My personal best with wireless was about 20Mb/s of UDP in one direction with iperf. Trying to xmit/rcv at the same time cut it down further, using TCP even more, etc. This is totally not a scientific result, just me playing around one day wondering how fast I could get wireless to go.
I can't say I've hit a bug in MySQL replication in years. It's not like I'm still trying to use 3.23. I'm not saying there aren't any bugs, just none that I've come across personally.
Can't speak for the GP but I've purchased a lot of cables from monoprice: ethernet, component, s-video, audio, HDMI, etc. I always test new ethernet cables on my Fluke Cable IQ before putting them into production; no faults in anything I've received thus far. I've been happy with every cable I've ordered.
The problem I have is with placing small orders because they're so darn cheap you end up paying more in shipping than the cable. Solution? I always try to order mass quantities.
Plenum rated cable is used in air handling spaces because it won't give off toxic gas if it burns/melts. It doesn't mean the quality is different, just the jacket is a different material. The wire inside is the same as non-plenum rated cable.
FYI "air handling space" includes spaces above drop ceilings if the building doesn't have ducted returns everywhere.
Personally, I wouldn't classify permanent install cable inside walls and going into an IDC connector on a wall plate or patch panel as "custom made". It's just "the right way."
I used to make cables in my younger days when I thought it was cool, but these days I find it's more convenient just to buy mass quantities from monoprice. They're so cheap that even if one does go bad (no problems yet) there's 50 spares waiting in the storage closet.
I see that as a core problem with postgres: too many replication tools that are potentially incompatible. The MySQL slaves I set up years ago have been going along just fine - even though upgrades - because it's part of the core. If I were to switch to postgres I'd have to bolt on some random tool to get basic functions back. What if the one I chose stopped updating out of nowhere and I'm stuck? It kills the postgres benefit of supposedly being more reliable than MySQL unless you're only using one database server. Does anyone know why the postgres team decided to omit replication?
Of course, at the end of the day, the customer doesn't care or even know what's going on behind the scenes as long as it keeps working. Just like how people sit on slashdot and claim MySQL could never operate something as big as that.
I started with MySQL because it had replication. Yeah, I know Postgres has slony or whatever it's called, but I cringe at the though of maintaining disparate pieces of software (that sometimes don't get along) on something as critical as a database. Maybe it's changed, maybe it hasn't, but that's how it was when I chose to use MySQL.
I was referring to SLA "gel cell" batteries. Coincidentally, I caught a set in an APC external battery pack overheating at my parent's house the other day. It took a while get them out of the UPS (the casing was horribly deformed), and they were too hot to handle after initially unplugging them. You could hear them hissing for hours. I can take pictures if you don't believe that an SLA battery can fail.
There were some AGM batteries in the same string that are older than the SLA's that failed and they are still going along just fine. Lesson: buy the expensive set of AGM batteries, they'll save you in the long run.
Amazon has warehouses in Fernley, NV (making free shipping to Reno next day PM since it's 25 miles up the freeway) but they don't put sales tax on my orders.
Though one could argue that T1 costs are absurd too, not to mention that T1s are absolutely shit in terms of bandwidth. 1.5mbit/sec is slower than my CELL PHONE. There is something wrong with paying $400-800 per month for service that is beaten by a $50/mo AT&T 3G plan on a $200 cell phone.
It's not beaten. Users of T1, T3, and lit buildings with Ethernet handoffs never deal with the silly tiers and caps that dominate the race to the bottom pricing customers. No restrictions, no limits, no filtering, no fine print. You get the whole thing all to yourself and a guarantee that you can to saturate it 24x7x365 if you want. You even get a guarantee that it'll work within a certain performance level.
For anyone in a lit building and can get Ethernet over SONET, it's really the way to go. Is it true that the cost of a T1 and T3 are absurd, although there's a whole world of legacy telco tariffs out there. It becomes obvious when you get two quotes for a lit building: Ethernet or DS3 handoff. The only difference is the card the carrier sticks in the SONET gear for your channel, but the price for the DS3 loop is typically more than double. Even your typical T1 doesn't go very far anymore before it ends up in a fiber ring. However, the pricing does help keep the typical "I need help with the power button" jackoffs away.
But if one is convinced a $50 AT&T 3G plan on a cell phone can beat a SONET delivered circuit, well, it's not for everyone.
"Sealed" just means that it's not a "wet" battery and doesn't need to be operated upright, doesn't spill, don't need to check the electrolyte level, etc. With enough internal pressure a failed SLA will vent/drool acid everywhere in an effort to prevent a catastrophic failure.
Well, Google is not about redundancy or uptime or retention. They have mass quantities of these things in such a manner that it just doesn't matter how many fail per day. And besides, it's search. If a search failed or didn't return useful results because the useful data got eaten by failed servers I bet most people don't even think twice about it.
Fear? What the hell? It's well known that infrastructure collapsing bugs are frequently introduced. Some trains of IOS have a horrible reputation depending on your platform. And playing in T train land? Good luck with that game of Russian roulette.
The whole "store card" thing never really did appeal to me. It's annoying to be asked everywhere to sign up for their cards, but it would probably be more so to actually juggle 20 different store cards and the bills that go with them.
I routinely give away full months of free service to my customers if they have a problem - even for tiny things like management interface issues that prevent them from doing something they should have. For major issues, the only real way to compensate a customer is to give them x amount of their money back. Google's "you may get 15 days" SLA is very, very weak. Purely my opinion; obviously many people think it's the most awesome generous thing a company could ever do. If you look around though, it's bottom of the barrel.
Everyone but Apple has a 4-hour same day on-site or better service option when you buy from them. I've seen them show up in under 30 minutes with new parts. Hell, they usually send you everything "just in case" what you though was the problem was wrong.
I once has a power supply fail in a Dell rack server. No biggie, fired up their online chat thing and said the PSU was dead. They sent me a replacement PSU, motherboard, CPU, backplane, risers, everything. They just said start with the PSU and keep replacing stuff until it works. Return unused/replaced parts within 30 days. It wasn't even the "gold" support or whatever, it was a next-day parts only contract (non-critical server).
It looks like ma.gnolia was a "yay, I like Macs, Apple is awesome lolomg!!!1!!11" operation run by someone who didn't know what he was doing, probably backed up by Time Machine. Tough cookies for them.
I wouldn't lump all outside vendors in the same boat as Google and their "even if you pay us the only recourse you get is up to 15 free days at the end of your term" SLA.
If any of my customers asks for their raw maildir tree, sure, no problem, where do you want me to send it to? Want it on a CD/DVD? A little more work for a price, but I can do that too. (Encrypted, of course, in case the mail gnomes take it.) I figure it's their data anyway so I don't see a problem with giving it to them for backup purposes. You just have to make the right choice in who trust. Even then, don't trust and make sure you have some kind of backup you can see (like the aforementioned DVD of a maildir in your hands that you can prove to yourself it works) for the stuff that really matters to you.
And yeah, free is usually worth exactly how much you paid for it. Although everyone expects the internet to be free and get the same quality of service as if you were paying a million a month for it.
With a transactional database who cares how long it takes - the state isn't going to change. If you're backing up your 500GB MyISAM tables, well, you're asking for trouble. Since you mention MySQL, use innodb tables with the dump option "--single-transaction".
Slashdot Mirror
Don't forget register globals. Worst option ever.
My personal best with wireless was about 20Mb/s of UDP in one direction with iperf. Trying to xmit/rcv at the same time cut it down further, using TCP even more, etc. This is totally not a scientific result, just me playing around one day wondering how fast I could get wireless to go.
I can't say I've hit a bug in MySQL replication in years. It's not like I'm still trying to use 3.23. I'm not saying there aren't any bugs, just none that I've come across personally.
Can't speak for the GP but I've purchased a lot of cables from monoprice: ethernet, component, s-video, audio, HDMI, etc. I always test new ethernet cables on my Fluke Cable IQ before putting them into production; no faults in anything I've received thus far. I've been happy with every cable I've ordered.
The problem I have is with placing small orders because they're so darn cheap you end up paying more in shipping than the cable. Solution? I always try to order mass quantities.
Plenum rated cable is used in air handling spaces because it won't give off toxic gas if it burns/melts. It doesn't mean the quality is different, just the jacket is a different material. The wire inside is the same as non-plenum rated cable.
FYI "air handling space" includes spaces above drop ceilings if the building doesn't have ducted returns everywhere.
Personally, I wouldn't classify permanent install cable inside walls and going into an IDC connector on a wall plate or patch panel as "custom made". It's just "the right way."
Gigabit Ethernet uses all four pairs.
I used to make cables in my younger days when I thought it was cool, but these days I find it's more convenient just to buy mass quantities from monoprice. They're so cheap that even if one does go bad (no problems yet) there's 50 spares waiting in the storage closet.
I see that as a core problem with postgres: too many replication tools that are potentially incompatible. The MySQL slaves I set up years ago have been going along just fine - even though upgrades - because it's part of the core. If I were to switch to postgres I'd have to bolt on some random tool to get basic functions back. What if the one I chose stopped updating out of nowhere and I'm stuck? It kills the postgres benefit of supposedly being more reliable than MySQL unless you're only using one database server. Does anyone know why the postgres team decided to omit replication?
Of course, at the end of the day, the customer doesn't care or even know what's going on behind the scenes as long as it keeps working. Just like how people sit on slashdot and claim MySQL could never operate something as big as that.
I started with MySQL because it had replication. Yeah, I know Postgres has slony or whatever it's called, but I cringe at the though of maintaining disparate pieces of software (that sometimes don't get along) on something as critical as a database. Maybe it's changed, maybe it hasn't, but that's how it was when I chose to use MySQL.
I was referring to SLA "gel cell" batteries. Coincidentally, I caught a set in an APC external battery pack overheating at my parent's house the other day. It took a while get them out of the UPS (the casing was horribly deformed), and they were too hot to handle after initially unplugging them. You could hear them hissing for hours. I can take pictures if you don't believe that an SLA battery can fail.
There were some AGM batteries in the same string that are older than the SLA's that failed and they are still going along just fine. Lesson: buy the expensive set of AGM batteries, they'll save you in the long run.
Amazon has warehouses in Fernley, NV (making free shipping to Reno next day PM since it's 25 miles up the freeway) but they don't put sales tax on my orders.
Though one could argue that T1 costs are absurd too, not to mention that T1s are absolutely shit in terms of bandwidth. 1.5mbit/sec is slower than my CELL PHONE. There is something wrong with paying $400-800 per month for service that is beaten by a $50/mo AT&T 3G plan on a $200 cell phone.
It's not beaten. Users of T1, T3, and lit buildings with Ethernet handoffs never deal with the silly tiers and caps that dominate the race to the bottom pricing customers. No restrictions, no limits, no filtering, no fine print. You get the whole thing all to yourself and a guarantee that you can to saturate it 24x7x365 if you want. You even get a guarantee that it'll work within a certain performance level.
For anyone in a lit building and can get Ethernet over SONET, it's really the way to go. Is it true that the cost of a T1 and T3 are absurd, although there's a whole world of legacy telco tariffs out there. It becomes obvious when you get two quotes for a lit building: Ethernet or DS3 handoff. The only difference is the card the carrier sticks in the SONET gear for your channel, but the price for the DS3 loop is typically more than double. Even your typical T1 doesn't go very far anymore before it ends up in a fiber ring. However, the pricing does help keep the typical "I need help with the power button" jackoffs away.
But if one is convinced a $50 AT&T 3G plan on a cell phone can beat a SONET delivered circuit, well, it's not for everyone.
"Sealed" just means that it's not a "wet" battery and doesn't need to be operated upright, doesn't spill, don't need to check the electrolyte level, etc. With enough internal pressure a failed SLA will vent/drool acid everywhere in an effort to prevent a catastrophic failure.
http://en.wikipedia.org/wiki/VRLA
Well, Google is not about redundancy or uptime or retention. They have mass quantities of these things in such a manner that it just doesn't matter how many fail per day. And besides, it's search. If a search failed or didn't return useful results because the useful data got eaten by failed servers I bet most people don't even think twice about it.
Fear? What the hell? It's well known that infrastructure collapsing bugs are frequently introduced. Some trains of IOS have a horrible reputation depending on your platform. And playing in T train land? Good luck with that game of Russian roulette.
What does 11.2 cents/kwh compare to in your area? That's what natural gas and coal power from the grid costs where I live.
The whole "store card" thing never really did appeal to me. It's annoying to be asked everywhere to sign up for their cards, but it would probably be more so to actually juggle 20 different store cards and the bills that go with them.
I routinely give away full months of free service to my customers if they have a problem - even for tiny things like management interface issues that prevent them from doing something they should have. For major issues, the only real way to compensate a customer is to give them x amount of their money back. Google's "you may get 15 days" SLA is very, very weak. Purely my opinion; obviously many people think it's the most awesome generous thing a company could ever do. If you look around though, it's bottom of the barrel.
Everyone but Apple has a 4-hour same day on-site or better service option when you buy from them. I've seen them show up in under 30 minutes with new parts. Hell, they usually send you everything "just in case" what you though was the problem was wrong.
I once has a power supply fail in a Dell rack server. No biggie, fired up their online chat thing and said the PSU was dead. They sent me a replacement PSU, motherboard, CPU, backplane, risers, everything. They just said start with the PSU and keep replacing stuff until it works. Return unused/replaced parts within 30 days. It wasn't even the "gold" support or whatever, it was a next-day parts only contract (non-critical server).
It looks like ma.gnolia was a "yay, I like Macs, Apple is awesome lolomg!!!1!!11" operation run by someone who didn't know what he was doing, probably backed up by Time Machine. Tough cookies for them.
I wouldn't lump all outside vendors in the same boat as Google and their "even if you pay us the only recourse you get is up to 15 free days at the end of your term" SLA.
If any of my customers asks for their raw maildir tree, sure, no problem, where do you want me to send it to? Want it on a CD/DVD? A little more work for a price, but I can do that too. (Encrypted, of course, in case the mail gnomes take it.) I figure it's their data anyway so I don't see a problem with giving it to them for backup purposes. You just have to make the right choice in who trust. Even then, don't trust and make sure you have some kind of backup you can see (like the aforementioned DVD of a maildir in your hands that you can prove to yourself it works) for the stuff that really matters to you.
And yeah, free is usually worth exactly how much you paid for it. Although everyone expects the internet to be free and get the same quality of service as if you were paying a million a month for it.
With a transactional database who cares how long it takes - the state isn't going to change. If you're backing up your 500GB MyISAM tables, well, you're asking for trouble. Since you mention MySQL, use innodb tables with the dump option "--single-transaction".
If you have ductwork you can always fish plenum-rated cabling through the ducts.
Next up is the Mars Science Laboratory. Bigger, better, and nuclear powered with radioisotope thermoelectric generators.
P.S. cleaning a solar panel is hard. Did you really think the designers overlooked it?
The next one will be; the Mars Science Laboratory will use radioisotope thermoelectric generators.