A Market Design for IaaS Cloud Resources

It has been some time since I last posted on this blog. The reason for that is that I've been very busy actually writing my thesis.
After completing the introductory chapter, I continued writing on the other chapters and I'm proud to say that as of 2 days ago, the thesis document and accompanyning prototype have been completed. The document itself can be viewed below, while the implementation of the market prototype can be found online.




I wish to thank everybody who helped me during the writing on this thesis. In particular, my promoter Prof. Dr. Jan Broeckhove, my co-promoter Dr. Kurt Vanmechelen and my mentor Ruben Van den Bossche. I also wish to thank my family and close friends for their continued support.

If you have any questions about the thesis, don't hesitate to contact me.

Writing the introductory chapter - final part

As it turns out, writing that introductory chapter took more time than anticipated (probably because I wanted it to be of a good quality). With a lot of hard work, I have finished it now. Next up: chapter 2! Chapter 2 deals with general market design principles. It starts with a definition of 'a market' and subsequently gives a short overview of some existing financial markets and their workings. Some desirable market properties are discussed (incentive compatibility, individual rationality, allocative efficiency, computational tractability), and some considerations about commodity market design are outlined. The chapter will finish with some first more concrete ideas about the IaaS market, based on the previous commidity market design considerations.

It is imperative that I finish this chapter this week if I don't want to get behind on schedule. With that in mind, I already did some work during the weekend. However, as the next week is already rather busy (I also need to make a rather elaborate assignment for an other course), a lot of hard work will be required to finish chapter 2 this week. But, as always, I'm eager to work hard and make it happen :-)

I'll post an update as soon as I know something more (=when I finish chapter 2).

Writing the introductory chapter - part 2

The last week I've had a bit of a writer's block, as I tried to get something on paper multiple times without a lot of success. Writing went a bit better during the weekend, but I cannot say that it went really well. The result of this is that I did not finish the first version of the introductory chapter yet, although I have made good progress.

As the next week is rather busy, it is difficult to make accurate projections for next week (and as it turns out, not every week goes according to plan...), but I will definitely finish the introductory chapter by then.

Not every day can be as productive as the others, but let's make the next days count :-)

Writing the introductory chapter - part 1

I planned to both work on the prototype and write on my thesis this week, but that didn't work out as I had to spend quite some time on an unexpected assignment for a course on webinformation systems. However, by putting in some extra effort over the weekend, I did manage to write 8 pages about cloud computing (what is it?, Iaas/PaaS/Saas, public/private/hybrid, challenges). Writing went rather smooth because I was able to reuse some parts I had already written last year in a general introduction on cloud computing.

In the coming week I'd like to finish the introductory chapter (which is a bit of a bold statement, as that would probably involve writing more than the usual 5 pages a week) as well as work a bit on the prototype (secondary objective). I will be focusing on some examples of real world cloud providers for the different cloud computing types (IaaS, PaaS, SaaS), as well as write about the power industry and it's resemblances to cloud computing. I will also try to write a clear problem statement for the thesis.

Let's make it a productive week :-)

Price Discovery and Prototype

The last week I turned up working on different topics than expected. With hindsight, I think I was righteously cautious when I said last week that it was maybe still a bit early to make any real statements about the price discovery mechanisms I have spend time on in the past weeks. As it turns out (without going into to much detail; that would require a rather long blogpost), although valuable, in practice some of the price discovery mechanisms I proposed have some difficult issues to solve. However, helped by the writing I did the last few weeks, I believe I found a valid, far simpler alternative. That is, since the a consumer can always choose to refuse or accept a proposed price, the CRA already provides a sufficient form of price discovery as part of the market as consumers can always discover prices by submitting a bid and subsequently refusing the matching offer. Although it may seem that I have been looking for a far more difficult solution before, it is only quite recently that I am quite certain that giving the consumer the possibility to reject matches does not have a negative impact on the market. The reason I doubted this before was because of a potential loss of allocative efficiency in a mostly consumer oriented market. However, I have come to understand that this is not necessarily the case in an environment such as the cloud computing market where there is an seemingly unlimited supply of resources (again, going into much detail would require a far lengthier post).
I discussed this with my co-promoter this week (overview of my preparations), and we agreed that at least for now it is indeed a good idea to go with this form of price discovery.

We also decided that it is valuable to put some more effort into the prototype, as I've always put much emphasis on the real world applicability of the market model. In the past few days, I have thus primarily been focusing on that prototype, while also having done some writing. Concerning that writing; I focussed some more about how the CRA model can be implemented (this helped me out when implementing the prototype), and I also wrote a bit about compound bids.

I worked quite a lot the past week, and although I've only written my weekly 5 pages, I think my efforts on the prototype will pay off in the coming weeks.

In the coming week, I'll start writing on the introduction chapter of my thesis (about cloud computing and the power industry and their markets), as well as continue to work on the prototype (still a lot to do!).

Slides and the Constraint Lookup Service

This post comes to you with a bit of unexpected delay. Nevertheless, with this post I want to give you a status update. On Wednesday 23Feb, I had to given a presentation covering my progress so far. The fact that the audience seemed to agree with most of my ideas, is another indicator that the use of the CRA mechanism is probably a solid choice.
The slides can be found on https://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=0B2ZFuPHiC22wYmU0M2Q0OTktYjA1NS00NzcyLWI1OGYtYzM0YTVhN2ExOGQ4&hl=en_GB

About the writing on the thesis itself: this week I started writing on the 'Implementation of the CRA' section. I managed to meet my 5 pages/week deadline again by writing about how seperate lookup services should be used to implement the constraint operations. The figure below gives an idea of how such a service interacts with the other market actors.

I also made a first rudimentary price discovery prototype. Although the prototype still needs a lot of work (I'll be improving on it this week), the first results look promising (maybe it is still a bit early to make any such statements).

Besides improving upon the pricediscovery prototype, I'll continue to write about the implementation details of the CRA this week. Update next week!

About problem specification and powerpoint slides

As the week comes to an end, I'm providing you with another update.
This week I've concentrated on clearly specifying the problems that occur when trying to use Dube's market model in the current cloud computing market. I must admit that I had more difficulties with this than I had anticipated. Although the problems in itself are not very difficult, they do tend to become rather complex (as many different factors play a role). This makes clearly explaining what the exact problem is hard, as the problem itself is composed of different subproblems, of which some are also hard to define.
However, I do believe that I have succeeded in clearly defining some of the issues (a lot of work is still to be done). The result of this is now part of my thesis document (I succeeded in adding another 5 pages to the document this week).

I've also created a slideshow presentation for the presentation I mentioned last week. I will put the slides online after I've given the presentation (which will take on Wednesday, Feb 23th).

Although creating the slideshow has helped me to (re)gain a better overview over some parts, it has also made me more aware of the fact that I should invest enough time to thoroughly explaining some concepts and ideas in the thesis itself. In particular, I think I should re-evaluate the time I spend building a working prototype, the time I spend exploring new topics and the time I spend on elaborating on existing ideas. After this week, I believe that the elaboration of existing topics should get some more attention.

Since I also have to give a presentation about my research internship on Wednesday, and I still have some preparation to do for that, I think it is safer not to make to many specific plans for the coming week. I *am* going to try and work on the price discovery issue (which I didn't get to in the past week), as well as write an additional 5 pages for the thesis itself.

As already mentioned, I will be posting the slideshow of Wednesday's presentation in the second half of the week.