Lecture Notes 5/2 Last lecture agenda - current research in OS - Hints for OS design - grad school for people interested - questions - next midterm monday evening Processings from the last OS conference included in the following topics - Security - memory protection - fault tolerance - overlays - given the internet, u can impose your own architecture on top - Procedings show what research is done where, might give people a good idea of where to go for grad school, if they are interested in specific research areas - File search, indexing - internet worms - security - intrusion recovery - I/O systems - primarily concerned with ernergy savings, partially due to ecological reasons etc google has a facility rite next to a power plant to save power cost - avoid waiting for file I/O - file sytem - bugs - scheduling/performance - disk performance - some of these performance issues were hot in 1970's not so much today - thin client computing - having huge servers and you just have a screen and a keyboard - mail service - web caching - energy optimization - high performance communication sytems - world is going towards a state of being monotonically network - multiprocessors - the "JOKE" - since the 60's multiprocessors have been a thing of the future - until recently multiprocessors were only in high end hardware, now they are found in consumer electronics - parallel programming is relaly difficult - army example - biggest multiprocesing entity really inefficient - capability based operating systems - naming systems - dynamic mobile systems - distributed virtual machines - routers - priority scheduling - OS kernal optimization - security for java - scalable microprocessors - distributed shared memory - RAID file systems - peer to peer storage - distributed storage system - garbage collection - interprocess communication - fault tolerance interface design - must satisfy conflicting requirements - should be complete - should be fast, have a small footprint - has to be simple - should capture minimum exceptions - neither abstraction or simplicity is a substitute for getting it right, must get it right - much better to have simple stuff run quickly than have complicated stuff to run slowly - more low level operations and less high levels - theres always going to be bottlenecks - normal to spend 80% of the time in 20% of the code and its almost never possible to figure out wat the 20% is apriori - dont hide power - allow useful low level functionality to be accessible - leave it to the client - as long it is cheap to pass control back and forth - keep basic interfaces stable - building something new, then the first one is not going to work, dont spend all your money on it - even when implementation is successful, it pays to go back to rethink design - compute in the background when possible processor scheduling - tweaking little parameters here and there very often doesnt do anything at all - large red button that indicates dissatisfaction of user - either improve service or kick user off - making a reliable system is not hard to create, but retrofitting a preexisiting one is difficult - make actions atomic or restartable - also called a transaction - either completes or has no effect grad school short talk - CACM - guys wrote a little program in which you enter your parameters and it will tell you wat your chances of getting into grad school are - why go to grad school? - 4 years of college doesnt bring you up to professional training - masters is what you need - good way to build your professional network - bachelors - basic technical work, masters - more independence, and design responsibility, phd - research - do what you like doing - low end stuff is being outsourced, so good idea to get a graduate degree - masters degree starting salary - much higher than bachelors - how long does grad school take - masters - 1 to 3 years - where to go for grad school? go to the best place you can - research experience and smarts - two different types of things admissions officers look for - hard to say anything in the personal statement that will get you in, but easy to say stuff that will keep you out - weaker school - hard to get financial support