Frankly, I've never bought that old CW about girls being worse at math than boys... especially since I met and married my math-major wife in college, who has always been much better at math than I am.
This tells us a lot more about your mathematical skills than your wife's.
For the love of all things holy, please study statistics for a minimum of 15 minutes before opening your mouth again.
When a more experienced tech at my current job started to take me under his wing and mentor me, and then completely stopped talking to me after I stopped wearing makeup
Why did you stop wearing makeup? Did you change any other habits/behaviors at the same time?
The only reason I ask is that the guy may have reacted to some other change in your demeanor, perhaps whatever it was that made you decide not to wear makeup in the first place.
A perfect example would be if you had two applicants for a unit, and you chose one over the other. If you have messages to/from both applicants, and the person who lost immediately sends you an email stating they are 'a protected class, and you have no right to deny me a rental' - it is reasonable to expect trouble.
I get that all the time. First of all, it's not true (not that you'd need to know this, but whatever). The law only says that I cannot reject an applicant specifically because he/she is a member of a protected class. It says nothing about that I cannot reject a protected class member--everybody's in a protected class because "sex" is a protected class.
My written policy says: "For everyone's protection, all requests for application scoring much be submitted in writing to the PO Box." This is for two reasons: 1. It takes no effort to yell over the phone, but submitting a request in writing takes 2 minutes' effort. People who tend to feel discriminated against tend to... well... if they were putting forth effort and succeeding in life, they wouldn't have to blame others for their problems. Indeed, in all my years of landlording, nobody has ever submitted a request in writing to the PO box for any item that my policy requires.
2. It creates a paper trail. If someone wants to sue me, fine. But there will be no "he said, she said" arguments over a phone conversation because as soon as someone gives any indication that he/she wants to sue me, I give the PO box address and hang up the phone.
If your case makes it up to federal civil court, then those judges will expect you to produce - and if you don't, then you weren't interested in winning your case after all (no excuses).
Fortunately, I don't get sued all that much, and have never had a fair housing claim. I really doubt a typical landlord/tenant issue would make to to federal court, but I guess stranger things have happened. Some people really want their damn security deposit back, no matter how badly they've damaged the unit.
Obviously, if you end up in this sort of trouble, the question is going to boil down to: "So, Dan541, you admit deleted your email to and from PersonXXX. You also admit that the relationship (business or otherwise) between you and PersonXXX was destined for trouble. How can you say your deletion of email met the requirements of the Federal Rules of Civil Procedure?"
"In order to control the cost of storage and backup, our company instituted an email retention policy where all email is retained for precisely 180 days. We do not examine the contents of each email that is deleted after 180 days, so we could not have possibly known that deleted email would be the subject of legal action. Obviously we suspended our automatic deletion process once we were made aware of this litigation so as not to inadvertently purge discoverable evidence."
Personally, I think such an explanation would carry more weight if the retention time were greater than 180 days. What constitutes a reasonable retention time probably varies by industry, but 6 months sounds awfully short.
As it turns out, I'm in an industry that has conflicting rules/guidance on retention times. As a landlord, I maintain private information on individuals, and I am supposed to purge that after it is no longer needed (so I don't accidentally release it). On the other hand, if I were subject to a Fair Housing claim, I would need all rental applications and scoring to prove that I do not discriminate against any protected class. My policy is to retain all applications since I might be subject to a future Fair Housing claim, and those applications would then be needed for discovery/defense.
As a father of 3, I cannot fathom what drives a person to do that to their own child. An adult can create conflict that may drive you to retaliate. A 3 year old cannot.
Your kids must have been extremely atypical as 3 year olds, or you must have an atypically lousy memory.
3 year olds are programmed to drive you out of your fucking mind. It's in their genes. I don't think that's what happened in this case, however.
I simply bill GoGrid as an "alternative" to EC2 with some additional features.
With GoGrid, can I make my own server images? I'd have to, because you don't support any Linux distros other than RedHat.
Do you have something analogous to EC2's "Availability Zones"? Please don't point to your SLA, because a usage credit could never cover my lost business and lost goodwill. With GoGrid, one careless backhoe operator could put your customers out of business for a week. EC2 doesn't provide an SLA, yet I trust their geographically-diverse infrastructure more than I trust anybody's single point of failure datacenter.
Don't believe a single disaster could take you and your customers offline for major amounts of time? Just ask theplanet if they are believers now.
I don't know enough about GoGrid to point out any other deficiencies, but I imagine they're there. I realize that I'm coming off as a bit of an asshole, and perhaps that is an accurate characterization. On the other hand, I might be inciting some healthy competition here, which can only benefit everyone in the end.
If GG supported custom server images, I could see using you as a cold spare for EC2. In the event of a serious EC2 or S3 outage, someone could fail over to GG by altering a few DNS entries and spinning up a few GG servers. Does GG offer storage that isn't associated with any server instance? If so, that would be a real possibility.
In terms of GoGrid vs. EC2 comparisons, there are, obviously, some differences between the services. More info here.
You know, it's kind of funny. That comparison was one of the many reasons I never evaluated GoGrid. I figured, if you were that misleading in your claims there, you must be misleading in all of your other claims as well.
For example:
You assert that "1 GB RAM/1 Xeon Core server deployed for 1 hour" costs $0.10 on both GG and EC2. However, further investigation reveals that an EC2 instance actually does cost $0.10 per hour, however, a GG instance costs $0.19 per hour, or nearly 2x the cost of EC2! To ice the cake, an EC2 instance has 1.7GB of RAM, which is superior to GG's 1.0 GB--also by nearly a factor of 2x.
Also, you assert that "Outbound data transfer/GB" is $0.25 for GG and $0.17 for EC2. However, it seems that while the first GB of outbound transfer for EC2 is correctly quoted, the first GB outbound from GG is exactly 2x the price in your chart, or $0.50.
You also assert that it is impossible to deploy Windows on EC2. This is a falsehood.
You further assert that an EC2 instance can't be booted in under 5 minutes. Also false.
So if you are willing to be so generous in describing your capabilities in areas that are provably false, why should I believe any of your other marketing claims?
We don't yet have enough data to know what our load looks like, or when/why it will vary.
Well, this is where the rightscales of the world (I forget the names of the other automated ec2 scalers are.. scalr or something?) fit in. If you don't want to roll your own monitoring/scaling, you can rent somebody else's.:) Or rent someone else's until you have the time to roll your own.
But, anyhow, if you have such variable load, you'd most likely save money (I haven't seen your numbers, so who can say for sure?) using some type of rentable CPU. During the holiday seasons, do you really want to buy new hardware that will sit idle for the rest of the year? Or do you want to sign a 12-month contract on a dedicated server that you won't need come January?
The usual complaint against EC2 and the like is, "I can get an el-cheapo server from Dell for $1000, and I can get colo from whomever for $50/mo including 1TB of traffic. That's $1600 to get a server for a year, or I could spend $242.75/mo getting the same setup from EC2 and not even own the server!" Well, guess what, if that's your need, then Elastic Cloud Computing is not for you.
On the other hand, if you need 2 servers from January through October, but you need 3 servers in November and December, all of a sudden you are spending $3000 in servers + $1800 in colo for the whole year. Whereas you would be spending only $314.75/mo on EC2 from Jan through Oct and $386.75 in Nov/Dec = $3921.
True, you own the servers in the colo solution, but we also only added 1 server for the holiday season. What if you need 2 more or 3 more? What if you need more at different times of the day instead of the year?
I have a client who does a huge amount of processing at the end of each month, and the results need to be available within 24 hours. When I told them that I could bring up a server farm with 400 64-bit Xeon CPUs with 140GB of RAM in under 5 minutes at a cost of $16.00/hr (paid only when in use), they thought I was making fun of other consultants who "promise the world" and fail to deliver. I had to tell them I was serious.
This is what EC2 is really useful for. I understand it can be used for web hosting, but the real value is when you take full advantage of the "E" in EC2.
EC2 gives me almost none of that, without something like RightScale.
Ummm... maybe you should be using RightScale, then?
If you need hundreds of thousands of calls per second, you're going to need a lot of resources... I don't know what rightscale costs, but I don't think it's very much.
For your purposes, do you really need something like auto-scaling? If your load doesn't vary much, just spin up a bunch of EC2 instances and run a monitoring program. Overloaded? Spin up a few more instances.
I've never used glassfish. I'm curious why you recommend it so strongly over JBoss.
For app servers and servlet containers, I've used Weblogic, Websphere, JBoss, Jetty, and Tomcat, and I couldn't say that one or the other of them made my life 1000x easier.;)
Keep in mind, that's not the only way to do it -- dynamic DNS is still feasible, so that's another way to narrow the window during which clients have to figure the situation out themselves.
Doesn't dynamic DNS take more than 15 minutes to propagate though the Internet? I think most DNS servers disregard TTL values of under 60 minutes, no?
Thanks for that link. It's great to have my curiosity settled regarding whether or not they could have processed all of those TIFFs in 24 hours on EC2.
Do you happen to know which OCR software they used? It wasn't clear to me from the article.
Are online services cheaper and easier to use than the enterprise? In some cases, certainly. Some internal IT departments require a $10k+ tax on top of server purchases to cover IT installation and provisioning costs, and then take 2 weeks to 2 months to bring the server online.
This, I think, is the Big Deal, here.
My current client, in the mid '90s, embarked on a huge digitization project. They hired an external vendor to scan hundreds of thousands of documents for them, and the deliverable was approximately 100GB of TIFF images that needed to be OCR-ed into PDF. To accomplish this, they bought 5 servers, 5 Windows licenses, and 5 copies of Adobe Acrobat Capture (or whatever it's called) + data center costs. I think it took a month or two to process all of those TIFFs.
If this client had to do such a project today, assuming they were OK with their documents being outside the corporate firewall (which is a HUGE assumption), I would have done this using EC2/S3/SQS. I'd build a quick script that uploaded the TIFFs to S3, plopped a metadata message on the queue (start/end page number for each document, etc.) for each document, and then I'd build a script on an extra-large CPU intensive that pegged the CPU with OCRing, popping a message off the queue, grabbing the TIFFs from S3, and plopping the PDFs back to S3. I would just add more and more OCR instances until the processing rate was acceptable, and when the project was over, I'd download the PDFs and be done with it.
I'm guessing that using EC2, the project could have been done in a week, and the processing done in a day. It probably would have been less than 1/10th the cost, too. Do you know what my client did with those servers after the PDF conversion? They just left 'em in service in case they needed to convert any more TIFFs into PDFs in the future (they did not) and promptly forgot about their existence. So they paid for all that hardware, software, electricity, and data center space, for 10 frickin' years, when they could have paid for it for 24 hours.
I'm convinced that cloud computing is going to be huge. Not necessarily the EC2s of the world, but even inside corporate data centers. Using VMWare, or some of the other virtualization packages, a company can really scale down its hardware requirements. Currently, the servers supporting business-hours computing and after-business-hours processing is on separate hardware, with much of that hardware sitting idle for half the day. Why not use resources from the idle webservers for overnight batch processing and vice versa?
Amazon's Elastic IP, for example, can take 15 minutes to switch between instances -- something like 10-12 minutes during which requests are sent to the old instance, then 2-3 minutes during which all traffic is dropped on the floor and no instance is reachable, and then, finally, traffic is routed properly to the new instance.
Have you experimented at all with round-robin DNS records pointing to two different elastic IPs in different availability zones? If one availability zone goes tits-up, does round-robin yield acceptable performance, routing clients to the good zone?
Speaking as a lawyer, I am always impressed by how much control slashdotters think we have over our clients.
I am a consultant, and I have a ton of control over my clients. Ultimately the client is the boss, of course, but I can be pretty persuasive.
Try this one next time: "I don't know how to impress upon you just how serious this is, but I'm going to try. Your trial and sentencing are going to be over in 3 weeks. These 3 weeks are the most important 3 weeks of your life thus far. One cock-up will mean jail time, and there's nothing I can do about it. Get drunk and manage to get yourself a minor consumption? You're gonna spend the next few years of your life locked in a cage. Stay sober, go to some AA meetings, etc."
I know you can't control your client, but you can certainly shock 'em.
I'm going off-topic here, but I still don't understand why you americans accept this crazy 'no alcohol before you turn 21' thing. I was under the impression that you are an adult when you turn 18? Why make an exception for alcohol?
Most Americans think it's stupid, including myself. Only reason I brought it up is that it is still illegal, and if Lipton had been unfortunate enough to be cited for consumption before sentencing, it would have meant him getting locked in a cage instead of a $50 fine.
Off topic, but my wife and I let our kids drink responsibly, as in a glass of wine with dinner. But if we were to allow one of their friends a glass of wine with dinner? We'd be facing jail time. Fucked up, eh?
Now, it may not be so when prosecutors dredge up photos unrelated to, older, than, or from a different person with the same name, so this only argues for more transparent ways for hosts, services, and users to find unshakeable ways to authenticate what happens under their aegis. opt-in automatic encrypted transmission watermarks, anyone?
Only problem is these photos were not used as evidence. The trial was already over. Only sentencing remained.
Those photos never could have been admitted as evidence at his trial unless you got the photographer to take the stand and say that he witnessed Lipton partying, took the pictures, and that the subject was Lipton, etc. After all, you can't cross-examine a photograph.
Hello? Did partying now become some sort of grounds for a harsher verdict? What should he have done? Mourn and weep for at least 2 years or whatever the court deems "appropriate"?
The verdict never changed. It was the sentencing.
Lipton nearly killed someone, and was given an appropriate sentence. A lot of times, if a convict shows serious remorse, enrolls in alcohol treatment programs, etc., a judge will reduce the sentence because the convicted has already had some personal justice. Nothing new here.
In this case, Lipton showed no remorse, so the judge simply gave an appropriate sentence for his crime, rather than a reduced sentence.
The only "news" here is the fact that the prosecutor used Lipton's facebook profile to document Lipton's lack of remorse. The same thing would have happened had he prosecutor brought in witnesses who attended the party, or if Lipton got a minor consumption ticket (he is only 20, so he shouldn't have been drinking at all), etc.
So what we have is a guy who was known for drinking alcoholic beverages, now drinks non-alcoholic Red Bull instead.
It was only 2 weeks after he nearly killed someone because of his partying antics. His lawyer is lousy, all right, but only because he should have made sure lipton:
1. Did not go out partying at all. 2. Enrolled in Alcoholics Anonymous and started attending meetings. 3. Enrolled in any other local alcohol treatment programs might be useful. 4. Sure as shit stayed away from alcohol. We don't know he was drinking at that Halloween party, but I'm just saying, he was 20 years old. If he would have gotten a minor consumption ticket, the consequences would have been jail time.
Apparently, the lawyer couldn't convince his client to lay low and pretend to be remorseful for just two frickin' weeks. It doesn't matter if Lipton was drunk in those pics or not. It showed he was out partying while the woman he nearly killed was still in the hospital.
I'm unfamiliar with MySQL's partitioning -- is it radically different from postgresql's partitioning?
Yes, it is radically different. MySQL has partitioning, PostgreSQL does not.
I'm using inheritance to implement table partitioning with a rather large (50+ gig) PostgreSQL/PostGIS database. Constraint exclusion allows the query planner to use CHECK constraints to avoid even looking at tables where conditions contradict the constraints.
That's not partitioning, that's an ugly hack that will save you some I/O.
Slashdot Mirror
Fast forward to today. They're housewives. I'm a Software Engineer. It's sad and disheartening. [...] It's like puberty fried their brains completely.
What's sad and disheartening? That you have to work and they don't?
Whose brains got fried, again?
Frankly, I've never bought that old CW about girls being worse at math than boys... especially since I met and married my math-major wife in college, who has always been much better at math than I am.
This tells us a lot more about your mathematical skills than your wife's.
For the love of all things holy, please study statistics for a minimum of 15 minutes before opening your mouth again.
When a more experienced tech at my current job started to take me under his wing and mentor me, and then completely stopped talking to me after I stopped wearing makeup
Why did you stop wearing makeup? Did you change any other habits/behaviors at the same time?
The only reason I ask is that the guy may have reacted to some other change in your demeanor, perhaps whatever it was that made you decide not to wear makeup in the first place.
A perfect example would be if you had two applicants for a unit, and you chose one over the other. If you have messages to/from both applicants, and the person who lost immediately sends you an email stating they are 'a protected class, and you have no right to deny me a rental' - it is reasonable to expect trouble.
I get that all the time. First of all, it's not true (not that you'd need to know this, but whatever). The law only says that I cannot reject an applicant specifically because he/she is a member of a protected class. It says nothing about that I cannot reject a protected class member--everybody's in a protected class because "sex" is a protected class.
My written policy says: "For everyone's protection, all requests for application scoring much be submitted in writing to the PO Box." This is for two reasons:
1. It takes no effort to yell over the phone, but submitting a request in writing takes 2 minutes' effort. People who tend to feel discriminated against tend to... well... if they were putting forth effort and succeeding in life, they wouldn't have to blame others for their problems. Indeed, in all my years of landlording, nobody has ever submitted a request in writing to the PO box for any item that my policy requires.
2. It creates a paper trail. If someone wants to sue me, fine. But there will be no "he said, she said" arguments over a phone conversation because as soon as someone gives any indication that he/she wants to sue me, I give the PO box address and hang up the phone.
If your case makes it up to federal civil court, then those judges will expect you to produce - and if you don't, then you weren't interested in winning your case after all (no excuses).
Fortunately, I don't get sued all that much, and have never had a fair housing claim. I really doubt a typical landlord/tenant issue would make to to federal court, but I guess stranger things have happened. Some people really want their damn security deposit back, no matter how badly they've damaged the unit.
Obviously, if you end up in this sort of trouble, the question is going to boil down to: "So, Dan541, you admit deleted your email to and from PersonXXX. You also admit that the relationship (business or otherwise) between you and PersonXXX was destined for trouble. How can you say your deletion of email met the requirements of the Federal Rules of Civil Procedure?"
"In order to control the cost of storage and backup, our company instituted an email retention policy where all email is retained for precisely 180 days. We do not examine the contents of each email that is deleted after 180 days, so we could not have possibly known that deleted email would be the subject of legal action. Obviously we suspended our automatic deletion process once we were made aware of this litigation so as not to inadvertently purge discoverable evidence."
Personally, I think such an explanation would carry more weight if the retention time were greater than 180 days. What constitutes a reasonable retention time probably varies by industry, but 6 months sounds awfully short.
As it turns out, I'm in an industry that has conflicting rules/guidance on retention times. As a landlord, I maintain private information on individuals, and I am supposed to purge that after it is no longer needed (so I don't accidentally release it). On the other hand, if I were subject to a Fair Housing claim, I would need all rental applications and scoring to prove that I do not discriminate against any protected class. My policy is to retain all applications since I might be subject to a future Fair Housing claim, and those applications would then be needed for discovery/defense.
As a father of 3, I cannot fathom what drives a person to do that to their own child. An adult can create conflict that may drive you to retaliate. A 3 year old cannot.
Your kids must have been extremely atypical as 3 year olds, or you must have an atypically lousy memory.
3 year olds are programmed to drive you out of your fucking mind. It's in their genes. I don't think that's what happened in this case, however.
I simply bill GoGrid as an "alternative" to EC2 with some additional features.
With GoGrid, can I make my own server images? I'd have to, because you don't support any Linux distros other than RedHat.
Do you have something analogous to EC2's "Availability Zones"? Please don't point to your SLA, because a usage credit could never cover my lost business and lost goodwill. With GoGrid, one careless backhoe operator could put your customers out of business for a week. EC2 doesn't provide an SLA, yet I trust their geographically-diverse infrastructure more than I trust anybody's single point of failure datacenter.
Don't believe a single disaster could take you and your customers offline for major amounts of time? Just ask theplanet if they are believers now.
I don't know enough about GoGrid to point out any other deficiencies, but I imagine they're there. I realize that I'm coming off as a bit of an asshole, and perhaps that is an accurate characterization. On the other hand, I might be inciting some healthy competition here, which can only benefit everyone in the end.
If GG supported custom server images, I could see using you as a cold spare for EC2. In the event of a serious EC2 or S3 outage, someone could fail over to GG by altering a few DNS entries and spinning up a few GG servers. Does GG offer storage that isn't associated with any server instance? If so, that would be a real possibility.
In terms of GoGrid vs. EC2 comparisons, there are, obviously, some differences between the services. More info here.
You know, it's kind of funny. That comparison was one of the many reasons I never evaluated GoGrid. I figured, if you were that misleading in your claims there, you must be misleading in all of your other claims as well.
For example:
You assert that "1 GB RAM/1 Xeon Core server deployed for 1 hour" costs $0.10 on both GG and EC2. However, further investigation reveals that an EC2 instance actually does cost $0.10 per hour, however, a GG instance costs $0.19 per hour, or nearly 2x the cost of EC2! To ice the cake, an EC2 instance has 1.7GB of RAM, which is superior to GG's 1.0 GB--also by nearly a factor of 2x.
Also, you assert that "Outbound data transfer/GB" is $0.25 for GG and $0.17 for EC2. However, it seems that while the first GB of outbound transfer for EC2 is correctly quoted, the first GB outbound from GG is exactly 2x the price in your chart, or $0.50.
You also assert that it is impossible to deploy Windows on EC2. This is a falsehood.
You further assert that an EC2 instance can't be booted in under 5 minutes. Also false.
So if you are willing to be so generous in describing your capabilities in areas that are provably false, why should I believe any of your other marketing claims?
(RAID arrays are expensive)
You are doing something wrong, then.
Redundant Array of Inexpensive Disks
We don't yet have enough data to know what our load looks like, or when/why it will vary.
Well, this is where the rightscales of the world (I forget the names of the other automated ec2 scalers are.. scalr or something?) fit in. If you don't want to roll your own monitoring/scaling, you can rent somebody else's. :) Or rent someone else's until you have the time to roll your own.
But, anyhow, if you have such variable load, you'd most likely save money (I haven't seen your numbers, so who can say for sure?) using some type of rentable CPU. During the holiday seasons, do you really want to buy new hardware that will sit idle for the rest of the year? Or do you want to sign a 12-month contract on a dedicated server that you won't need come January?
The usual complaint against EC2 and the like is, "I can get an el-cheapo server from Dell for $1000, and I can get colo from whomever for $50/mo including 1TB of traffic. That's $1600 to get a server for a year, or I could spend $242.75/mo getting the same setup from EC2 and not even own the server!" Well, guess what, if that's your need, then Elastic Cloud Computing is not for you.
On the other hand, if you need 2 servers from January through October, but you need 3 servers in November and December, all of a sudden you are spending $3000 in servers + $1800 in colo for the whole year. Whereas you would be spending only $314.75/mo on EC2 from Jan through Oct and $386.75 in Nov/Dec = $3921.
True, you own the servers in the colo solution, but we also only added 1 server for the holiday season. What if you need 2 more or 3 more? What if you need more at different times of the day instead of the year?
I have a client who does a huge amount of processing at the end of each month, and the results need to be available within 24 hours. When I told them that I could bring up a server farm with 400 64-bit Xeon CPUs with 140GB of RAM in under 5 minutes at a cost of $16.00/hr (paid only when in use), they thought I was making fun of other consultants who "promise the world" and fail to deliver. I had to tell them I was serious.
This is what EC2 is really useful for. I understand it can be used for web hosting, but the real value is when you take full advantage of the "E" in EC2.
$16/hr. Outrageous!
There's an open source cloud engine. Eucalyptus, or something like that.
EC2 gives me almost none of that, without something like RightScale.
Ummm... maybe you should be using RightScale, then?
If you need hundreds of thousands of calls per second, you're going to need a lot of resources... I don't know what rightscale costs, but I don't think it's very much.
For your purposes, do you really need something like auto-scaling? If your load doesn't vary much, just spin up a bunch of EC2 instances and run a monitoring program. Overloaded? Spin up a few more instances.
I've never used glassfish. I'm curious why you recommend it so strongly over JBoss.
For app servers and servlet containers, I've used Weblogic, Websphere, JBoss, Jetty, and Tomcat, and I couldn't say that one or the other of them made my life 1000x easier. ;)
Keep in mind, that's not the only way to do it -- dynamic DNS is still feasible, so that's another way to narrow the window during which clients have to figure the situation out themselves.
Doesn't dynamic DNS take more than 15 minutes to propagate though the Internet? I think most DNS servers disregard TTL values of under 60 minutes, no?
Thanks for that link. It's great to have my curiosity settled regarding whether or not they could have processed all of those TIFFs in 24 hours on EC2.
Do you happen to know which OCR software they used? It wasn't clear to me from the article.
Are online services cheaper and easier to use than the enterprise? In some cases, certainly. Some internal IT departments require a $10k+ tax on top of server purchases to cover IT installation and provisioning costs, and then take 2 weeks to 2 months to bring the server online.
This, I think, is the Big Deal, here.
My current client, in the mid '90s, embarked on a huge digitization project. They hired an external vendor to scan hundreds of thousands of documents for them, and the deliverable was approximately 100GB of TIFF images that needed to be OCR-ed into PDF. To accomplish this, they bought 5 servers, 5 Windows licenses, and 5 copies of Adobe Acrobat Capture (or whatever it's called) + data center costs. I think it took a month or two to process all of those TIFFs.
If this client had to do such a project today, assuming they were OK with their documents being outside the corporate firewall (which is a HUGE assumption), I would have done this using EC2/S3/SQS. I'd build a quick script that uploaded the TIFFs to S3, plopped a metadata message on the queue (start/end page number for each document, etc.) for each document, and then I'd build a script on an extra-large CPU intensive that pegged the CPU with OCRing, popping a message off the queue, grabbing the TIFFs from S3, and plopping the PDFs back to S3. I would just add more and more OCR instances until the processing rate was acceptable, and when the project was over, I'd download the PDFs and be done with it.
I'm guessing that using EC2, the project could have been done in a week, and the processing done in a day. It probably would have been less than 1/10th the cost, too. Do you know what my client did with those servers after the PDF conversion? They just left 'em in service in case they needed to convert any more TIFFs into PDFs in the future (they did not) and promptly forgot about their existence. So they paid for all that hardware, software, electricity, and data center space, for 10 frickin' years, when they could have paid for it for 24 hours.
I'm convinced that cloud computing is going to be huge. Not necessarily the EC2s of the world, but even inside corporate data centers. Using VMWare, or some of the other virtualization packages, a company can really scale down its hardware requirements. Currently, the servers supporting business-hours computing and after-business-hours processing is on separate hardware, with much of that hardware sitting idle for half the day. Why not use resources from the idle webservers for overnight batch processing and vice versa?
Amazon's Elastic IP, for example, can take 15 minutes to switch between instances -- something like 10-12 minutes during which requests are sent to the old instance, then 2-3 minutes during which all traffic is dropped on the floor and no instance is reachable, and then, finally, traffic is routed properly to the new instance.
Have you experimented at all with round-robin DNS records pointing to two different elastic IPs in different availability zones? If one availability zone goes tits-up, does round-robin yield acceptable performance, routing clients to the good zone?
2. furthermore, nntp is a dying protocol
Ahh, but does Netcraft confirm it?
Speaking as a lawyer, I am always impressed by how much control slashdotters think we have over our clients.
I am a consultant, and I have a ton of control over my clients. Ultimately the client is the boss, of course, but I can be pretty persuasive.
Try this one next time: "I don't know how to impress upon you just how serious this is, but I'm going to try. Your trial and sentencing are going to be over in 3 weeks. These 3 weeks are the most important 3 weeks of your life thus far. One cock-up will mean jail time, and there's nothing I can do about it. Get drunk and manage to get yourself a minor consumption? You're gonna spend the next few years of your life locked in a cage. Stay sober, go to some AA meetings, etc."
I know you can't control your client, but you can certainly shock 'em.
I'm going off-topic here, but I still don't understand why you americans accept this crazy 'no alcohol before you turn 21' thing. I was under the impression that you are an adult when you turn 18? Why make an exception for alcohol?
Most Americans think it's stupid, including myself. Only reason I brought it up is that it is still illegal, and if Lipton had been unfortunate enough to be cited for consumption before sentencing, it would have meant him getting locked in a cage instead of a $50 fine.
Off topic, but my wife and I let our kids drink responsibly, as in a glass of wine with dinner. But if we were to allow one of their friends a glass of wine with dinner? We'd be facing jail time. Fucked up, eh?
Now, it may not be so when prosecutors dredge up photos unrelated to, older, than, or from a different person with the same name, so this only argues for more transparent ways for hosts, services, and users to find unshakeable ways to authenticate what happens under their aegis. opt-in automatic encrypted transmission watermarks, anyone?
Only problem is these photos were not used as evidence. The trial was already over. Only sentencing remained.
Those photos never could have been admitted as evidence at his trial unless you got the photographer to take the stand and say that he witnessed Lipton partying, took the pictures, and that the subject was Lipton, etc. After all, you can't cross-examine a photograph.
Hello? Did partying now become some sort of grounds for a harsher verdict? What should he have done? Mourn and weep for at least 2 years or whatever the court deems "appropriate"?
The verdict never changed. It was the sentencing.
Lipton nearly killed someone, and was given an appropriate sentence. A lot of times, if a convict shows serious remorse, enrolls in alcohol treatment programs, etc., a judge will reduce the sentence because the convicted has already had some personal justice. Nothing new here.
In this case, Lipton showed no remorse, so the judge simply gave an appropriate sentence for his crime, rather than a reduced sentence.
The only "news" here is the fact that the prosecutor used Lipton's facebook profile to document Lipton's lack of remorse. The same thing would have happened had he prosecutor brought in witnesses who attended the party, or if Lipton got a minor consumption ticket (he is only 20, so he shouldn't have been drinking at all), etc.
So what we have is a guy who was known for drinking alcoholic beverages, now drinks non-alcoholic Red Bull instead.
It was only 2 weeks after he nearly killed someone because of his partying antics. His lawyer is lousy, all right, but only because he should have made sure lipton:
1. Did not go out partying at all.
2. Enrolled in Alcoholics Anonymous and started attending meetings.
3. Enrolled in any other local alcohol treatment programs might be useful.
4. Sure as shit stayed away from alcohol. We don't know he was drinking at that Halloween party, but I'm just saying, he was 20 years old. If he would have gotten a minor consumption ticket, the consequences would have been jail time.
Apparently, the lawyer couldn't convince his client to lay low and pretend to be remorseful for just two frickin' weeks. It doesn't matter if Lipton was drunk in those pics or not. It showed he was out partying while the woman he nearly killed was still in the hospital.
I'm unfamiliar with MySQL's partitioning -- is it radically different from postgresql's partitioning?
Yes, it is radically different. MySQL has partitioning, PostgreSQL does not.
I'm using inheritance to implement table partitioning with a rather large (50+ gig) PostgreSQL/PostGIS database. Constraint exclusion allows the query planner to use CHECK constraints to avoid even looking at tables where conditions contradict the constraints.
That's not partitioning, that's an ugly hack that will save you some I/O.
I guess not. Oh well, at least I amuse myself. And apparently there is one mod out there somewhere who thinks I'm funny.