Comments for What's Killing Twitter: Twitter Gets Help Digging its Grave

What's Killing Twitter: Twitter Gets Help Digging its Grave

2008-05-24T21:33:15Z

Our beloved Twitter has been digging its own grave for several months now. Recently, the service has experienced an absurd amount of downtime. There have been numerous posts calling for Twitter to be replaced. With so many problems plaguing Twitter, it seems the company could care less. Here are a few factors that are helping to kill Twitter.

]]> Too Many Twitter Apps

There are thousands of Twitter applications out there. In fact, there's a new one that pops up everyday. Twitter's API is constantly hammered by a humongous crop of Twitter apps with well over 600 requests per second. On top of the growing number of users that use Twitter, I'm sure the majority of us use 2-3 different Twitter apps to provide better functionality that Twitter refuses to provide out the box. The constant hemorrhaging of the Twitter API can put a severe strain on the service. Think of it like the three little pigs. Twitter is one of the first two pigs. If they'd built the service out of bricks, banging the crap out of the API wouldn't' blow the whole house down. Unfortunately, Twitter's architecture seems to be made of straws.

Lack Of Communication From Twitter

Communication is important for any business. Users appreciate being informed about why a service they use isn't working the way it should be, regardless of whether the service is free or paid for. Twitter is no exception to this. How many times has Twitter informed you of why it may have been down? I can only count on one hand the number of updates I've seen from Twitter about their downtime. Twitter users are up in arms with the lack of communication from Twitter about the service's problems. Up until recently, Twitter in no way, shape, or form provided reasonable information to users as to what was wrong with Twitter. Come to find out, the company doesn't know what's wrong with Twitter.

Architecture and Scaling

It seems the Twitter team was not prepared for the exponential growth that Twitter has and is still achieving. There has been no easy way to scale Twitter though numerous ideas are floating around as to how. We've even discussed Twitter's architectural problems before in "Twitter and the Architectural Challenges of Life Streaming Applications". Alex Payne, one of the companies software engineers, has also gone public with a blog post about the frustrating problems Twitter has had with scaling Twitter based in the current architecture that the service is using. Payne also notes the following:

We've made progress, and we're more scalable than we were a year ago, but we're not yet reliably horizontally scalable. Why? Because there are significant portions of our system that need to be rewritten to meet that goal.

Twitter's architecture was not made for a messaging system, which is what Twitter has become. Instead, the architecture was structured as a content management system. This plays a huge part with Twitter's scalability issues. However, it seems odd that they didn't create the architecture for mass messaging since Twitter is essentially a messaging system for the masses.

Many Solutions, Not Enough Being Done

Even with the numerous solutions floating around, not enough is being done to stop Twitter from digging its own grave. Payne states that they intend to replace the existing system component by component. However, the will take too long for Twitter users and possibly result in even more downtime. With numerous other factors to consider, it's more than clear that Twitter is digging its own grave with tons of help from both inside and outside the company. The question then becomes: where will you go if or when Twitter dies?

Comment from Dave Sanders on 2008-05-24

2008-05-24T22:01:49Z

Like Web 1.0 went through its own issues with load balancing, multiple server farms, lack of redundancy in server farms and more, Web 2.0 is going to go through and find and fix all of these new problems. Whenever you change the (ugh) "paradigm" you change the problem and the environment and have to find all the new bugs.

This shouldn't really be surprising that twitter, one of the more popular Web 2.0 applications, is the first to really experience these things in a major and very public way. Couple that with a small team (which I don't know if they have any previous experience in high load architectures) and an unproven in high-load underlying infrastructure (Ruby on Rails) and you have the perfect storm.

But, like all things, the team will learn from this experience and all of the bugs will get worked out over time. And their pain will be a blessing to the hundreds of other teams out there over the next few years, as we all learn how to build better interfaces between services. Kudos to them for keeping it up - especially for a "free" application.

Comment from The Masked Millionaire on 2008-05-24

2008-05-24T23:12:16Z

I wonder why all of the people that relentlessly tout twitter never mention that the system goes down all of the time?

They also fail to mention that their "strong love" for twitter is based on a fantasy in their heads and not on communication with the company.

I don't use twitter becasue of the the hyper hype that surrounds the company. It just makes me want to stay away.

But, I am pro business and hope they can figure out a way to make the company work, become profitable, and then sell the whole shooting match to Google.

Live From Las Vegas
The Masked Millionaire

Comment from rick on 2008-05-24

2008-05-24T23:15:03Z

"Twitter's architecture was not made for a messaging system, which is what Twitter has become. Instead, the architecture was structured as a content management system."

I've heard this said before and it makes NO sense. Twitter is an application that asks you a simple question (What are you doing?) and gives you 140 characters to answer. It allows people to subscribe to your answers. What else IS that BUT a messaging system in purpose? So why on earth would you architect the system supporting that product description as a CMS vs a messaging system.

I appreciate that the solution to scaling Twitter and similar apps is decidedly non-trivial, but it's been, what, 2 years since they were discovered at SXSW? Precisely how long should they get before we call this what it is - a nice little app on an architecture that can't possible support what people are doing with it.

I don't know if Pownce or Jaiku could either... but I'd like to see RWW and other prominent blogs start talking about Twitter alternatives more and quit acting like Twitter is the only thing in town. One of the faults of the tech blogs is that you guys anoint one app in a hot category and the others fall by the wayside... whether deservedly or not.

Comment from Allen Stern on 2008-05-24

2008-05-24T23:17:13Z

Corvida - it's nice to see you discuss Twitter's problems without talking about the valley darling of the day - FF. It seems like some of your partner writers have gone from "twitter = g-d" to "ff = g-d" and it's really refreshing to see a discussion of twitter's issues without any other services coming into play.

I've said all along that they opened up too soon. Instead of building a robust infrastructure they said "we have 2 tires and 2 flat ones but hey, cmon jump on anyway!"

Comment from Akhmad Fathonih on 2008-05-24

2008-05-25T01:54:49Z

@rick
agreed. Some part was internet media fault for making twitter the king of publicity :D.

Like FF was said to be the Google of lifestream, twitter would be the Google of web messaging. Multiply that with who's joining on the platform. It will likely cast other player to oblivion ;)

Comment from Andy DeSoto on 2008-05-24

2008-05-25T02:31:22Z

It's really getting tiring reading these negative articles. Normally ReadWriteWeb has a great optimistic perspective that makes me glad to be part of the web's frontier. Not so much with this Twitter debate lately.

Comment from Rob on 2008-05-24

2008-05-25T02:48:13Z

@Andy Desoto - I agree completely. I didn't expect such a negative post from ReadWriteWeb. It feels like a way overstatement of the case just for the sake of having something to write. I think the Twitter community as a whole understands that Twitter is having some scaling problems. Yes, it's frustrating, but overall I think we like the Twitter service enough to stand behind them and encourage them instead of blasting them and calling for their death. If you go read the comments on the Twitter blog, you'll see that most of them are supporting Twitter. Let's keep this all in perspective and cheer on our beloved Twitter team as they work through this. They are a startup after all, not a cold giant corporation.

Comment from atomic1fire on 2008-05-24

2008-05-25T03:07:51Z

What twitter needs is a buyout
some company that can purchase it, give it the backend it needs, and monetize it in a way that wont detract its users

Comment from Luigi Montanez on 2008-05-24

2008-05-25T03:27:49Z

Twitter was designed to be a microblogging platform, not a messaging platform. That's why it was designed as a content management system, because it was supposed to manage content (in 140 character chunks).

When adoption exploded early this year, that's when we realized that it's really a messaging platform. RWW called it a microblogging platform here:

http://www.readwriteweb.com/archives/10_micro-blogging_tools_compared.php

It really irks me when pundits who had no involvement in the technical development of Twitter think that they're qualified to second guess the technical decisions the dozen or so Twitter employees have made in the past year and a half. Unless you work on a project, you have no idea what goes into it, and you shouldn't act like you do.

Comment from Corvida on 2008-05-24

2008-05-25T04:23:16Z

Luigi, while there's no denying that we've stated that it is a microblogging tool, I was paraphrasing from the Twitter dev blog itself which stated that the direction of Twitter is taking on that of a messaging system.

@Andy and Rob We will definitely make an effort to be more positive about Twitter, however the frustrations of the service and the lack of communication (up until recently) have been profusely vented on the web. Maybe we should write a post about how we can help Twitter out? Any suggestions on that?

@Rick We will look into covering Twitter alternatives.

Thank you all for your feedback

Comment from spuds on 2008-05-24

2008-05-25T04:28:02Z

Perhaps it is time one of the big boys took over twitted. For example would we be having all these problems if google owned it.

Comment from Noah on 2008-05-24

2008-05-25T04:50:17Z

"Where will you go if or when Twitter dies?"

Brightkite, most likely. It seems to closest in spirit to Twitter, in terms to connecting people to each other via web & SMS.

Comment from Al Willis on 2008-05-24

2008-05-25T06:13:04Z

First, Twitter is going to be fine. They understand what the problems and bottlenecks are and they're dealing with it. I'm not expecting I'll have to move from Twitter anytime soon. Plus my "social graph" is there and that's not so easy to move.

What the Twitter team is attempting to do--keeping the service running amid database crashes and the like while replacing the key components--is incredibly difficult, especially while tens of thousands of people are using the service.

While Twitter is becoming increasingly important to me personally and professionally, it doesn't bother me (much) when it's down. I'll deal.

Comment from COP on 2008-05-25

2008-05-25T08:17:57Z

MAKE it OPEN SOURCE.... and CROWD SOURCE your problems... ppl ll gladly help yaaa... 4 free... come on..

Comment from Marina Martin on 2008-05-25

2008-05-25T08:21:10Z

I'm not much of a programmer, but if third-party apps are making API calls, then an end-user is generally making those requests, yes? Most of us may end up with 2-3 core Twitter apps, but that doesn't mean people aren't constantly trying others. The sheer number of apps that people have developed is a pretty strong testament to the community that has grown on Twitter, and I'd say that it's a large reason why many people stay with Twitter instead of playing microblogging hopscotch.

When you consider that the Twitter founders also started Blogger, their initial vision of Twitter as a microblogging CMS seems, well, Obvious (forgive the pun).

I don't understand the energy that goes into these "dump on Twitter" posts. It's one thing to devote 100 characters venting on Twitter that Twitter is being slow, and it's another to go through the trouble of writing, editing, and formatting a full-length piece. The Twitter community has done a lot of incredible things, like the Frozen Pea Fund. The community continues to thrive via phone, SMS, email, and yes, FriendFeed chat rooms, while Twitter has its growing pains, but I just don't realistically see a mass migration away from the platform in the near future.

Comment from Sushant Kumar on 2008-05-25

2008-05-25T13:51:07Z

The downtime is very frustrating for frequent twitters. I recently mentioned this on my blog

Comment from sull on 2008-05-25

2008-05-25T17:45:00Z

twitter may decide to apply some limitations to the service while it makes core component replacements over time.
the first place to look are the 3rd party apps that add the most strain on system resources.

some features could be temporarily removed, like rss of public timelines.

they could also take the top 100 active/popular users and segment that data out of the twitter core... a sort of satellite twitter engine to deal with the most intense queries. that is assuming that indeed some of the problems are caused by relational data of these top users (ie. +15k followers/following). this is how web host providers work with the most popular sites on a shared host.

this is of course low hanging fruit on the knowledge tree.
a total migration plan, hardware infrastructure and re-analysis of all back-end technologies as well as the twitter product and what twitter really is NOW as opposed to what it was when it was launched... plenty to figure out. they have funding, community and increasingly brand-recognition. twitter will be here in 2 years, and i cannot imagine it being in worse condition moving forward.
the worst is behind them... every day.

Comment from Andy DeSoto on 2008-05-25

2008-05-25T19:26:53Z

@Corvida Absolutely, that sounds like a great idea. I admit I really have no understanding of how web infrastructure functions, but sure, something more constructive would be great for the entire community, I think! Instead of highly-respected tech bloggers going around saying "we should kill this" or "this is digging its grave," maybe folks can use their expertise to make the internet better for everyone.

Gotta let some of this stuff sink in, maybe I'll comment more later. Thanks for replying!

Comment from rick on 2008-05-25

2008-05-25T19:32:53Z

@Andy, Marina et al...

So are you say RWW and the tech blogs/press should only be positive? That we should only see glowing articles and should sweep problems under the rug? Really?? Because I don't want a tech press that's simply a large PR effort for the industry I want informed, reasoned reporting on what's going on - the good, the bad, the new and cool.

And sorry Luigi, but you don't get to play the "you don't know what's going on so shut up" card. If you want to bask in the warm glow of all of the positive press, you need to expect some brickbats when your service starts seeing repeated downtime. The way to deal with this is known - be as open as you can with what's happening. Fix the issues, yes, but most of us will cut you significant slack if you're transparent. What doesn't work is remaining silent and then lashing out at any criticism.

And I need to somewhat correct my earlier point about Twitter having seen growth for a while now. Yes, they were discovered at SXSW a year and a half ago and took off there. But they HAVE seen a significant increase in traffic to twitter.com in 2008:

That doesn't count API calls either. So I can understand why they're seeing more issues lately. Of course, open is the question of why I'm posting this here vs seeing this kind of from the Twitter people.

Comment from rick on 2008-05-25

2008-05-25T19:34:15Z

Bah... the embed code didn't work above... i should have previewed.

http://siteanalytics.compete.com/twitter.com+jaiku.com+pownce.com/?metric=uv#

is the source for what I was talking about.

Comment from Clay Newton on 2008-05-25

2008-05-25T20:41:40Z

My money is on Twitter not dying. I think they are in a rough spot, but will pull through it. Of all the services out there, they actually implement the least end user functionality (balanced by robust external API capabilities). This will be their saving grace.

Comment from Marina Martin on 2008-05-25

2008-05-25T22:03:26Z

@Andy Of course I don't want a news site to only have positive articles. On the other hand, I *do* expect a new site to write about, y'know, *news*. A seeming unending stream of "Twitter is dying" posts != news and is not the superb-quality articles I'm used to seeing on ReadWriteWeb, which is usually one of the first RSS feeds I read each day.

Comment from Boredcollegekid on 2008-05-25

2008-05-25T23:25:03Z

Excellent post, my sentiments exactly, its sad to see the service suffer, I hope the people can either shape up or find someone who can do it for them because its painfully obvious they can't

Comment from Larry L. on 2008-05-26

2008-05-26T07:56:31Z

Call me crazy, but I didn't realize twitter was this big. I need to get with the times! Thank you to both the article and comments for the good info.

Comment from Andrey Golub on 2008-05-26

2008-05-26T10:14:08Z

I think twitter.com will soon end up as the mass-phenomenon, as it happens with some pioneers that miss the moment to make a commercial project from their idea/ service. Several BETTER applications and services will come to out place twitter and I see this as just the evolution of the Web 2.0!

plz guys let's not mix MicroBlogging, an IDEA 2.0, and TWITTER, a web site (twitter.com)/ company (Twitter, Inc) that has just implemented earlier/ better than others (?) the microblogging idea.
The bloggers that judge another microblogging products as "twitter clones" really make me smile- what is twitter ("twitter.com"="Twitter, Inc")?! and what's microblogging?!

pay attention to this difference, web 2.0 crowd! let's talk about Web 2.0, not brands/ companies. We use products, of course we can't just use protocols and APIs :), but we do it 'coz we like the IDEA.

I bet the future is in the integration of microblogging to "where it could be useful" and it's also in Mobile MicroBlogging (take a look at BeeMood).
And in order to dream about integration "a company-powered platform" to somewhere else, it must have been RELIABLE during the free/trial phase. twitter.com wasn't reliable we know it, but it was "the spirit of microblogging" so we love it and will probably still love twitter.com when another companies with better business models will launch better products implementing microblogging! This is what I wish to happen soon!

Comment from jansegers on 2008-05-26

2008-05-26T14:16:18Z

What about a new approach ? just remembering the last 20 posts destined for any user ?

i20=i19
i19=i18
i18=i17
...
i2=i1
i1=new

Wouldn't it reduce the scalability problem ? Request for more history are relatively rare I believe. (Based on a comparison with search engines - most people don't go beyond the first page of results, some go the page 2, very few to page 3 and virtually none to page 4.)

It would mean some information would be retained in more versions but the total amount of this redundancy would be quite manageable.

It's just an idea...

Comment from Roger on 2008-05-26

2008-05-26T15:35:42Z

Looking at the bigger picture, the Twitter concept is here to stay, no doubt. Conceptually, Twitter's appeal is rooted in its function as an information management system. And therein lies its weakness. In an era where avalanches of information descend upon us daily, Twitter's simplicity allows us to structure the chaos. So Twitter is just too useful. It won't die. But in what form will it continue to live on...? Watch this space!

Comment from Luciano Evaristo Guerche (Gorše) on 2008-05-26

2008-05-26T16:32:35Z

Caring about it, booked accounts in bkite.com, pownce, jaiku and others as redundant ways of communicating to my networks/communities.

Comment from Baxter Tocher on 2008-05-26

2008-05-26T18:36:11Z

If $15m funding can't get them sorted, we're all stuffed. Can't they at least buy a few new servers with part of their new-found funding? It's not as if what we're doing on Twitter is highly audio or graphics intensive, is it? What really *is* the problem?

Comment from Todd on 2008-05-27

2008-05-27T14:56:34Z

"There's no such thing as bad press" - P.T Barnum

Comment from Michael on 2008-05-27

2008-05-28T03:06:55Z

I think most of Twitters scaling problems are just because no one has really dealt with something like this before, it is such a database intensive service that it is a little bit more complicated then it seems.

They'll get it all figured out and working smoothly just like all the other services out there that had scaling issues in the beginning.

Comment from Laura on 2008-05-28

2008-05-28T13:36:59Z

Long live Twitter! Or at least the concept!

Comment from Andrey Golub on 2008-05-28

2008-05-28T14:54:32Z

x Michael, author of the post containing "no one has really dealt with something like this before"

maybe you meant difficult to scale with Open-Source or scale with no money investments? (for somebody passing to Oracle from MySQL seems to be a real enhancement of the infrastructure :lol: ), but also here most CTO of social projects will respond you "it doesn't really metters of $$$, it's a matter of heads/ hands"!

so that problem of twitter.com is something different from the technically too complex problem to solve , it's for sure.

Comment from C. Enrique Ortiz on 2008-06-08

2008-06-09T04:36:43Z

Bad design, but perhaps, bad capacity planning....

ceo