Only the cloud can scale

By Sean McGrath  7 comments

Only the cloud can scale

Step a little bit into the future. Not too far. Just a little bit. A handful of years perhaps. A group of web application developers get together in a bar to discuss their current projects...

"We are building a new .com website. We are hosting it ourselves. Doing our own storage layer and our own replication/fault tolerance, managing our own operating system instances, the works.", says one of their number.

"Oh that is so cute!", says another. "I'd love to be working on something retro like that! Yup. Those where the good old days of web development..."

Retro? Cute? Que pasa?

Step a little bit into the past. Not too far. Just a little bit. A handful of decades perhaps. A group of aviators get together in a bar to discuss their current projects...

"We are building a new aeroplane. It has two engines for fail-over. 6 seats. Two thousand mile range. We are building some landing strips for it too, complete with wind speed monitors!"

I suspect the history of aviation and the history of web applications will look somewhat similar at some point in the future. The early days of aviation were full of one-off aircraft and ad-hoc infrastructure. Over time, it became apparent that in order for aviation to go mainstream, massive infrastructure would be required at massive cost. Only countries could afford to own planes and run airlines and airports.

The early days of web applications - the current era - look similar to me. Lots of "one-off aircraft" in the form of bespoke application stacks. Lots of "ad-hoc infrastructure" in the form of DIY load balancing and redundancy and fail-over and so on. However, it is becoming apparent that in order for a modern web application to go mainstream, massive infrastructure is needed.

Today, a successful website may have to cater for many thousands of hits a second. Scaling to those loads requires some serious infrastructure. Multiple data centers spread around the world. Maybe tens of thousands of worker nodes during peak usage. Ability to grow and shrink with demand....

Are you really going to try to build all that yourself? Even if you could, would your application fly - metaphorically and commercially speaking - with all that financial overhead?

I doubt it. Being a success on the Web is becoming a terrifying prospect! Overnight you might have to scale from a few hits per minute to thousands of hits per second. To be successful you need to stay upright during flash foods. You cannot predict when they will come. If your site falls over, they eyeballs move elsewhere. You only get one shot at success in the fickle world of eyeballs.

What to do? There is only one way. Rent access to huge infrastructure. The modern web is seeing the emergence of a number of infrastructure superpowers. The web-infrastructure equivalents of Boeing, of FedEx, of Amtrak, of Qantas. We can debate which ones will succeed but I don't think we can sensibly debate their necessity.

The future is shared infrastructure. The future is the cloud. Not because it is trendy but because only the cloud can scale.

Fasten your seatbelts.

7 comments

    Anonymous 3 years ago
    Sean,A small typo correction for you:The Australian Air carrier is..Quantas -- QANTASQueenslandAndNorthernTerritoryAerialServicesQantasCheers from down under,Matt.
    ITworld staff
    ITworld staff 3 years ago in reply to Anonymous
    All set. Thanks for the head's up!
    Sean McGrath
    Sean McGrath 3 years ago
    Julien,You are right about Skype and other P2P networks. I missed that one. Thanks.I am thinking primarily of the classic multi-tiered web app with front ends, mid-tiers and back-ends.I should have been more explicit in my use of the term "scale". When I use it I am referring to not just the technical/theoretical aspects but also the economic aspects.The way I see it, a classic web app that becomes a serious traffic node on the Web needs a lot of OS instances, a scalable repository, geographic replication etc. etc. Sure you could put all that in place yourself and manage it all yourself but would your business model stack up? That is the nub of it for me.The cloud is a rental, pay-as-you go model.With so many of the business models on the web being based around eye-balls and advertising I have difficulty seeing how new enterprises will be able to make a profit if they own and operate all their own infrastructure rather than rent a managed service.regards,Sean
    Sean McGrath
    Sean McGrath 3 years ago
    Anonymous,Yes indeed. Real gems do indeed get hacked together in corners and I don't see that changing.What I do see changing is that high traffic, mainstream nodes on the net will need serious infrastructure and it will be more economical to rent it as a managed service, rather than run it yourself.regards,Sean
    kcurry
    kcurry 3 years ago in reply to Sean McGrath
    I agree with you, Sean. Furthermore, I would go on to say that managing traffic to web sites is hardly the only driver. The cloud also provides the ability for individuals and small orgs to leverage enormous amounts of computing space and power for the purpose processing massive amounts of data in a cost-effective way. (See also, Chris Anderson's article in Wired 16.07) But, I'm not sure drawing comparisons with aviation are constructive. We in the United States certainly do not want to create a future that resembles our present state of aviation! :^)
    Anonymous 3 years ago
    "However, it is becoming apparent that in order for a modern web application to go mainstream, massive infrastructure is needed."Perhaps that's the difference. We may need some mainstream apps, but diversity means that our real gems are still hacked together in the corner by a group of like minded individuals.
    Anonymous 3 years ago
    So for you, the only future for us is to strike 40 years back in the mainframe world where only happy few HAVE computers and the rest of us are slaved to beg for computing resources?So Skype success in a (quasi) center less design is just a dying cygnus singing the end of a person's owned computers?Thanks for this post, I'll return back my lonely cavern self lamenting about this fantastic linguistic acrobacy. For me, "clouds" are in the sky. In the internet/interweb are just computers big or small but always ONE's, be it Google's, Microsoft's, Amazon's, yours or mine.Centralizers did not succeed with "the network is the computer" motto but seemed to be more successfull with the weather metaphor.Let's see when/if it'll rain.

      Add a comment

      Post a comment using one of these accounts
      Or join now
      At least 6 characters

      Note: Comment will appear soon after you have activated your account.
      Obscene/spam comments will be removed and accounts suspended.
      The information you submit is subject to our Privacy Policy and Terms of Service.

      ITworld LIVE

      DevelopmentWhite Papers & Webcasts

      White Paper

      HP NonStop SQL Fundamentals whitepaper

      This whitepaper offers a detailed look into the fundamentals of HP NonStop SQL solutions. See how this system delivers unprecedented levels of application availability with fail-safe data integrity and meets the needs of enterprises with large-scale business critical applications.

      White Paper

      Nebraska Medical Center case study

      See how the Nebraska Medical Center implemented a SQL solution to make information more readily available to streamline operations, improve patient care and facilitate medical research with an enterprise solution running on HP NonStop servers.

      White Paper

      Concepts of NonStop SQL/MX

      For DBAs and developers who are familiar with Oracle solutions and want to learn about NonStop SQL/MX, this whitepaper provides an overview of the similarities and differences between the two products-with a specific focus on implementation.

      White Paper

      6 Things Your CIO Needs to Know About Requirements

      If your organization is not predictably successful on technology projects, there is likely an issue in requirements. CIOs must take action and own requirements maturity improvement. There are 6 main things a CIO must know about requirements.

      Webcast On Demand

      User Experience Monitoring

      In this webinar, you will learn hints & tips for improving end-user response times from Forrester Research analyst, Jean-Pierre Garbani.

      Sponsor: Nimsoft

      See more White Papers | Webcasts

      Ask a question

      Ask a Question