J. Talbot. Turning the AIX operating system into an MPcapable OS. In Proc. USENIX Technical Conference, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....when other resources are shared, and study the performance of our system for real applications. 8 Related work A number of papers have been published on performance issues in shared memory multiprocessor operating systems, mostly in the context of resolving specific problems in a specific system [5, 6, 9, 22, 26, 28] . These systems were mostly uniprocessor or small scale multiprocessor systems trying to scale up to larger systems. Other workon locality issues in operating system structure were mostly either done in the context of earlier non cachecoherent NUMA systems [8] or, as in the case of Plan 9, were ....

J. Talbot. Turning the AIX operating system into an MPcapable OS. In Proc. USENIX Technical Conference, 1995.

....the communication system performance requires that the speedup obtained from parallelism outweighs the associated synchronization and context switching overheads. The CPU support is important. The load and reserve and store conditional instructions have replaced the traditional test and set [Chesson94, Talbot95]. They can be used to build all the classical synchronization facilities, lock free list manipulations, TCP IP statistic counters, etc. An appropriate use reduces the need of locks and improves the communication system performance. The parallelism model is also essential. Schmidt94] identifies: ....

....either of bulk type for throughput measurements, or of echo type for latency measurements (the client sends data and waits for an echo) The machines are a 42 MHz Power DPX 20 and a quadri processor 66 MHzPowerPC601 ESCALA. The operating system is either AIX 325, a monoprocessor system, or AIX 4. 1 [Talbot95], a multiprocessor system that also exists in a monoprocessor flavor. 4.1 Comparison of the BSD and classic STREAMS stacks Figure 5 shows the performance of the BSD TCP IP stack versus that of a classic STREAMS TCP IP stack. There is one TCP connection over a loopback interface (1460 byte MSS) ....

J.Talbot, "Turning AIX operating system into an MP-capable OS", USENIX, New Orleans, January 1995.

....when other resources are shared, and study the performance of our system for real applications. 8 Related work A number of papers have been published on performance issues in shared memory multiprocessor operating systems, mostly in the context of resolving specific problems in a specific system [5, 6, 8, 21, 24, 26] . These systems were mostly uniprocessor or small scale multiprocessor Unix systems trying to scale up to larger systems. Two 1 2 4 8 12 16 Processors 1 10 Slow Down a) sgi convex ibm sun numa 1 2 4 8 12 16 Processors 1 10 b) 1 2 4 8 12 16 Processors 1 10 c) 1 2 4 8 12 16 Processors 5 1 Slow ....

J. Talbot. Turning the AIX operating system into an MPcapable OS. In Proc. USENIX Technical Conference, 1995.

....For example, gcc spends over 20 percent of its time in the kernel under Ultrix [8] which would limit the useful size of a multiprocessor to 5 processors if all kernel requests were serialized. As a result, a fully preemptable (and fully parallelized) system software base is generally required [9, 29]. Fine grained locks are generally needed to achieve a high level of concurrency and improved performance; however, the finer the granularity of locks, the larger the number of locks that must be acquired to complete an operation, resulting in higher overhead even if there is no contention for the ....

....software must generally switch from using simple flags to using full locking, with significantly more expensive special memory synchronization instructions. This difference is illustrated by the cost of a lock unlock pair in AIX on the PowerPC, which is 100 times more expensive that a cached store [29]. 5 Another result of the high cache miss latency is the movement towards larger cache lines of 128 or even 256 bytes in length. These large cache lines are an attempt to substitute bandwidth (which is relatively easy to design into a system) for latency (which is much harder to design in) ....

Jacques Talbot. Turning the AIX operating system into an MP-capable OS. In USENIX 1995 Technical Conference Proceedings, January 1995.

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