July 25, 2006

I successfully defended my Stanford EE PhD.

Posted in Category: School — Amr Awadallah @ 2:02 pm | link | | comment (0)
Share

The showdown happened yesterday (7/24/2006) around 11am at Stanford University in Packard 202.

The chairman for the orals was Prof. B.T. Khuri-Yakub, the other examiners included Prof. Mendel Rosenblum (my research advisor and founder of VMware), Prof. Hector Garcia-Molina, and Prof. David Mazières (both Hector and David are also on my reading committee).

The public part of the orals defense took about 1 hour, this is where I get to do a presentation about my work, and the public is invited to attend. After that we took a 5 min break then the private part started, where the profs got to hammer me with all types of questions to reveal whether I really know what I am talking about ;). I think that took about 45 minutes. They then asked me to leave the room and spent about 15 minutes deliberating (felt like eternity). Prof. Khuri-Yakub (with his very pleasant smile) was the first one to come out and give me the good news.

BTW, according to Google Scholar, my first paper in this series (published in 2002), is the most relevant result for the query “Dynamic Content Distribution” (thanks Google), and has 35 citations so far.

There is still one step left to get the PhD degree, and that is to finish the thesis (the book). It’s currently about 90% done, and I am targeting end of August to complete it, so if everything goes smoothly, I should be done by end of this summer quarter.

One thousand thanks goes to a very long list of folks that helped me achieve this (hard to list them all), thanks also goes to Stanford for being such an amazing entrepreneurial School, and to Yahoo! for being a company that cares about my career just as much as they care about the work I do for them.

You can get a copy of the presentation and my papers at this link:

http://vmatrix.awadallah.com

From that page you can also download a hi-quality video for the talk, or you can watch the low-quality version at Yahoo video search (also embedded below):



— Amr

Share

• • •

July 23, 2006

Tomorrow (Mon 7/24/2006 at 11am) is my Stanford PhD Orals defense …

Posted in Category: Work — Amr Awadallah @ 10:35 pm | link | | comments (6)
Share

I can’t believe this day has finally come 🙂 please wish me good luck.

The EE PhD orals at Stanford is two parts, the first part lasts about 1 hour and its a public talk were everybody is welcome to attend and hear what my work is about (if you happen to be in the area, then please pass by). The second part of the orals is private, its about 2 hours, where the four profs (including my research advisor) get to drill me to make sure I know what I am talking about.

Please find attached below the talk announcement, I also included a map with easy directions on where to park and how to find the Packard building. Essentially take the University Ave exit from 101, then continue crossing ElCamino. University Ave then becomes Palm Drive which is the main road into Stanford campus. Make a right on Campus Drive and follow the map from there to the Packard EE building (room 202).


Special University Oral Examination

APPLICATIONS OF VIRTUAL MACHINE MONITORS FOR SCALABLE, RELIABLE, AND INTERACTIVE INTERNET SERVICES

Amr A. Awadallah

Department of Electrical Engineering

Monday, July 24th, 2006

Packard 202

Refreshments at 10:45am

Talk begins at 11:00am

———————————————————————————–

Abstract

Today most Internet services are pre-assigned to servers statically, however, using the classic operating systems concept of virtual machine monitors (VMM) we are now able to dynamically move such services between servers. A VMM encapsulates the state of the machine in a virtual machine file, which could then be activated on any real machine running the VMM software.

In this thesis we introduce the vMatrix framework, which is an overlay network of virtual machine monitors. We then illustrate, through practical implementations, that such a network can improve the interactivity, scalability, and reliability of existing internet services without requiring significant code or architecture changes (i.e. backward compatible).

The main applications that we demonstrate in this thesis are:

  • Dynamic Content Distribution: Moving services closer to the Internet edge hence reducing latency and rendering such services more interactive for end users
  • Server Switching: Sharing a pool of servers between more than one service, hence leveraging the benefits of statistical multiplexing to reduce overall system cost
  • Equi-ping Game Server Placement: Placing game servers at optimized locations to improve the fairness of multiplayer first-person-shooter games.

We also demonstrate additional side benefits including on-demand replication for absorbing flash crowds (in case of a newsworthy event like a major catastrophe) and faster recovery times for improved overall reliability.

Click here to access relevant papers and presentations.


Packard 202

Share

• • •

July 20, 2006

Google Q2-2006 sequential growth slows down to single-digit

Posted in Category: Work — Amr Awadallah @ 12:46 pm | link | | comment (0)
Share

Google just reported their Q2-2006 earnings, and their sequential growth is just 9% over Q1-2006 (the wallstreet whisper number was 10%). This confirms that the search slowdown is industry wide. Note that almost half of that gain in revenue for Google is not natural growth, its one-time growth coming from more ads placed on top of the natural web results (specifically the state where they have 3 ads, e.g. the flowers query).

So for Q3 over Q2 their sequential growth will be even lower than 9%, except if they increase their ads significantly and go to 4 ads above web results (like what yahoo does), and/or start placing ads at the bottom of the page (not just above and to the right). However, we now have to wait and see whether wallstreet will get tricked into believing this 9% growth is natural and sustained, the stock price tomorrow will tell us that. Wallstreet got tricked before, and then it bit them back for Q4-2005 once the revenue accelerating changes caught up and only the natural growth remained.

Unlike Yahoo, Google does not give forward revenue guidance, but if they did, I bet they would be revising the numbers down a bit just like yahoo did.

Cheers,

— amr

PS: Many of us at Yahoo are sad at how severley wallstreet punished our stock, -20%, but we are hanging in there. Our executive management team choose to focus on the long term, and not try to rush things on the Panama front given how much of our long term value is tied to the new platform, they want the Panama launch to be perfect. I am proud to work at a company with that style of leadership, and I back them up 100%.

Share

• • •

July 17, 2006

Hilrious, yet amazing, Math Skills.

Posted in Category: Fun — Amr Awadallah @ 4:07 pm | link | | comment (0)
Share

A friend forwarded these hilarious mathematical deductions along to me, they truly speak for them selves:

  1. Find X

    Where is x?

  2. Limit Theory

    limit

  3. n cancels out with n

    sin six

— amr

Share

• • •

July 13, 2006

Prediction for Google Q2-2006 Earnings …

Posted in Category: Work — Amr Awadallah @ 11:09 pm | link | | comment (0)
Share

Well, I have to admit that its very hard to predict the Google earnings for Q2 due to the reasons I list below. I think its really up in the air, they might do very well, but they also might disappoint horribly, since the expectations (the whisper expectations) are very high this time around.

The reasons are:

  1. Google Adwords launched a serious of very complicated updates in Q2:
    • Google increased the number of pages with 3 listings in the north of their SERP (i.e. above the web results), this change leads to more revenue and was covered by Imran Khan from JP Morgan.
    • Google tightened their matching algorithms such that less, more relevant, results are returned. You can read more about that change on the Google Adwords Blog. This change leads to less revenue, however it does lead to improved user quality (kudos to them for doing that).
    • Google adwords now lets you specify a lower target position for your ads. This change will tend to reduce revenue since it will most likely be used by advertisers who are getting excellent performance and thus ranked very high, but they do not want to show up high since that leads to too many clicks, which can consequently exceed their advertising budgets.
    • Google adwords now lets you specify which time of day to activate your ads. This is also for advertisers who want to reduce their spend and only target theirs ads during certain times of day. Some say that this change might lead to more competition during that time of day, and hence higher CPCs and more revenue, so its not really clear in which direction this will go.
  2. Comscore qSearch numbers might be giving a very false signal for Q2 over Q1 of 2006. They are saying that Google (and search overall) experienced double digit growth in search volume (20%+ sequentially). That kind of growth is simply not possible based on historical trends for this time of year, and given the growth slow down Comscore showed for Q3-2005 into Q4-2005, and again for Q4-2005 into Q1-2006. I think Comscore might have updated their stratification formulas around April 2006 and that shifted the whole baseline up a bit (for Q2 vs Q1 that is).

So in summary, the risk is very high this time around, very wide range of expectations for sequential growth. The secret expectations might be too high and folks will get disappointed (by Google that is, I obviously cant comment on Yahoo, but Yahoo is more than search, Google isn’t). That said, I really hope that Google stock price will fall, I am very jealous ;). Their valuation is now $124B, that is half the valuation of Microsoft with all their experience and diversified revenue streams, it simply does not make sense.

— amr

Share

• • •
Next Page »
Powered by: WordPress • Шаблон: ADMIN-BG • 33 queries. 0.509 seconds.