by Anna Filatova
08/25/2005 | 11:53 PM
The today’s keynote was presented Justin Rattner, Director of Intel Corporate Technology Group, who shared his vision of the future platform with us. And I have to admit that he spoke with such passion that I got really carried away and enjoyed this presentation a lot.
He was very excited to get us all in front of the technology and to keep us and you, our readers, on the leading edge of it. The question asked was: what will the platforms be like in, say, 5 or 10 years from now? What kind of platforms will these be?
The answer is simple: these will be user aware platforms. But what does Intel actually mean by this?
If you are a science fiction movie fan, then you probably can recall almost all the most illustrative examples of self-aware platforms. In most motion pictures or books, for instance, these “characters” very often got the main heroes into trouble. The issues of self awareness can be quite dangerous. Remember the Terminator movie? I believe everyone knows it well enough, but just in case let me remind you what the story is: Skynet virtual system comes online and gets out of human control. It is equipped with dreaded neural processors that begin to learn at a geometrically growing rate. Skynet becomes self aware and attempts to destroy human race. Remember now?
And this is not the only example we can actually bring here. The notion of turning against the user is a fundamental plot line in lots of movies and books. But it actually doesn’t have to be like this.
Some of you may have seen the Forbidden Planet. And to everybody’s surprise Justin Rattner introduced to the public the Robby:
He was a wonderful friend and a selflessly obedient servant. He cared more about users than about himself. And he was a great example of a self aware robotic system.
Unfortunately today’s systems are still quite far from sophistications of the robots like this guy.
If you look at the way you work and interact with today’s computer systems you will realize that you have to be very explicit, and make sure that you are very careful when you type in names, when you program command, or do some other routine work processing. If you have a constant power source you will only lose time in case you do something wrong, but if it is a portable platform you work on, you will also lose battery life. Namely you have to worry about the careful “feeding” of the platform. And just think about the last time your system fell under a virus attack? So, what he was driving at, is that we really need to get to point when the systems support the user, rather than user supports the systems.
In other words, we need the system to be user-aware :
It basically means that the systems should be able to take care of itself, weight out its resources and use them efficiently. It needs to anticipate with high level or certainty what needs to be done. It needs to sense the environment, to better understand the user and environment needs. It needs to learn and adapt to our needs. And this last thing is probably the most important. When we achieve this level of sophistication, we will fully unleash the power of the systems we built.
An example of a user aware system presented by Justin was actually very common and very illustrative. I think all computer users face this problem once in a while. I think most of us have digital cameras and we all take a lot of pictures. Digital photography is wonderful but it is so-called “user broken experience”. Why? - We have too many pictures and can never find anything when we need it real fast (unless we are organized, and sort pictures to properly named and dated folders with detailed content description right away). I am not like that, and it means that I am among those users who wouldn’t mind some help in searching for digital pictures :) Let’s see how this can be done.
The screenshot above shows a Diamond Project application. In this demo there are about 85,000 photos. But we need to find only one, and I our example it was supposed to be a picture of Justin Rattner in a blue IDF shirt delivering a speech. The photo is unlabelled, of course. One of the best known techniques that could be applied here is face detection. Then we create a new filter on the fly, by showing the system examples of shirts colors. And as a result we limit the search to just a few dozens photographs.
Diamond Project is a collaborate research product developed by CMU and Intel. And it is not the only one of the kind. According to Intel, they are now working on similar tools that could be used for sorting out medical images and graphical information in medical institutions.
Another aspect of the user aware systems is the ability of a system like that to take care of itself . This feature of a user aware system will be of great practical value in datacenters. We can be talking about tens of thousands of discrete server elements within a limited space. And this server space needs constant monitoring, so that we could identify the failure and respond to it before any data loss occurs. Here is an example:
On the picture above you can see a thermal scan of the data center room. Should the air conditioning fail, we need to act quickly. This is not a kind of situation that human administrator can respond in time to prevent problems. This is when platforms can handle the change of the environmental conditions and respond appropriately, like enable reserve air conditioning units, for instance and backup the data from the system in risk.
IBM introduced the notion of autonomic computing to describe systems like that. For the past years we’ve gone from dozens to tens of thousands of servers that need to be managed. We are at the point now when IT shops use 80% of the investment on maintenance. When you look at the environments today, there are different hardware pieces, it is a tremendous conglomeration of technologies. In this environment there is lack of standards so we need some kind of instrumentation to resolve problems. Autonomic computing is about building intelligence into the systems at different levels. Besides, it has to be an open standard approach and then you can build intelligent behavior on top of that. Intel delivered hundreds of self managing features in numerous products. One of these features is the mechanism for capturing event info.
Take, for example, Intel’s AMT Active Management Technology, which gives us the tools for reading and analyzing the events. Intel does a lot of research of the autonomic computing potential. One of the latest developments looks as follows.
There are wireless Intel modes that measure ambient conditions around the servers in a datacenter (server utilization, internal temperature, external temperatures and humidity in the room).
Say we have a situation when we’ve got two servers, and one of them is gets extremely dangerously overheated. What is going to happen in this case? Intel modes should pick up the changes in the ambient thermal conditions. The server that got overheated will be detected, and the load will migrate from it to another server, so that the system could hibernate without threatening to lose all the data.
Power consumption is another essential issue for contemporary data center. Actually, data centers operate more on a human time scale. If you zoom into the system you will see that changes in the platform power consumption and user requests are happening at a minor grain. Even though a demand is going up and down in just tens of microseconds, the power supply cannot respond fast enough and thus we are losing a lot of power. This may seem just a tiny power loss, because the time considered is less than a tenth of a microsecond, however, if we sum up all the tiny losses like that the total will not be negligible any more:
What of we would like to have a more optimal power use? Something like that, for instance:
How can we do it?
Main components of every mainboard are powered by voltage regulators. It rakes a lot of space and power. What if we removed all the voltage regulators and associative components into the “soon to be obsolete technology” waste basket. This is what will remain from the board:
Look how much space we could save this way on the mainboard PCB!
And here is what Intel suggests to use instead of all those obsolete-to-be voltage regulators and components:
This embedded solution will be only 35x35mm big, will operate at 100MHz frequency, will be 85% efficient and allow switching the voltage level. CPU, chipset and CMOS voltage regulator all-in-one. Intel even demonstrated a prototype mainboard platform for it, which was much smaller than the regular boards, I wish I had managed to take a picture of it, because they rapidly moved it off the stage after that: much be top secret, I assume :).
Another example of a user aware system important ability is the knowledge of the location relative to other devices and relative to resources of the network. It is called location awareness . It is important that the system can use precision location in order to find those assets. You will probably think that a GPS could help, right? But GPS cannot give 100% reliable position especially indoors.
What are other options we could resort to trying to teach our system to determine the location? We could actually use Wi-Fi. We could locate the access points and then locate the assets from there. How do we do it? Very simple: by measuring the signal strength:
But as you can se from the graph the results of these calculations are not really accurate, the measuring error is a way to big.
Another approach is to use the time arrival of the signal to a location and then from there calculate the distance to it. The experimental results suggest that with this approach the accuracy is much better. Take a look yourself: we can get very high precision, so that the deviations get reduced down to 1meter.
The demonstration of the system location awareness was actually very illustrative. Say, you are watching a movie, and then you take your tablet, and go to another room. The system calculates your location and transfers the playback of the movie to a screen in another room. You keep moving, and when you get into the hallway, the only screen you have around is the one on your tablet, so the system continues the movie playback on your tablet screen. As soon as you get outside the house the movie is paused, which can also be a good protection technique, against unauthorized access into your home network from outside.
As we have already indicated in the beginning of this presentation, another and probably one of the most important aspects of any user aware system is doing no harm to the user . It is not dramatic at this point, but has significant financial importance for network security. Computer viruses cause billions of dollars of damage around the world. The problem is that the viruses propagate so quickly that if you do not respond quickly enough you will lose control over the situation. The diagram below shows how fast the “witty worm” got spread when it hit one of the networks:
The solution at this point represents combined efforts of hardware and software products developed by Intel. This system below is equipped with Intel manageability engine silicon that analyzes the network connections and isolates the infected platforms.
According to Intel, this solution has been already tested in real conditions for about 8,000 hours and hasn’t revealed any false positives so far. In other words, no worms have escaped it :)
Well, we have just seen how greatly the system intelligence develops nowadays. It looks like in another 10 years we will create highly sophisticated user aware systems to help humans control important and complex spheres of our life. However, doesn’t this scenario look familiar to you? Machines get smart and take control over the human race completely. I could mention about a dozen examples when this happened (literature, motion pictures). I know you may say: hey, this is just science fiction crap, do not worry about it. But there were times, when an airplane was considered a wonder, and now we have them all over the place.
Technology breakthroughs are exciting. The power they give to us is huge. But maybe one day we should question ourselves: do we know what we are doing? Just a thought :)
P.S.: More news from the IDF showcase is coming, so stay tuned! :)