webbies. Not that anyone cares what The Man thinks, anyway.
Slashdot Mirror
[Insert I have no moderator points, but consider this an outstanding posting here]
WTF is up with using an open-caret close-caret sequence? I've seen other people do it. I gave up just now, but does one need to use HTML entities to make that work even with "plain old text" formatting?
Our mail server is currently handling about 1M messages a day. IO became a serious issue. We're still using sendmail, and I'm not going to give it up (we know it, we have a custom builds for strange applications, it works). As others have noted, load average doesn't mean much here - I have some machines with a load average at 4 that are actually idle and fine, and others at .2 that need tuning. Ignore it and concentrate on what matters.
Assuming IO matters, I am putting my full faith (and job) on Mylex controllers. I love them. I only have one in production, but am about to deploy 5 more, and we'll come in at about 600G managed by them. They just work. The DAC960SXi I have in production (for 7months now) has been flawless, delivering wire speed doing RAID 5 without any effort after initial config (which is a bit annoying, to be sure).
My production system using it is doing far too many things - mail, staging server, enterprise backup. This is changing - lack of time and historical accident made it that way. The point is that the Mylex handles it with no grief.
If you're building these, be aware that Mylex external controllers need to be mounted in a box with "internal" style connectors. For good RAID cases, check out http://www.storagepath.com/ - they are what I'm using. They look low rent, but the boxes are nice (if a bit expensive).
Down to specifics. For a mail only machine doing the sort of volume you're talking about, I'd deploy a dual processor box with three SCSI busses (one for spool, two for mbox/system access - system access is pretty cheap in comparison) attached to two harware RAID setups. Granted volume allows, I'd go RAID 5 for spool (with 18G disks, that's ~65G spool) and hot spares. For mboxes, I'd do 0+1, for as much space as needed. Stripe disks on independent controllers, mirrored to each other. Striped mirrors can grow, as you need them to (RAID 5 can't, easily). You don't want to lose anyone's mail. Hot spares for each.
Assuming 100G of mboxes, that's a total of 17 18G disks. Add three Mylex DAC9660SXis and (initially) 3 rack mount cases, and that's something around ~24K.
Availability beyond disk is a different question, that gets platform specific. I do mainly Solaris now, so I can't talk much about Linux for this. Mylex controllers can do dual active/dual host configurations, but things get more complex, and
a summary here doesn't make sense.
Other options like A1000s (Sun specific) and Netapps require different approaches - they're very different beasts. We have all of the above, and treat them very differently. We'll buy them all again - they're all decent - but are good at different things.
If you can, buy raw Mylex contollers through a reseller like TechData or similar - you'll save a lot.
Hope this helps some.
-j
If Mr. T and Jesse Ventura got in a fight, who'd win?
When did Brian get an InterCap?
-j
For the most part, if you're buying x500s, you're probably looking at Ax000 disk arrays, too.
Typically devices like these run journaling file systems, so you don't fsck them.
You're better of using a distributed peer-to-peer chat protocol with potential anonymizers in the mix than praying for "something that doesn't expose your IP". Better to protect your box and use a protocol that doesn't depend on a server than to depend on them to protect your IP address. No, there isn't a protocol that does this yet. Keep your eyes open, though. You might see one soon. -j
Diamond age, here we come. Ads on chopsticks. Every surface moving, with ads. Ads. Ads.
Banner ads are failing as a product industry wide because (a) they didn't fulfill the promise of giving advertisers microsegmented audiences and (b) because the total number of banners available for sale is rocketing as everyone's hit counts continues to rise. The reason why software specific banners could work much better is simple - the market sample is self selecting to a finer gradient than most websites. If you're looking at an online cook book, you're more likely to be interested in cookware. If you're looking at a JDK reference... You get the idea. This doesn't work as well on web pages because microsegmenting claims can't really be backed up with statistics very well. One wonders why print magazines do so well.
Shared storage is an issue. We bought a Netapp for that... that of course isn't within everyone's budget, nor am I completely happy with it (Short answer - a Netapp is nifty, and mostly functional. Mostly functional is Ok for something that costs 1/10th of the price. Hell, maybe I'm doing something wrong.). For most web related things, carbon-copy machines are fine until you get to the Yahoo range of traffic. But if you're that bleeding edge, you have tons of consultants telling you what to do with your servers, and thus different problems.
Bah. You don't know what toasters is.
toasters@mathworks.com. does. Net Apps seem to be about spending a lot of money, fretting over the fruits of that cash (cache?), and endlessly, well, fucking with the parameters for this box that was advertised as simplified storage management*. *I may be poisoned, but I have yet to see a network-beast (aside from Cisco) that actually was what it was advertised to be. Netapp, etc. are not there yet (autonegotiating on standard hubs would be a start). My Filer 720 still spews NFS errors whenever I ask it to copy more than a few K. I suspect that's because I can't make it behave like the other Solaris box accessing it; which does so over a crossover cable. Go figure. I guess my next storage device is a Sun server with RAID. At least that works. Wait, it is cheaper, too. I can hope, tho. Some day, this netapp box will cease to be a lame ass bottleneck. I just hope I'm not dead then. Oh, yeah - slashdot nanotech will be there by then. Sorry. -j
Indeed, superbad is the shit. Those who use the web should hang heads in shame if superbad doesn't get beyond a nomination for the Webbies. Not that anyone cares what The Man thinks, anyway.
Um, what has the GPL to do with the BSD license?
Nevermind the fact that many BSD people have given
up the credit clause.
Please, come again?
Um... Wait. Because Be's browser can be easily
decoupled, and M$'s can't, that's good.
Wait, isn't that a cornerstone of the case
against M$?
He's hoping that the near future
is more evenly, um, coupled.
-j
Erm... TAFKAP (erm, Prince) earns that
label. I believe he's the first "big time"
artist to do so (for certain values of
big time...).
Depending on your definitions, there's also
Madonna.