categories

HOT TOPICS

Thought Leaders In Cloud Computing: Willie Tejada, Senior VP And GM Of The Enterprise Cloud Division, Akamai (Part 6)

Posted on Monday, Sep 12th 2011

Sramana Mitra: Even on the private network, I am not sure that is working so well. One of the popular collaboration tools these days is Google Docs. You open up a spreadsheet in Google Docs, and 10 people around the world are able to see it. It is a smooth, real-time, convenient function. In the CAD world or a video editing world, it is just a completely different animal.

Willie Tejada: Yes, but wouldn’t you say that part of the elegance in the Google Docs environment is how lightweight the editing scenario [is]? What you have to edit is something significantly lighter in weight than what you have in the CAD.

SM: Absolutely. I think the bulk of the problem is in the data set.

WT: Right, and until those things actually move, [it will remain difficult.] Back to the entrepreneurship associated with that, what are the types of applications we have seen proliferate as SaaS applications? We have not seen a lot of what I would just call CAD-centric environments, even from the traditional players, become services that have best pronounced themselves as SaaS applications.

SM: Absolutely, anything that is IO intensive and real time has not done so well.

WT: No, but I believe strongly that is one of those areas where this is an evolution of what makes it out into the cloud. It happens organically. As we watch, even the transaction-based applications migrate, piecemeal. The front end, other logic that’s shared, and maybe eventually the data set migrate to the cloud. The types of applications that will move there or happen as well, in a fashion, will trail based upon how well tailored they are to exist in the cloud. When you take a look at the CAD applications and their computing intensity and IO and things of that sort, I think those will take – and you take a look at the inherent challenges even with collaboration – those might be the last applications that we’ll see stood up. But that certainly represents an opportunity, because we know it’s a problem. 

SM: Yes. Now, switching topics a bit, what do you think is the evolution of a public cloud telecom service like Skype? What needs to happen for that infrastructure to withstand the pressure of the demand because they are now powering Facebook video chat? Well, that is extremely  bandwidth-intensive stuff. The level of quality of service they are delivering on simple things like voice collaboration and voice conferencing has started to go down. So, what is the evolution of Skype, for instance?

WT: These are areas that I think are very interesting use cases for Akamai, [but] I do not want the dialogue to be focused on Akamai use cases. I think what has happened, in some cases, is the popularity and the use of the service has now outpaced what the Internet itself can deliver in terms of reliability and consistency. That is just managing communications on the client and, of course, on the server and the way that architecture is built. My point is that there are underlying premises of the internet such as the BGP routing protocols, how it handles pack and loss in congestion, that I think it ultimately end up being challenges that the services that scale have to overcome. They have to overcome them by either writing that intelligence into the clients and the servers or using facilities like overlay services from a company like Akamai or other people who can get into the business. The carriers and others along those lines may be able to do that. But typically they are associated with their one network. A [company] like Skype, they want to go everywhere.

The bottom line is that an overlaying network on multiple networks is likely the architecture that is best suited for delivering. But I think you get to a point where because of the reach and the popularity of the service and the bandwidth contention in relation to it, the requirements for quality of service, the variance for resiliency on things like pack and loss and congestion, ultimately have to actually move into fortification services to the service itself. They can only do so much of that inside the client and the server.

This segment is part 6 in the series : Thought Leaders In Cloud Computing: Willie Tejada, Senior VP And GM Of The Enterprise Cloud Division, Akamai
1 2 3 4 5 6 7 8

Hacker News
() Comments

Featured Videos