(former NYT Engineer) Some potentially disruptive maintenance that can be localized to a single team (say, storage maintenance) would be pushed to an overnight change window. However, emergency reachability for certain teams at NYT can problematic, so some (expected) low-impact work requiring multiple teams is done during the day. They have multiple data centers and everything is fully redundant, so updates can happen to service-drained components, be tested and then flipped once verified. Apparently that could not or did not happen here.
FTR, the entire Digital Core Infrastructure team (responsible for Storage, Virtual Infrastructure, OS Mgmt, App Support and some networking) resigned approximately 3 months ago.
Former NYT Digital, Core Infrastructure Engineer here. The outage yesterday was a problem with their load balancers caused by internal action. Stop reading the nonsense posted by Fox Business News.
Anyone that gripes about wanting Flash on their phone/mobile device *HAS NEVER HAD* Flash on a mobile device. If they had, they wouldn't want it so bad.
*raises hand*
Here, please. I have an N800 and an N900. I *do* want flash on my devices and I *do* use it. Both of those devices have adblock plus to combat annoying ads with the built-in browser. The main thing is, I can still use flash if (and when) I want to. That's my choice to make - not the manufacturers.
Oh look! It uses hover states for mouse tracking - something that isn't supported on a touch interface...
Oh look! The N900 has a touch interface that supports hover!
Google Voice, Gizmo5 account (if you already have one; they're not accepting new requests) or a SkypeIn US number. Get a GSM smartphone with a simcard from a Canadian provider (since you're there most of the time). When in the states, get a pre-paid sim card with voice/data or just data. If you need the SkypeIn, it will set you back $30 for the year.
You now set up GV to forward calls to your US number to your VoIP account (Gizmo/Skype) while in Canada. Calls will be delivered via data. When in the states, you can continue with the same method, but with a prepaid simcard OR you can just forward via voice.
Note that while data plans for Canada or pre-paid US may be capped/metered you only need to use the GSM data when you are out-and-about. Any decent smartphone these days will happily shuffle data through wifi instead.
My Nokia N900 might be a bit too pricey, but will do everything here seamlessly with the built-in Skype and SIP integration.
The only thing this doesn't cover is porting your existing number to GoogleVoice...
I would add that the "floor mat" excuse always sounded like BS to me.
There may have been more than one conclusion or actual root cause for the stuck accelerators. I can confirm that on one of my (non-Toyota) cars I have personally seen 2 different aftermarket floor mat styles slide forward and end up in a position that stuck the gas pedal down.
Not yet...maybe soon...Part of their whole Project Dark thing is for them to start carrying high-end handsets. Rumor is that they'll have the Nokia N900 in Q1 '10.
TMobile is very friendly to unlocked phones. If you buy one of their subsidized phones, they will unlock it for you 90 days from purchase. They even have a gray market phone support department; they support the phones they don't even sell...
I would mod up if I could. I've got two separate Isilon clusters at work and they are indeed awesome. Just one of the features that go above and beyond standard RAID is what they call FlexProtect. I allows for the ability to handle your redundancy not just at a node/disk level, but right down to a directory/file level. Only need N+1 except for some critical data? You can give that file or whole directory N+4 while keeping the rest of the cluster N+1.
It appears that you misunderstand the aphorism...obscure != secret
There is some, albeit minimal value in security through obscurity...but that's not what you're talking about. Securing something via obscurity is akin to giving an attacker 100% of what they need to break the security and putting a layer of obscurity on top. Encoding a message with ROT13, for example. Giving an attacker the ROT13 encoded text and having the ROT13 algorithm available is everything they need to decipher the message.
Having a security system based on a secret key is *not* security through obscurity. The attacker simply does NOT have all of the information needed to break the system. Now, the system is only as secure as long as the key is secret. With a large enough keyspace, you can be more than reasonably assured that the system is secure as long as the key remains a secret.
...even if you don't want to spend for the HD, a TiVo Series 2.5 DT does dual tuning goodness at standard def and has been available for how many years?
Don't take this the wrong way, because I love RT, but it does have some serious performance problems in the large scale.
It's absolutely perfect for the volume that I'm currently administering (3K-4K tickets per year, roughly 15K correspondence, 70K transactions) and once you realize why certain things are laid out the way they are, the API is *very* easy to extend and customize.
The last RT shop that I worked in was a completely different story. We were doing 100K+ tickets per year easy (with that increasing at a rapid rate) and that was 6 years ago. When I left, about a year and a half ago, there were approx 700+ queues, many thousands upon thousands of users and millions of correspondence. There was quite a bit of performance tweaking we needed to do to the code and, of course, the DB just to make it useable. Simply iterating over the queue list using the API was taking upwards of 3 seconds. 3 seconds. Not to render the list in a browser. Not to access the names of each queue. Just to iterate through the list. That install is now plowing through tickets faster than even and is well past the 1M ticket mark, likely has in the tens of millions of correspondence and is performing adequately.
RT is a *great* tool, but you're not going to see performance at those numbers unless you have a firm grasp on DB tuning, perl, Mason and RT's architecture and get your hands very dirty - or you pay a hefty sum to Best Practical for support.
* VMware ESX - enterprise grade virtualisation server. Combined with vmware infrastructure, you run a bare minimum hypervisor (no overhead from a standard linux or windows OS host), store your virtual machines on a SAN or NFS, have a pool of physical servers and automatically load-balance your VMs between them or even bring them back up automatically if a physical server goes bang. Nearly completely abstract your servers from the hardware, run 20 servers per actual piece of tin. Very much not free.
Almost. What you're describing is full "VMware Infrastructure". ESX is the bare-metal hypervisor - and has actually been "replaced" with ESXi. This is a stripped down version that has a smaller install footprint and therefore a lower exposure to exploits. Most of the old ESX patches were for things like Samba and CUPS on the service console. ESXi now comes from vendors like Dell in an embedded form even: 32MB on an SD card, pre-installed, no hard drive required.
You can use local storage with ESX and ESXi; just format it with VMFS. When you're dealing with live-migration (vmotion), automated resource balancing (DRS) and bringing VMs back up after a bang (HA), that's all part of VMware Infrastructure and Virtual Center.
Very much not free - BUT the ESXi installable hypervisor? Free. Go download it now if you want it. Due to the stripped down nature, it supports a more limited set of hardware than VMware Server (which relies on a regular host OS to work out the hardware details), but it performs much better. Depending on the host hardware and VM workload, you can get a 20:1 VM:host ratio with your eyes closed.
Please, someone shoot this idea down - why not just do it with infrared lights on the sidelines or something similar? All you need is some focused, non-visible light broadcast at field level and you could paint that visible using a chroma-key-like system. Calculating the angles of this and the deflection of the field, etc, etc seems far more complicated than it has to be...
I took great joy in sending an "Teh intarnets is down!!" email a few months ago to our network guy. Exchange servers worked...dedicated line to the main office worked...outbound link from our office? Not so much.:-) He got the email, I still couldn't get to google.com.
There are many video games that I've reviewed for Plugged In that I wouldn't let my kids--or anyone's kids, if I had such power--get within a country mile of.
"My views are so obviously correct that nobody should be allowed to hold others..."
Not at all. I've been a T-Mobile customer for well over 5 years. They've always had great customer service and retention. A (long) while back I wanted to upgrade from my old v60 to the T610. AT&T was giving away the T610 (T616 actually, same phone though) for free with a new contract. T-Mobile was selling it for $50 to new customers and $100 for existing customers (that's after the new contract discount). I called them out on it, told them how ridiculous that was and that I wanted the T610 for free from them, just like AT&T was offering, or I would simply walk over to AT&T and get it there. Instantly, the rep said "I can't do that - it's in the system at a set price....but I CAN give you the next 3 months of service free, which works out to be a little more than the cost of the phone..."
They're in it to make money just like everybody else and they'll do what they can to get it. The thing I like about them is the fact that they seem to understand that keeping their customers is cheaper than getting new ones - so they actually TRY to treat their customer well.
They have an entire division dedicated to working with phones that they "don't support" (like the China-Mobile firmware Motorola e680i I had for a while). I've had support techs refuse to get off the line until my Linux laptop was GPRS tethering properly off that same "unsupported" phone. Add all that up and it becomes a little bit clear why I've stayed with them for so long.
I got my dates backwards (2003 for my t610); summer of 2004 sounds about right. The phone was never released by a carrier in the States. That said, I bought my unlocked e680 from a company in Florida. T-Mobile FULLY supported me twice over the phone with that handset. Once for getting MMS settings working on their network; the other time for getting my Linux laptop working over bluetooth DUN w/ GPRS ("unsupported phone" *division* and someone willing to work through a problem with an "unsupported" OS - yeah, I'm a customer for life).
I never got a terminal on the phone myself; it was just a phone with more features in a smaller package than anything else at the time to me. From what I remember from a few motodev and linux geek forums, people were able to crack it open and get a terminal eventually. It had no wifi and no rj45, so that would have been a limited use for me anyway. I did have an ssh client for it, though.:-)
Regardless, the e680 wasn't even the first Linux based phone on the market as I recall.
(oh wait, it's the *only* *nix based phone on the market - that actually works)
My Linux-based Motorola e680 from 2003 would beg to differ. It worked wonderfully, thank you. Full touch screen, minimal buttons (keypad was in the touch screen), mp3 audio, mp4 video playback (this all sound familiar?)....oh yeah and a (vga:-/) camera, video capture, fm tuner and an SD card slot...
As a former waiter / bartender, the size of the bill is *usually* related to the length of time for the transaction. Leaving a dollar on that $5 coffee / bagel is fine because the server will be doing a significant number of transactions over places with larger ticket items. A server can complete that entire transaction in less than a minute on its own. If you get multiple, similar requests at or near the same time, the server will also combine these actions; so, while they can do a single coffee in 45 seconds from station to table, they can do 6 coffees in only 90 seconds.
Also, at a diner, you might be able turn a table over every 30 minutes. At a nice, elegant restaurant, you may only turn the table over two, *maybe* three times a night depending on scale. People eating an expensive meal tend to like to sit, talk and enjoy the food rather than getting quick body-fuel. That requires more attention and round-trips to make sure everything is going well, drinks are filled, etc.
Waiting tables is a Hell job. Long hours on your feet (I used to do 6 doubles@16+ hours each and a dinner on the 7th day) and being barked at by inconsiderate jackasses make it a largely thankless job. If you can get things as efficient as possible, your customers will appreciate it and tip in kind. If you just try to get through the hours, it will show and reflect in your tips. Those servers that just try to grind through without regard to their customers don't deserve 15%. Their choice to work in a customer-oriented business while note caring about said customers does not require me to subsidize their life. Remeber TIP = To Insure Prompt/Proper Service.
That said, I regularly earned 20% average per shift; on everything from $10 lunches to $500 two person dinners. I tip the same when I am being served. Good service to me will get you a minimum 20%. With supremely excellent service, I typically ignore the percentage completely and have been know to leave well over 100%.
Any place offering free wifi should have a person that will swing through the dinning room to ask if people want anything else. Put up a sign that says "complementary wifi with service" or some such. Make it an honor system and police it when you're not totally slammed.
Slashdot Mirror
We left for separate opportunities; all within 6 weeks of each other. 100% related to management interference / incompetence.
(former NYT Engineer) Some potentially disruptive maintenance that can be localized to a single team (say, storage maintenance) would be pushed to an overnight change window. However, emergency reachability for certain teams at NYT can problematic, so some (expected) low-impact work requiring multiple teams is done during the day. They have multiple data centers and everything is fully redundant, so updates can happen to service-drained components, be tested and then flipped once verified. Apparently that could not or did not happen here.
FTR, the entire Digital Core Infrastructure team (responsible for Storage, Virtual Infrastructure, OS Mgmt, App Support and some networking) resigned approximately 3 months ago.
Fox Business were the first to "break" the story with a "source close to the matter". Everybody went off their lead. Every single technical statement in the article is incorrect and/or laughable. http://www.foxbusiness.com/technology/2013/08/14/new-york-times-site-experiences-major-outage/
Former NYT Digital, Core Infrastructure Engineer here. The outage yesterday was a problem with their load balancers caused by internal action. Stop reading the nonsense posted by Fox Business News.
Re: mobile synch - That already exists with the N900 and Firefox with the Weave plugin.
Anyone that gripes about wanting Flash on their phone/mobile device *HAS NEVER HAD* Flash on a mobile device. If they had, they wouldn't want it so bad.
*raises hand*
Here, please. I have an N800 and an N900. I *do* want flash on my devices and I *do* use it. Both of those devices have adblock plus to combat annoying ads with the built-in browser. The main thing is, I can still use flash if (and when) I want to. That's my choice to make - not the manufacturers.
Oh look! It uses hover states for mouse tracking - something that isn't supported on a touch interface... Oh look! The N900 has a touch interface that supports hover!
AT&T runs the one (1) and only GSM 3G network in the United States of America.
T-Mobile would be very shocked to hear this...
Google Voice, Gizmo5 account (if you already have one; they're not accepting new requests) or a SkypeIn US number. Get a GSM smartphone with a simcard from a Canadian provider (since you're there most of the time). When in the states, get a pre-paid sim card with voice/data or just data. If you need the SkypeIn, it will set you back $30 for the year.
You now set up GV to forward calls to your US number to your VoIP account (Gizmo/Skype) while in Canada. Calls will be delivered via data. When in the states, you can continue with the same method, but with a prepaid simcard OR you can just forward via voice.
Note that while data plans for Canada or pre-paid US may be capped/metered you only need to use the GSM data when you are out-and-about. Any decent smartphone these days will happily shuffle data through wifi instead.
My Nokia N900 might be a bit too pricey, but will do everything here seamlessly with the built-in Skype and SIP integration.
The only thing this doesn't cover is porting your existing number to GoogleVoice...
I would add that the "floor mat" excuse always sounded like BS to me.
There may have been more than one conclusion or actual root cause for the stuck accelerators. I can confirm that on one of my (non-Toyota) cars I have personally seen 2 different aftermarket floor mat styles slide forward and end up in a position that stuck the gas pedal down.
4) install distccd on a bigger system on the same network; emerge distcc as soon as possible on the smaller system
Nokia's handwriting calculator running on an N810 or the upcoming N900. Add in a bluetooth keyboard if you want something full size to type on...
Not yet...maybe soon...Part of their whole Project Dark thing is for them to start carrying high-end handsets. Rumor is that they'll have the Nokia N900 in Q1 '10.
TMobile Even More Plus.
TMobile is very friendly to unlocked phones. If you buy one of their subsidized phones, they will unlock it for you 90 days from purchase. They even have a gray market phone support department; they support the phones they don't even sell...
I would mod up if I could. I've got two separate Isilon clusters at work and they are indeed awesome. Just one of the features that go above and beyond standard RAID is what they call FlexProtect. I allows for the ability to handle your redundancy not just at a node/disk level, but right down to a directory/file level. Only need N+1 except for some critical data? You can give that file or whole directory N+4 while keeping the rest of the cluster N+1.
It appears that you misunderstand the aphorism...obscure != secret
There is some, albeit minimal value in security through obscurity...but that's not what you're talking about. Securing something via obscurity is akin to giving an attacker 100% of what they need to break the security and putting a layer of obscurity on top. Encoding a message with ROT13, for example. Giving an attacker the ROT13 encoded text and having the ROT13 algorithm available is everything they need to decipher the message.
Having a security system based on a secret key is *not* security through obscurity. The attacker simply does NOT have all of the information needed to break the system. Now, the system is only as secure as long as the key is secret. With a large enough keyspace, you can be more than reasonably assured that the system is secure as long as the key remains a secret.
...even if you don't want to spend for the HD, a TiVo Series 2.5 DT does dual tuning goodness at standard def and has been available for how many years?
Don't take this the wrong way, because I love RT, but it does have some serious performance problems in the large scale.
It's absolutely perfect for the volume that I'm currently administering (3K-4K tickets per year, roughly 15K correspondence, 70K transactions) and once you realize why certain things are laid out the way they are, the API is *very* easy to extend and customize.
The last RT shop that I worked in was a completely different story. We were doing 100K+ tickets per year easy (with that increasing at a rapid rate) and that was 6 years ago. When I left, about a year and a half ago, there were approx 700+ queues, many thousands upon thousands of users and millions of correspondence. There was quite a bit of performance tweaking we needed to do to the code and, of course, the DB just to make it useable. Simply iterating over the queue list using the API was taking upwards of 3 seconds. 3 seconds. Not to render the list in a browser. Not to access the names of each queue. Just to iterate through the list. That install is now plowing through tickets faster than even and is well past the 1M ticket mark, likely has in the tens of millions of correspondence and is performing adequately.
RT is a *great* tool, but you're not going to see performance at those numbers unless you have a firm grasp on DB tuning, perl, Mason and RT's architecture and get your hands very dirty - or you pay a hefty sum to Best Practical for support.
Almost. What you're describing is full "VMware Infrastructure". ESX is the bare-metal hypervisor - and has actually been "replaced" with ESXi. This is a stripped down version that has a smaller install footprint and therefore a lower exposure to exploits. Most of the old ESX patches were for things like Samba and CUPS on the service console. ESXi now comes from vendors like Dell in an embedded form even: 32MB on an SD card, pre-installed, no hard drive required.
You can use local storage with ESX and ESXi; just format it with VMFS. When you're dealing with live-migration (vmotion), automated resource balancing (DRS) and bringing VMs back up after a bang (HA), that's all part of VMware Infrastructure and Virtual Center.
Very much not free - BUT the ESXi installable hypervisor? Free. Go download it now if you want it. Due to the stripped down nature, it supports a more limited set of hardware than VMware Server (which relies on a regular host OS to work out the hardware details), but it performs much better. Depending on the host hardware and VM workload, you can get a 20:1 VM:host ratio with your eyes closed.
Please, someone shoot this idea down - why not just do it with infrared lights on the sidelines or something similar? All you need is some focused, non-visible light broadcast at field level and you could paint that visible using a chroma-key-like system. Calculating the angles of this and the deflection of the field, etc, etc seems far more complicated than it has to be...
I took great joy in sending an "Teh intarnets is down!!" email a few months ago to our network guy. Exchange servers worked...dedicated line to the main office worked...outbound link from our office? Not so much. :-) He got the email, I still couldn't get to google.com.
Reviews: http://www.guitarpraise.com/reviews.php/
There are many video games that I've reviewed for Plugged In that I wouldn't let my kids--or anyone's kids, if I had such power--get within a country mile of.
"My views are so obviously correct that nobody should be allowed to hold others..."
Nope. That doesn't smack of ignorance or bias.
Not at all. I've been a T-Mobile customer for well over 5 years. They've always had great customer service and retention. A (long) while back I wanted to upgrade from my old v60 to the T610. AT&T was giving away the T610 (T616 actually, same phone though) for free with a new contract. T-Mobile was selling it for $50 to new customers and $100 for existing customers (that's after the new contract discount). I called them out on it, told them how ridiculous that was and that I wanted the T610 for free from them, just like AT&T was offering, or I would simply walk over to AT&T and get it there. Instantly, the rep said "I can't do that - it's in the system at a set price....but I CAN give you the next 3 months of service free, which works out to be a little more than the cost of the phone..."
They're in it to make money just like everybody else and they'll do what they can to get it. The thing I like about them is the fact that they seem to understand that keeping their customers is cheaper than getting new ones - so they actually TRY to treat their customer well.
They have an entire division dedicated to working with phones that they "don't support" (like the China-Mobile firmware Motorola e680i I had for a while). I've had support techs refuse to get off the line until my Linux laptop was GPRS tethering properly off that same "unsupported" phone. Add all that up and it becomes a little bit clear why I've stayed with them for so long.
I got my dates backwards (2003 for my t610); summer of 2004 sounds about right. The phone was never released by a carrier in the States. That said, I bought my unlocked e680 from a company in Florida. T-Mobile FULLY supported me twice over the phone with that handset. Once for getting MMS settings working on their network; the other time for getting my Linux laptop working over bluetooth DUN w/ GPRS ("unsupported phone" *division* and someone willing to work through a problem with an "unsupported" OS - yeah, I'm a customer for life).
:-)
I never got a terminal on the phone myself; it was just a phone with more features in a smaller package than anything else at the time to me. From what I remember from a few motodev and linux geek forums, people were able to crack it open and get a terminal eventually. It had no wifi and no rj45, so that would have been a limited use for me anyway. I did have an ssh client for it, though.
Regardless, the e680 wasn't even the first Linux based phone on the market as I recall.
(oh wait, it's the *only* *nix based phone on the market - that actually works)
:-/) camera, video capture, fm tuner and an SD card slot...
My Linux-based Motorola e680 from 2003 would beg to differ. It worked wonderfully, thank you. Full touch screen, minimal buttons (keypad was in the touch screen), mp3 audio, mp4 video playback (this all sound familiar?)....oh yeah and a (vga
As a former waiter / bartender, the size of the bill is *usually* related to the length of time for the transaction. Leaving a dollar on that $5 coffee / bagel is fine because the server will be doing a significant number of transactions over places with larger ticket items. A server can complete that entire transaction in less than a minute on its own. If you get multiple, similar requests at or near the same time, the server will also combine these actions; so, while they can do a single coffee in 45 seconds from station to table, they can do 6 coffees in only 90 seconds.
Also, at a diner, you might be able turn a table over every 30 minutes. At a nice, elegant restaurant, you may only turn the table over two, *maybe* three times a night depending on scale. People eating an expensive meal tend to like to sit, talk and enjoy the food rather than getting quick body-fuel. That requires more attention and round-trips to make sure everything is going well, drinks are filled, etc.
Waiting tables is a Hell job. Long hours on your feet (I used to do 6 doubles@16+ hours each and a dinner on the 7th day) and being barked at by inconsiderate jackasses make it a largely thankless job. If you can get things as efficient as possible, your customers will appreciate it and tip in kind. If you just try to get through the hours, it will show and reflect in your tips. Those servers that just try to grind through without regard to their customers don't deserve 15%. Their choice to work in a customer-oriented business while note caring about said customers does not require me to subsidize their life. Remeber TIP = To Insure Prompt/Proper Service.
That said, I regularly earned 20% average per shift; on everything from $10 lunches to $500 two person dinners. I tip the same when I am being served. Good service to me will get you a minimum 20%. With supremely excellent service, I typically ignore the percentage completely and have been know to leave well over 100%.
Any place offering free wifi should have a person that will swing through the dinning room to ask if people want anything else. Put up a sign that says "complementary wifi with service" or some such. Make it an honor system and police it when you're not totally slammed.