Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : The New QLogic (ANCR) -- Ignore unavailable to you. Want to Upgrade?

To: Craig Stevenson who wrote (13422)	1/3/1998 11:38:00 AM
From: George Dawson	Read Replies (3) \| Respond to of 29386

Craig, I also happened upon this interesting latency information on the Sun site: sun.com Under the NUMA section, about 7/8 down there is this excerpt: "NUMA inter-node interconnects vary in bandwidth and latency. The Sequent NUMA-Q, for example, is characterized by a 1ÿGB/sec. interconnect bandwidth and a variable remote-memory latency, typically 3.3ÿusec. to as high as 10ÿusec. With a local memory latency of 180ÿnsec., this yields a memory access latency ratio of 18:1. By way of comparison, the Sun UltraÿEnterpriseÿ10000 uses a switch-based interconnect with 12.8 GB/sec. aggregate bandwidth and typical 0.5 usec. latency." From a previous section, Symmetric Multiprocessing (about 1/4 down): "In addition to high bandwidth, low communication latency is also important if the system is to show good scalability. This is because common data warehouse database operations such as index lookups and joins involve communication of small data packets. Sun provides very low latency interconnects. At only 400 nsec. for local access, the latency of the Starfire Gigaplane-XB is 200 times less than the 80,000 nsec. latency of the IBM SP2 interconnect. Of all the systems discussed in this paper, the lowest latency interconnect is on the Ultra Enterprise 6000, with a uniform latency of only 300 nsec. When the amount of data contained in each message is small, the importance of low latencies is paramount." For comparison, from the MKII page: ancor.com The following latencies are noted: Class 2 or Class 3 frame: < 0.6 microseconds Class 1 frame: < 0.7 microseconds Class 1 connect: < 1.3 microseconds Class 1 disconnect: < 0.7 microseconds These statements illustrate how important latency is in high I/O systems. To me it also suggests that if you are connecting an FC switch to this system there is an inherent advantage to having a switch latency that is the same order of magnitude, although the advantage may be limited by the I/O discrepancy. I think these data mining applications may be what Cal was referring to when he said that some applications work better with smaller frames. George D.