Article View: comp.sys.novell
Article #5961Re: Internal bridging throughput
From: Gary Heston
Date: Thu, 28 Jan 1993 22:45
Date: Thu, 28 Jan 1993 22:45
45 lines
2828 bytes
2828 bytes
In article <CC-MAIL+1/26/93_10#c#35AM+Internal_bridging_throughput*/S=BLANSFIE/OU=EM-01/O=CCMGW/PRMD=GOV+USDOE.G02/ADMD=ATTMAIL/C=US/@mailgw> /S=BLANSFIE/OU=EM-01/O=CCMGW/PRMD=GOV+USDOE.G02/ADMD=ATTMAIL/C=US/@mailgw.er.doe.gov writes: Gad, I'm already beginning to dislike X.400, and I'm not even using it yet... > I have heard that putting multiple NIC's in a server and using > internal bridging gives you better throughput than having only 1 NIC. Yes. This is exactly the case. > I heard also that you should have only 30 connections per NIC. No, 30 physical connections per 10Base2 segement. Those 30 physical connections (one would be the server) can attach to 16-port 10BaseT hubs, giving about 464 logical connections (nodes). I wouldn't load one NIC quite that heavily, though.... > Someone from our LAN ops group said that internal bridging is slow and > since you have to put each NIC on a different segment you have 2 hops > from the root. Internal bridging is quite fast. The primary bottleneck in a server is the I/O bus, and having intelligent or well-buffered adapter cards is the best way to relieve the constriction. I'm not sure what is meant by "2 hops from the root", though; anyone on either segement will be no hops away from the servers' executive (i.e., they'll have equal access; you wouldn't gain speed by having them all on one segement). If the users on segement "a" want to access a host or another server on segement "b", then they'd be *one* hop away. Odds are they wouldn't notice; I can't tell the difference.... > Having multiple NIC's makes sense to me but I need more information > can someone out there tell me where to look. I need some hardcore > data and/or formulas. I don't have benchmarks or formulas, but I do have four servers, each of which has *five* NICs installed, supporting a manufacturing line. They work well (servers are 486SX-25 w/8MB and a 540MB SCSI), fast, and reliably. I'll probably be installing a couple more in the near future, the way things are going. -- Gary Heston SCI Systems, Inc. gary@sci34hub.sci.com site admin The Chairman of the Board and the CFO speak for SCI. I'm neither. Windows NT documentation: "Command to change password CTL-ALT-DEL"
Technical Details
Message-ID:
<1993Jan28.224517.23325@sci34hub.sci.com>Path:
arc1-ca.pugleaf.net!sparky!uunet!sci34hub!garyReferences:
<CC-MAIL+1/26/93_10#c#35AM+Internal_bridging_throughput*/S=BLANSFIE/OU=EM-01/O=CCMGW/PRMD=GOV+USDOE.G02/ADMD=ATTMAIL/C=US/@mailgw>Show Full Headers
Complete Article Headers
Newsgroups: comp.sys.novell Path: arc1-ca.pugleaf.net!sparky!uunet!sci34hub!gary From: gary@sci34hub.sci.com (Gary Heston) Subject: Re: Internal bridging throughput Message-ID: <1993Jan28.224517.23325@sci34hub.sci.com> Reply-To: gary@sci34hub.sci.com (Gary Heston) Organization: SCI Systems, Inc., Huntsville, Al. References: <CC-MAIL+1/26/93_10#c#35AM+Internal_bridging_throughput*/S=BLANSFIE/OU=EM-01/O=CCMGW/PRMD=GOV+USDOE.G02/ADMD=ATTMAIL/C=US/@mailgw> Date: Thu, 28 Jan 1993 22:45:17 GMT Lines: 45