sorted by: recent | see : popular

Content Tagged with startups + infrastructure

In planning for last Wednesday’s Structure 08 conference, we at GigaOM had our heads in the cloud. We aimed to draw attention to the resurgence of hardware underlying the various software and web services that consumers and businesses now use, and hoped to define the emerging set of offerings that comprise cloud computing.

That definition is important. But not as important, I realized, as figuring out which business models will win out. Because while everyone wants to push their own definition of cloud computing, at its heart, cloud computing is about moving, storing and delivering data on demand.

After moderating two panels, watching almost all of the speakers and having numerous conversations, I came away with the belief that most people view cloud computing not as access to computing resources, but access to services ranging from application-specific offerings such as Salesforce.com to compute grids like that of GoGrid or EC2. And when it comes to buying into such data services (be they software, a platform, storage or compute time), there are certain questions that need to asked, among them:

How do I get my data into the cloud? Maybe it’s as simple as calling up Salesforce.com, or a bit more complicated, like using an API to tap into EC2, but to use a cloud you’re going to need bandwidth. Whether it’s figuring out how to measure and appropriately charge people for bandwidth as Google is attempting to do with their structured meta data, or contracting with a CDN to lower latency, the delivery of data in and around the cloud represents one of its biggest costs and is subsequently one of the areas that’s ripest for innovation.

What format does that data need to be in? Different clouds work with different software. Some clouds work with Windows and others are only friendly to the LAMP stack. Various people expressed the idea that the industry would divide along the lines of low-margin, general purpose clouds like Amazon Web Services, and high-margin, special-purpose clouds such as Heroku’s Ruby on Rails testing environment (which is built on AWS). The key is to know what you need from a cloud before investing.

How can I change and move that data? The differing programming languages or operating systems accepted by various clouds are only part of the issue. The still undecided fight will be between proprietary formats such as BigTable and open standards that are truly standard, as in used by many, many developers. It’s a young effort, so there are no set standards yet, but until there are, transferring data kept in the cloud will never be as seamless as the bank analogy pushed by Sun CTO Greg Papadopoulos.

So while I spent most of my time trying to figure out which areas of compute infrastructure can be offered as a service while providing the highest margins or most defensible market share, I should have been keeping my eye on the data, because how providers treat the data will determine how their business models evolve.

Solarflare Communications, a chip startup in Irvine, Calif., has raised $26 million in a third round of funding. That brings the total the company’s raised to $126 million, which is a lot of money for a chip startup, even when you consider that the amount includes money raised by Level 5 Networks, which Solarflare acquired in April 2006. But the startup is hoping to use that money to attack a big problem in the data center at prices lower than the current technology offers. And if it succeeds, it’ll make computing faster and data center operations more flexible.

Like many other communications chip companies, Solarflare is working on a way to deliver 10 Gigabit Ethernet over copper, which is cheaper than delivering it via fiber. That enables the high-speed transport technology to move outside of the telecommunications networks, where companies such as Infinera are already pursuing 100 Gigabit Ethernet over fiber, and into mass adoption in the data center. Getting the technology into servers at a reasonable cost would create a market 10 times bigger than that of networking switches.

Others chasing mass adoption of 10 GigE on the server side are Intel and Broadcom, which like Solarflare, have controller chips. Broadcom and Solarflare also have PHY chips sampling with customers. Solarflare CEO Russell Stern plans to integrate the PHY with the controller chip in 2009, beating Broadcom to the market. He will use some of the funding for that purpose.

It’s likely Broadcom will end up attempting an integrated 10 GigE over copper chip as well. Broadcom doesn’t talk about its chips until they’re sampling, but the company did make a mint by cornering the market for integrated 1 Gigabit Ethernet chips for servers. However, success for Solarflare or Broadcom is probably three years out and depends on creating an energy-efficient chip at the 32 nanometer process node, according to Bob Wheeler, an analyst at The Linley Group.

Power consumption is a big challenge for these chips because unless it’s managed properly, they run too hot for servers and switches. And because technology doesn’t stand still in the data center, where virtualization and ever-increasing amounts of data are screaming for fatter pipes, hybrid forms of networking technologies that mix fiber or Fibre Channel with Ethernet are emerging to bridge the Gigabit gap between servers and networking equipment. Broadcom has several products that take advantage of such a hybrid networking environment. Startups such as Arastra and Woven Systems are also in that sector, and may see gains at the expense of a unified 10GigE world, which means Solarflare’s market opportunity could fragment if cheap, integrated 10 GigE takes too long.

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

On the eve of Structure 08 in San Francisco, the folks behind many of today’s cloud computing initiatives will be gathering at CloudCamp. GigaOM is one of the evening’s sponsors:

CloudCamp was formed in order to provide a common ground for the introduction and advancement of cloud computing. Through a series of local CloudCamp events, attendees can exchange ideas, knowledge and information in a creative and supporting environment, advancing the current state of cloud computing and related technologies.

If you’re in the city, come check it out. The event, which will be held from 5-9 pm on June 24th, is community-backed and free to attend. And it will be a great primer on cloud computing.

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

Twitter, our favorite tool for narcissism and the eponymously named San Francisco company behind the service may not have a business model, but it surely has the buzz. Whether it is their new round of funding or their inability to keep the service running — the blog world loves to twitter about Twitter.

After talking to some of sources, I have a theory that could help Twitter solve its scaling conundrum and also help the company make money. (I am sure there are others who are thinking along those lines.) And in order to do that, I will use fellow blogger Robert Scoble, who has over 25,000 followers, as an example.

Robert is the perfect embodiment of what is wrong with Twitter, but he also offers the best hope for the company to figure a way out of their current infrastructure-scaling conundrum. I am not picking on Scoble, but using him as an example of “extreme” user who can put any system through a major stress test. Leo Laporte is another such extreme example, and has 37,000 followers on Twitter.

First let’s start with: What is the problem? Instead of using my words, lets go with Twitter’s self-acknowledged infrastructure problems.

Twitter is, fundamentally, a messaging system. Twitter was not architected as a messaging system, however. For expediency’s sake, Twitter was built with technologies and practices that are more appropriate to a content management system.

First, I am glad to see that they are not passing the blame onto their hosting providers, like Joyent. They have recognized a fundamental problem with their service, as pointed out by Assetbar in an earlier post, who wanted to offer a proxy service for Twitter.

The way Twitter is architected, when Scoble sends out a “tweet” it is sent to 25,000 of his followers — whether they are checking it from a desktop client, a mobile phone, Chat client or on the web. The message goes into a database, which then figures out how those messages are to be delivered to each of the followers. This causes the database to behave like an overweight man who gorged on a buffet at local Chinese restaurant.

Dare Obasanjo explains how it stresses out the database in his post, and correctly points out that by giving ability to add an unlimited number of followers, Twitter might have brought all the troubles on to themselves. Facebook, on the other hand, is smart to restrict you to 5,000 friends. Why? Because to process the social graph of 5,000 friends is compute-intensive, and costly.

Anyway, to put Scoble and his Tweets in context, let’s assume for a minute that he always has 25,000 followers and he sent them 12,000 updates which are all 140 characters long, the maximum size allowed by Twitter. Again, hypothetically speaking, assuming each update is 100 bytes, then 12,000 updates generated used up 30 GB of data. (12000 updates * 100 bytes)* 25,000 = 30000000000 (30 GB)

So here we come to the good part. This massive database of followers is what Twitter should turn into a business. Twitter should charge Scoble, Leo, me, Michael Arrington and anyone else who has more than 100 friends and followers. How about something simple? $10 a month for 1,000 subscribers. 25,000 subscribers means someone like Scoble should be paying them around $250 a month.

Let’s take it a step further. Twitter should limit people to 500 free messages a month. Any more should come in a bucket of, say, 1,000 messages for $10. Businesses like Comcast that want to use the service for commercial reasons should pay for the service, and so should startups like Summize, which want to build their businesses based on Twitter’s API.

This would also fit the Freemium business model that Twitter investor Fred Wilson so loves. And at the same time, it would help Twitter overcome its abhorrence for adding advertising to the messages. I think many of us have a lot to gain from the service: My alerts about my posts on the system are a form of advertising for my work, and generate enough attention that paying for the service makes lot of sense.

There are some who are going to argue that this will kill the service. I don’t think so. First of all, average people don’t have 25,000 followers. Most have about 25-50 friends and possibly an equal number who are a degree removed but are still part of social environment. I think that for the average person Twitter will remain free. I think offering a premier-tier service will help stop abuse of the system by curbing the random following that has become rampant on the system. It will force many of us with excessive number of followers to be more selective. By doing so, Twitter is also going to help lower the noise and make the system more usable. This will give them time to figure out how they are going to become a real messaging-based company.

Will Twitter be brave enough to make such a move? Chances are, no: They are stuffed with VC dollars and signs of wild growth (including outages) can help them flip the company, making it someone else’s problem. Still, I wanted to throw it out there. It is a holiday weekend, after all!

Recommending reading:

* Dare Obasanjo on Twitter and its scaling issue.
* Hueniverse: Scaling a Microblogging Service.
* Twitter-proxy: Any Interest?

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

Neither Om nor I are shy about talking infrastructure, but the High Scalability blog has gone totally geek and parsed the details of how Facebook plans to scale its new Jabber chat service to 70 million members using a hella lot of servers and Erlang. As Sandy Jen over at Meebo can tell you, chat is a challenge to scale because it requires a constantly open connection to the servers and low latency. That’s a recipe for a lot of hardware and some flexible architecture. Good thing Facebook has $100 million to spend, but bad news for the firm if the money spigot closes.

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

Ask any IT professional what they dread most, and they’ll likely tell you that it’s change. Specifically, the act of putting a new application into production: There’s simply no way to know what will happen.

Companies spend a tremendous amount of time and money in staging environments, trying to see whether or not their new code will work. Only the largest banks and utilities have the resources to completely replicate the production systems, and none can truly simulate the real world.

Jonah Paransky, VP of marketing for testing-sandbox startup StackSafe, says that roughly 43 percent of IT problems can be traced back to lack of pre-production testing. “Over half of the changes we see in our research never end up being tested against the end-to-end IT service.” StackSafe tackles the problem by copying an entire application into a virtual appliance, letting IT change it, and testing the result.

Virtualization is an important enabler for pre-production testing. IT has longed for a sandbox in which to let new applications play, but doing so in the physical world has been too costly. Increasingly, however, IT doesn’t have a choice: Best practice frameworks like the IT Infrastructure Library (ITIL) include processes for testing, and certain industries, such as power utilities, require that IT changes are tested and logged.

Server consolidation and disaster recovery firm Platespin also provides tools for sucking physical machines into the virtual world, but Paransky insists that StackSafe’s focus is different: In addition to creating a virtual clone of a production application, it offers a suite of testing tools to quickly see whether the clone is broken.

Easier pre-production testing couldn’t come at a better time. Because virtualization makes it trivial to create a new machine, IT administrators face a flood of newly minted virtual servers they have to manage. “Now that it’s almost no work to deploy new servers, the first thing that happens to new virtualized environments is an incredible growth in server sprawl,” said Paransky.

Sandboxing approaches like StackSafe have an important limitation, however. In order to insulate the real world from the servers being tested, the sandbox is completely cut off from surrounding systems. Otherwise, testing the sandboxed servers would have real-world consequences, like buying a ticket or selling shares. “Even if you could virtualize much of a production environment,” said Paransky, “there may be components, like SOA, that can never be pulled into that virtualized environment.”

Consequently, pre-production testing on the distributed Internet becomes increasingly difficult — you can’t put a copy of the Internet into your sandbox in order to test it. Which is a challenge firms like StackSafe still have to tackle if they’re going to improve the reliability of tomorrow’s applications outside of the traditional enterprise data center.

The cloud is growing up. Its rite of passage comes this morning with the announcement that Amazon Web Services will now provide support for users of its Simple Storage Solution, Elastic Compute Cloud and Simple Queue Services products. Amazon, with its launch last week of persistent storage, was clearly wooing enterprise users, and the offer to provide support signals a formal courtship.

This is a romance that’s been played out across technology for decades, most recently in the open-source market. New technology gets launched and academics, hobbyists, and other early adopters play with them. Eventually businesses start wondering if they might be able to play, too. But downtime, glitches and the sense that you’re on your own are big turnoffs for corporate buyers.

Amazon Web Services has decided it’s time to grow up and play nice with business. It’s offering two different service levels: One starting at $100 a month and the other, at $400. While smaller companies such as Nirvanix already offer support and better usability for business users, the Amazon brand will bring cachet to its offerings, no matter what else is out there.

Austin, Texas-based startup NextIO has scored $18.8 million in a third round of funding, bringing the I/O virtualization systems maker to almost $40 million in total venture capital raised since 2003. It’s attacking one of those nitty-gritty technical problems in data centers and tossing around today’s favorite buzzword to do it.

NextIO makes chips and software to tackle I/O virtualization for the PCI Express protocol. Virtualized I/O allows a group of servers and/or storage systems to run Ethernet, FibreChannel, InfiniBand or whatever other flavor of interconnect technology through one box. This makes it easier to manage a network without having to map each server to a set endpoint. NextIO expects to ship products this year to OEMs that make servers and storage systems.

It’s not alone in attacking aspects of I/O virtualization. Others include Neterion, NetXen, 3Leaf, Xsigo and even some 10Gig-E players who have added I/O virtualization capabilities such as Solarflare Communications are trying to bring virtualization across the data center.

I was reading this week about Blackwave, an Acton, Mass.-based startup that just changed its name (from Acionen)Blackwave and got a whopping $16 million from Sigma Partners, Globespan Capital Partners and IDG ventures, and all I could think was: What makes this company worthy of such a massive cash infusion?

As it turns out, two words: online video. Blackwave has come up with a new kind of hardware that combines all sorts of network elements — load balancers, storage systems, streaming servers — into one package, topped off with a Mensa-quality management software layer.

This isn’t the first time a startup has taken a do-it-all, “God box” approach to building hardware, mind you. And since most of them have failed, Blackwave’s future is far from secure. Still, their approach does have a certain logic to it.

When it comes to web infrastructure, most companies typically buy boxes from different sources — storage systems from the likes of Network Appliance (NTAP) or Isilon (ISLN), for example; load-balancing gear from Cisco (CSCO), switches from Foundry Networks (FDRY) and servers from Hewlett-Packard (HPQ), Dell (DELL) or Sun (JAVA). They then try to build a bespoke solution, managed with specialized software.

Blackwave has basically taken all of the boxes, put them into one device and made the device good at one thing — delivering rich media content (loose translation: online video) as smartly and cheaply as possible. “We can reduce the hardware costs by a factor of ten,” says CEO Robert Rizika. Why? Because the system can adjust itself according to traffic needs — almost like pants that adjust to the size of the meal.

Let’s say, for example, that an online video destination is hosting a Britney Spears video clip. The system has allocated 100 concurrent users to stream the video when suddenly, it gets linked to Matt Drudge’s web site, and the number of people who want to watch the video balloons to 100,000. The system will automatically reallocate resources to that video. Rizika claims other benefits. “Most single rack (servers) can handle, say, 2,000 concurrent streams, and we can handle 20,000 concurrent streams,” he boasts.

It is hard to assess how realistic those claims are since the company refuses to talk about how it’s getting things done inside its boxes; Rizika refused to even let us know if Blackwave was using its own chips. “We use off-the-shelf components,” is all he would say.

At this point I would typically become skeptical about a company’s claims. But Blackwave has two things going for it — Chief Technology Officer David Carver, and overall market trends.

Carver was previously director of research and development at video-on-demand company SeaChange (SEAC); prior to that, he worked at video conferencing supplier PictureTel Corp. In other words, he knows digital video very well, and understands its complexity.

The second thing working in Blackwave’s favor is the fact that the market is crying out for a solution like theirs. For quite a while, bandwidth (despite steady price declines) took up a major chunk of the money spent on delivering content over the Internet. Hardware, thanks to rapid commoditization, commanded a far smaller share of the monies spent. Now the pendulum is starting to swing in the other direction, with hardware consuming more of the total budget.

Blackwave backer Jonathan Seeling of Globespan Capital Partners and co-founder of Akamai told The Boston Globe:

“Most of the stuff out there today to store Web video is really systems that were always designed and always intended” for businesses that needed to store and sort through masses of data, not movies.. “People weren’t thinking about traditional storage systems as being used for delivering very large monolithic files of video.”

Rizika declined to give details as to who is currently using Blackwave’s boxes, but he did hint that the three-year-old startup counts a social network, a media company, and a user-generated content web site among its customers. Some speculate that the media company might be NBC. They are hoping to get other media companies and content delivery network operators to buy its gear.

They better do it fast - the cushion of venture capital cash can deflate very quickly.