MTGOX so let me get this straight....

[smoothie]

You halt trading for 12 hours to "improve" your trading engine and to allow the market to "cool down"...

You bring the trading functions back up and you still lag like shit...

So where is the improvement and point of the halting?

[1713558755]

1713558755

[1713558755]

1713558755

[ShireSilver]

Why would you expect them to NOT lag when everyone who is at all interested will be picking this one time to be going to their site. Please have some frakkin sense.

[aantonop]

Why would you expect them to NOT lag when everyone who is at all interested will be picking this one time to be going to their site. Please have some frakkin sense.

Really? So they didn't increase their scale to at least the peak volume of the last week?

Oh wait, they're about 50% of peak and still lagging.

Never mind, keep makign excuses, they're laughing all the way to the bank.

[justusranvier]

I've never been an engineer in IT, but I did manage the maintenance department of a manufacturing facility where the cost of unexpected equipment downtime ran into the tens of thousands of dollars per hour.

Sometimes we had to wait for months in order for a particularly critical machine to be idle so that we could do preventative maintenance/repairs/upgrades. For all the critical equipment we had lists of all the things we wanted to do the next time we got an opportunity. We bought all the material well ahead of time, and double and triple checked that everything was ready to go so that we could be in and out as quickly as possible, with as much margin as possible for unexpected difficulties.

The implementation details differ here, but essentially that's how this upgrade should have gone. These are known scalability problems existed for at least two years. There's no excuse for why they weren't able to put together a plan a long time ago so they'd be ready to execute it in the 12 hours they were down.

[ShireSilver]

Why would you expect them to NOT lag when everyone who is at all interested will be picking this one time to be going to their site. Please have some frakkin sense.

Really? So they didn't increase their scale to at least the peak volume of the last week?

Oh wait, they're about 50% of peak and still lagging.

Never mind, keep makign excuses, they're laughing all the way to the bank.

So they should have known a month or two ago what the peak volume last week would have been?

Sure, they should have expected a mad rush of basically everyone who has ever been interested in bitcoins to be hitting their server all at the same time, and maybe they did. Maybe they calculated that since the peak they were going to get at the anointed time was going to be a one time event that they would only need to survive through and that regular usage would be much lower. Maybe they figured they didn't need to spend tons of money for the expected one time event when it would be a waste after that. There's a whole heck of a lot of decisions that have to be made when actually running a business that outsiders would never suspect.

Why not have a little damn patience and give them an hour at least before complaining about it. Heck, they probably needed at least an hour's worth of new data just to have enough information to fine tune the database.

[jasinlee]

We simply expect a company that makes millions of dollars to be capable of hiring 1 person who knows how to setup additional servers to handle the extra load. I bet they could pay at least 300 people on this forum to fly out there, PURCHASE AND INSTALL new servers within a week. But this has been ongoing since.......wait was there ever a time this wasnt an issue?

[crypTrade]

My conspricay theory, based on no facts or special knowledge but sheer speculation is this -- gox wasn't holding as many bitcoins as they should have been and were worried that they would have to buy bitcoins at $200+ each to fund withdrawls. (maybe they were planning to buy them when the rally ended but it never did and they got in dfeeper and deeper) So they staged a problem, crashed the price, bought up all the bitcoins they needed, and then resumed business as usual. They hope hte volatility will actually bring more speculators, becuase the more the market moves, the more opportunities for things like day trading there are, and nohitng makes gox more money than people who trade a lot.

[aantonop]

So they should have known a month or two ago what the peak volume last week would have been?

Sure, they should have expected a mad rush of basically everyone who has ever been interested in bitcoins to be hitting their server all at the same time, and maybe they did. Maybe they calculated that since the peak they were going to get at the anointed time was going to be a one time event that they would only need to survive through and that regular usage would be much lower. Maybe they figured they didn't need to spend tons of money for the expected one time event when it would be a waste after that. There's a whole heck of a lot of decisions that have to be made when actually running a business that outsiders would never suspect.

Why not have a little damn patience and give them an hour at least before complaining about it. Heck, they probably needed at least an hour's worth of new data just to have enough information to fine tune the database.

Um, yeah they shoud've known. Any half decent capacity planning would include flash-crash and 10x volume peaks. It would include a contingency plan for halting trading.

Do you work for them?


They've had 2 years to build a scalable architecture and are still showing signs of extreme amateurism.

I've actually managed data centers and have run businesses so don't patronize me like you know me. I've worked with financial services IT for about 18 years now.

"Fine tune the database"? What the hell is this, an MS-Access app?

you don't build a single stack architecture and expect it to scale. It won't. They have no clue. I have no patience with incompetence and amateurism when $2 billion economy is riding on top of it.

[repentance]

We simply expect a company that makes millions of dollars to be capable of hiring 1 person who knows how to setup additional servers to handle the extra load. I bet they could pay at least 300 people on this forum to fly out there, PURCHASE AND INSTALL new servers within a week. But this has been ongoing since.......wait was there ever a time this wasnt an issue?

Unfortunately, it's going to take a lot more than just installing some new servers to solve this problem.

[jasinlee]

We simply expect a company that makes millions of dollars to be capable of hiring 1 person who knows how to setup additional servers to handle the extra load. I bet they could pay at least 300 people on this forum to fly out there, PURCHASE AND INSTALL new servers within a week. But this has been ongoing since.......wait was there ever a time this wasnt an issue?

Unfortunately, it's going to take a lot more than just installing some new servers to solve this problem.

Installing competence is extremely difficult you do have a point.

[ShireSilver]

I've actually managed data centers and have run businesses so don't patronize me like you know me. I've worked with financial services IT for about 18 years now.

"Fine tune the database"? What the hell is this, an MS-Access app?

LOL, now I know you're nothing more than a troll. Anyone with experience in financial services IT would know that databases have a lot of configuration that needs to be messed with. Indexes can be added, tables can be pegged to memory, different table engines can make a difference; but these need data to be analyzed to determine if they will actually help. Are they using stored procs, and if so would recent changes in data usage mean they should be rewritten?

[aantonop]

I've actually managed data centers and have run businesses so don't patronize me like you know me. I've worked with financial services IT for about 18 years now.

"Fine tune the database"? What the hell is this, an MS-Access app?

LOL, now I know you're nothing more than a troll. Anyone with experience in financial services IT would know that databases have a lot of configuration that needs to be messed with. Indexes can be added, tables can be pegged to memory, different table engines can make a difference; but these need data to be analyzed to determine if they will actually help. Are they using stored procs, and if so would recent changes in data usage mean they should be rewritten?

You know, I didn't call you names.

If you think they can solve architectural scalability problems with fine tuning of a single instance database, you are wrong.

Trading Exchanges are not run on single instance databases. That is simply a fact. They are run by thousands upon thousands of distributed systems with multiple tiers of redundancy. Market data and reporting are completely separate from trading engines. Front-end stuff never even comes near the trading engine.

Oh, by the way: Fuck you.  I only criticized GOX based on the multi-year operations. You're making excuses for a bunch of idiots and making assumptions about people you don't know, based on a few words in a forum.

[superduh]

mtgox down again.
INSANITY

[Sage]

DECENTRALIZED EXCHANGE!

F$%&, how load does it have to be shouted!

All central exchanges are the constraint for Bitcoins growth.

Unprecedented exponential growth is ahead for Bitcoin.  No central architecture can handle it.

Imagine the demand when Germany dumps the Euro and brings back their own currency.  That event will happen unexpected.  It would kill any centralized architecture period.

We need a decentralized exchange now!

[dutt]

Gox logic: Lag brings down prices, let's lag it for 12 hours!

[superduh]

Gox logic: Lag brings down prices, let's lag it for 12 hours!

gox down

[repentance]

[ShireSilver]

If you think they can solve architectural scalability problems with fine tuning of a single instance database, you are wrong.

Trading Exchanges are not run on single instance databases. That is simply a fact. They are run by thousands upon thousands of distributed systems with multiple tiers of redundancy. Market data and reporting are completely separate from trading engines. Front-end stuff never even comes near the trading engine.

Oh, by the way: Fuck you.  I only criticized GOX based on the multi-year operations. You're making excuses for a bunch of idiots and making assumptions about people you don't know, based on a few words in a forum.

I didn't say that you can solve everything with just database tuning, but did imply that it was at least one of the things they might be doing. There's a whole lot of speculation going on here, and except for Gox mentioning that they were adding servers and moving their database off of the front end server, that's all it is - rampant and mostly pointless speculation. You're spewing condemnations when you don't really know what's going on. I don't know either, but as I have some experience in IT I can imagine all kinds of things that they might be doing to try to improve things, and I can imagine all kinds of things that could be preventing them from getting things straightened out quickly. I have some patience.

And no, I'm not affiliated with them in any way. I don't even have an account on any of the currency exchanges. I just really hate people bashing repeatedly with wild speculations when its a waste of time. Make your case, state it to the world once clearly, quit using their service, and get the fuck over it and move on. But you don't need to fill up these threads with repeated regurgitations of your pointless accusations.

[justusranvier]

and moving their database off of the front end server

They didn't do that yet according to an IRC chat log with MagicalTux.

[Maged]

and moving their database off of the front end server

They didn't do that yet according to an IRC chat log with MagicalTux.

That will be in done in the new trading engine, due out sometime in the next month.