S. D. Gribble. A design framework and a scalable storage platform to simplify internet service construction. Technical report, U.C. Berkeley, 2000. PhD Thesis.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....other services (cross interference) and hurt overall system performance. Second, if a system requires modi cations to the existing HTTP protocol [21] it may be impractical to deploy: the large number of deployed clients makes it dicult to change clients, and the increasing complexity of servers [25, 29, 31, 47, 56] makes it dicult to change servers. What is therefore needed is a prefetching protocol that (a) avoids interference at clients, networks, and servers and (b) does not require changes to the HTTP protocol and the existing infrastructure (client browsers, servers and networks) In this paper, we ....

....with demand requests [16] However, such mechanisms are not easily deployable because modifying the protocol implies modifying the widely deployed infrastructure that supports the current protocol. Furthermore, as web servers evolve and increase in their complexity by spanning multiple machines [25], CDNs [1] database servers, dynamic content generation subsystems [20, 33] etc. modifying CPU, memory, and disk scheduling to separate prefetch requests becomes increasingly complex. If interference were not a concern, a simple prefetching system could easily be built with the present ....

S. D. Gribble. A design framework and a scalable storage platform to simplify internet service construction. Technical report, U.C. Berkeley, 2000. PhD Thesis.

....a number of systems have proposed variants of the model. A common aspect of these designs is to structure an eventdriven application using a set of event queues to improve code modularity and simplify application design. One of the starting points of the SEDA design was the I O core framework [54] used by the Ninja system at UC Berkeley. This framework was used in Gribble s Distributed Data Structures (DDS) 52] layer and initial versions of the vSpace [139] cluster based service platform. In the Ninja I O core, a limited number of threads is used to process requests flowing through the ....

S. D. Gribble. A Design Framework and a Scalable Storage Platform to Simplify Internet Service Construction. PhD thesis, UC Berkeley, September 2000.

....and the O S multiplexes these virtual machines over hardware. Providing this abstraction entails a high overhead in terms of context switch time and memory footprint, thereby limiting concurrency. A number of studies have shown the scalability limitations of thread based concurrency models [6, 11, 21, 32], even in the context of so called lightweight threads. I O Scalability limitations: The I O interfaces exported by existing OSs are generally designed to provide maximum transparency to applications, often at the cost of scalability and predictability. Most I O interfaces employ blocking ....

....can be chosen at a per stage level, rather than for the application as a whole; this approach avoids wasting threads on stages which do not need them. For example, UNIX filesystems can usually handle a fixed number (between 40 and 50) concurrent read write requests before becoming saturated [6]. In this case there is no benefit to devoting more than this number Appears in Proceedings of the 8th Workshop on Hot Topics in Operating Systems (HotOS VIII) May, 2001. 4 accept connection connection parse packet read packet packet HTTP request receive packet HTTP request static ....

S. D. Gribble. A Design Framework and a Scalable Storage Platform to Simplify Internet Service Construction. PhD thesis, UC Berkeley, September 2000.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC