Hacker News new | past | comments | ask | show | jobs | submit login
Bloomberg Terminals Suffer Widespread Failures (nytimes.com)
132 points by svtrent on Apr 17, 2015 | hide | past | web | favorite | 66 comments

The outage isn't that bad. I've talked to 5 other firms in Canada this morning and haven't heard of a single one that is down.

Bloomberg's tickering plant is still working and their emsx(trading system) is up and running, albeit a bit slow right now.

I think it was on a stackoverflow podcast that Joel or Jeff said that it makes you wonder about people who complain about down time if they've ever built something of their own.

Google's been down, ditto for facebook, twitter and Azure. If you build a big service, something will eventually fail, Bloomberg is no different, if anything they are much better than Reuters or CapIQ for availability.

I saw this quote from another comment here:

> but the widespread reliance on the $10k+ per year terminal in the financial services industry means that this thing is ripe for disruption.

I completely disagree, Bloomberg isn't some slow lethargic big company. They iterate relentlessly, their help desk is the best I've ever had the pleasure of dealing with.

If I've got a problem with the trading system, their first level help will get me to a technical specialist in that area in under 30 seconds. If they can't solve it then I'll talk to a developer in 30 minutes.

Bloomberg is the leader in teh financial space, not only because their tech is the best in teh business, but because they listen to their customer better than their competitors, they release more frequently, and when they want to enter an area( say risk management) they go out and buy someone who is already a leader in that field.

In short they never stop running, if anything they are increasing the distance between them and their "peers".

As one example, they just posted a message to every Bloomberg terminal user telling us that the issue was an internal networking issue that has been solved.

EDIT to answer this question:

> How do they protect the ticker tape thing from attacks, like say, a DDOS?

2 things:

1) This question really doesn't make sense. How do you DDOS a one way line? They send you quotes for the tickers that you request. You can't really DDOS them, unless you do something like request tickers over and over again in a tight loop, in which case their api will rate limit you.

2) its a closed system. Only terminal users can access it and only through the bloomberg terminal or their api, so they control all the access points.

> Google's been down, ditto for facebook, twitter and Azure

The first three, at least, are free services and availability isn't a priority for many of their customers. Bloomberg users have different requirements.

Boeing 747's are complex systems; it's not acceptable if one fails occasionally.

> Boeing 747's are complex systems; it's not acceptable if one fails occasionally.

Yet they have failed. Reality isn't always acceptable.

Google Apps is not a free service. The pricing for business is $5 or $10 per user per month. It's worth bearing in mind that when people complain about the service, they are often paying for it.

> Boeing 747's are complex systems; it's not acceptable if one fails occasionally.

You're right that 747s should not just fall out of the sky. However, parts of 747s fail constantly. In a machine that complex, they are guaranteed to have some downtime. A well manufactured 747 will fail cleanly and safely and recover smoothly.

I think comparing the acceptability of failure of a stock information system and a commercial aircraft is a bit unreasonable.

It was pretty widespread in the (European) morning

seemed completely and utterly down in London across a good sample of clients.

Presumably many institutions would have other sources of pricing data too - I know this is the case in many sell-sides, not sure about everywhere else. I've worked in a few sell-sides and you see bloomberg screens all over the place, but usually alongside front office trading systems like Fidessa which are powered by direct feeds a lot of the time.

Probably a much bigger issue for everyone who doesn't have exchange connectivity, but you'd also hope they have some sort of fallback for exactly this scenario. I.e. reuters or something.

Yes, many banks would failover to Reuters IDN_SELECTFEED

WRT 2) DOS does not require a public network; you can war dial access points or attack infrastructure to achieve the same effect.

Any insight on what happened today? I'm surprised there's no failover mechanism that tries to route traffic some other way in case of a communication failure.

Unfortunately BBG requires a bit more scrutiny because when their network goes down entire countries cannot finance themselves -- they had to postpone some debt sales in the UK today I read.

> it makes you wonder about people who complain about down time

HN in particular seems to have a "_____ is down" on the front page every few days. It's always filled with comments like this guy[1] saying how their architecture is clearly broken or how any fifteen year old could write something better.

[1] https://news.ycombinator.com/item?id=9394071

hn was never built to be reliable, and has always had reliability issues.. it used to (still is?) run from a single server.

The outage affected Europe and Asia during their trading day, which is a very big deal, wouldn't have expected Canada to see anything since they are 5-6 hours behind.

Bloomberg runs some of the most reliable technology in the world, and is one of the best at running distributed architecture, so the fact that a problem like this wasn't more contained is surprising, if not shocking.

How do they protect the ticker tape thing from attacks, like say, a DDOS?

If the data is mission critical to your company, you consume the feed via private leased lines or interconnects.

The same way that ALL important systems should be protected: by not connecting them to the Internet in the first place.

"Service has been fully restored. We experienced a combination of hardware and software failures in the network, which caused an excessive volume of network traffic. This led to customer disconnections as a result of the machines being overwhelmed. We discovered the root cause quickly, isolated the faulty hardware, and restarted the software. We are reviewing our multiple redundant systems, which failed to prevent this disruption."

https://news.ycombinator.com/item?id=9393096 previous discussion from earlier today.

LOL. Maybe it was a can of coke???

   But reports from inside the company suggested
   that a spilled can of Coke in one of the server
   rooms had been responsible for knocking out
   systems across two continents.

I honestly can't believe some (semi reputable) news sites are spewing this bullshit.

Fuckin please.

Random anecdote: Was told by a trader a few years back that one trading desk had both Bloomberg and Thomson Reuters systems next to one another and once they were able to capitalise on Bloomberg being a few seconds faster to make a net gain.

This is really not possible. If there's any truth to this story at all you have forgotten the real version.

Having a Reuters and Bloomberg terminal on the desk next to each other is commonplace, I did have, plus a Datastream.

As for a few seconds delay, that's probably how they're interacting with them, not how they could use them. Both have both very different UIs and very different Excel plugins (assuming someone has both on their desk, from experience, they probably use the Excel plugins a lot/mission critical functionality).

<onion> bloomberg network unable to report crash of major financial information service </onion>

Moderately off-topic (and don't say: "Today? Nowhere."):

If I wanted to play around with a Bloomberg for a few hours in North America, just to see what the fuss is about, how could I go about doing this?

New York Public library (business/finance library) on Madison Ave between 33-34th (NYC) has 2-3 terminals for open public use. Usually a short wait but you can use them.

The Graduate School of Business Library at Stanford used to have a couple of terminals, in case you are in the bay area.

Since we're moderately off-topic: I once made a gnuplot script that emulates the look of a Bloomberg terminal. Super hacky, but it works: http://folk.ntnu.no/asmunder/bloom.gnu and you need this file as well http://folk.ntnu.no/asmunder/fader.png

Bloomberg starts at $1,750 a month

At that level of cost there is an expectation of almost perfect up time and multiple redundancy - the only reason people pay that much is they believe they can get more value than that out of it..

Cost isn't necessarily tied to uptime expectations. It's tied to the value people get from it. Bloomberg is so far ahead of it's competitors that this minor event probably won't lose them any business.

It's rare enough that it got a news report. I'd say that qualifies as "almost perfect."

Decades of crusty code - C++ wrapped around C wrapped around Fortran, home brewed versions of the STL, nih syndrome across the tooling. Actually surprised outages not more common. Still, given their interview process is basically parlour trick questions about pointers and char arrays, it is quite telling!

Bloomberg started out as a network system independent of the Internet. Michael Bloomberg was a network engineer. I'm not sure they ever really switched to proper TCP/IP. This company sure looks like a dinosaur these days. However its hard to imagine a company being able to replace even a chunk of their infrastructure. They are tightly integrated into all institutions and there is no other service which offers a complete coverage of all the world's financial assets.

Yes, we use proper TCP/IP :)

The issue was probably something like a broadcast storm because of defective fault detection on some redundant network kit. That won't stop people blaming it on the fact that the stack isn't built on the latest hipster frameworks.

While I see your point do we have to throw the word "hipster" at everything new?.

I'm not a huge fan of Javascript and it's proliferation of frameworks but some of that stuff is objectively both good and useful.

It's human nature to want to label things. IIRC a while ago I was watching one of those "house hunter" type shows and a husband on there was really annoying. I found a discussion online and people were calling him a "hipster douche". So, perhaps hipster by itself isn't so bad. :)

Back a long time ago the applicable pejorative was "yuppie". I haven't heard that one recently. I suspect that many people reading here would have to look it up in Wikipedia.

Edit: BTW at the time I was partial to calling people "yuppie scum". Sigh. It's so hard to keep up with the changing putdowns! :)

Using the term 'hipster' to refer to the newest and shiniest web framework always felt like a bit of a misnomer to me. A solid portion of hipsterdom is rooted in recycling of the obscure and of the vintage (vinyl, the popularity of the lumber sexual look, etc). It seems like the 'hipster' thing to do would be to write a web framework in PL/I or cobol or something.

New and obscure are all too frequently synonymous.

That said, there might be a market for a web framework in PL/I there is a whole lot of IBM big iron out there :-P

...among other things.

Yes I think to replace them someone needs to create a marketplace for different aspects of Bloomberg and you buy access to specific services and data as you need it.

Would be an amazing business that, supposing you could get buy in from a load of other firms.

Really don't see an ala carte solution being worth it for most firms unless they're really trying to cut down on cost. Integrating and maintaining separate services alone is going to require additional developers. Cost associated with that coupled with quality of service and reliability (outside of this situation) of Bloomberg's current offering makes it a very tough sale. Also, guys who have been using terminals for years are really not going to want to switch.

I'm more thinking "if someone was to compete with Bloomberg how would they even come close to what they offer" may not be practical in reality!

interesting view - what makes Bloomberg brilliant (aside from the support) is that i have, for example, an array of my portfolios that i can access from every function. So if you could create a service that took all of the information that was relevant to you (portfolios, news feeds, models etc) and normalize it to make it interoperable with a whole host of 3rd party systems, you could be on to something!

It's crusty old code because it works and it's tried and true. If you have a backend that works, why mess with it?

Sure, C++/C/Fortran/some home grown template library is what it's built on. What would they gain by rewriting it some node framework of the week or whatever other shiny distraction? If they did do that and that caused downtime, there would be the same naysayers here complaining about that instead.

I swear I should start a consultancy in the control of software complexity because I am seeing this sort of thing everywhere, and the downages and security issues will only continue (or get worse, as we continue to rely more on software)

yep, lets not forget 700mb + executables...

Only 700mb? Maybe back in the perkin elmer days, these days they're up to 4gb, constrained only by the various linkers, file size and address spaces available...

Server-side is very big - in fact, a long time ago the part called simply "the Big" got too big and now there are multiple Bigs.

When I was there (2008, intern), they were in the process of moving server code from Bigs into separate services (BAS's, if I recall).

And this is one reason why I don't like C++. I only hear about executables getting too big in C++ projects.

That equate from what I hear to 128gb+ memory usage!

Is this the first crack in the Hoover Dam that is Bloomberg Terminals? Probably not - but the widespread reliance on the $10k+ per year terminal in the financial services industry means that this thing is ripe for disruption. The key isn't to build a terminal replacement though. fintech startups need to chip away at little bits and pieces - and need to inter-operate as much as possible. Only then will the bbg dominance be reined in.

You can write the software, but where are you going to get the data? There are only a few companies that own it and they're not going to give it away. The data is very costly to collect. That's a lot of the cost of any financial software is data licensing fees.

$10k/terminal/year? Absurd. It is a lot more than that.

As another commentator mentioned, chip away. That is how Bloomberg made it as a company. By being better and more sane at fixed income in a very new market.

Find a niche and chip away at that. Data is expensive. Personally, I think making a niche in something coming to becoming mainstream (like R) which is already there in some boutique areas is where it could be.

But don't quote me. I'm still doing COBOL>

one of the reasons i'm generally Ok with Slack's valuation is that there is a ton of room for a messaging service to offer a service that is secure and auditable at the level that the SEC/FINRA and Banks require. Like the article mentions, if you talk to the people using BBG, it's really the chat functionality that provides the main value for literally everyone, at least in my experience (former front office dev)

I'm unfamiliar with these terminals (have kind of always wanted to play with one), but at first blush, a "secure and auditable" messaging system does not seem that difficult to build.

Is it a legitimate compliance or practical implementation problem, or is it just the network effect?

its also easy to keep it secure when it doesnt use the internet

Network effect.

Its not just the network effect..Bloomberg's messaging system is tightly integrated with the other functionality of the terminal..for example one can execute a trade via a message or get live streaming quotes.

“What I miss is the instant Bloomberg chats, which I rate higher than trading or data feeds. The fact is, Bloomberg connects 100 percent of the Street, and all that human intelligence is what makes markets hum.”

I enjoyed the article and I have squirrelled this quote from a trader away for future use - the need to meet and exchange information is why we used to have stock exchanges of course.

Except this information is apparently being exchanged in a closed, proprietary market.

Probably just a large sugary beverage having its revenge on Mike Bloomberg.


But if you want to cripple the financial system, Bloomberg's a good single point of failure to attack. That and Seamless, which also had a hiccup today.

Amazing how propaganda machine NY Times is fast to cover this story.

That crash was totally not strategic at all. /s

What weird conspiracy theory are you flogging? I can't even make sense of the accusation here.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact