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 (17794)	8/25/1998 9:46:00 PM
From: Roger Arquilla	Read Replies (3) \| Respond to of 29386

Craig, I sure wish I could even attempt to understand any of what you just wrote. It might as well have been in Japanese for me. None the less, I am sure that there are many who do understand it and appreciate your continued insight. Your last statement about a fibre channel to ethernet bridge reminds me that your suggestions to Ancor management should have been more closely scrutinized in the past. Perhaps the stock wouldn't be priced less than the cost of a Big Mac if they had. We are well into the third quarter at this point. Do you, or anyone else who is reading this, have any idea if Ancor has sold anything this quarter? I am beginning to get the feeling of panic, that we may be headed towards another revenue report like the last quarter. A 50% drop from this price would just about wipe this company out, not to mention many investors like myself who are hanging on by only the smallest of margin. Maybe the engineers wives (assuming most are men) could hold a cookie bake to boost revenues.

To: Craig Stevenson who wrote (17794)	8/25/1998 11:05:00 PM
From: George Dawson	Read Replies (1) \| Respond to of 29386

"The interrupt issue that Dr. Rauechle discussed is a key element of Alteon's approach to Gigabit Ethernet. (They are pushing Jumbo frames for this very reason.)" Craig, I had the chance to talk to him in person at the shareholders' meeting this year. He pointed out the I/O differences, SCSI support, and why proprietary solutions would not be useful. He also told me the "market has spoken" in terms of FC rather than GE in the SAN. His rebuttal covers that pretty well. FC is 4 and 0 with support from the major disk vendors. In addition to rebutting articles, FC companies should probably be proactive and have the appropriate FC vs. GE white papers on their site (some do now). It also might be useful to post real life benchmarks of FC vs. GE in the same enterprise (I don't think the overhead of GE switches + adapters is necessarily any less). The problem of course is that the big ATM guys are also vying to be the big GE guys. You won't find any FC references on the Cisco site. Now if we can just sell them a few switches..... George D.

To: Craig Stevenson who wrote (17794)	8/25/1998 11:46:00 PM
From: Greg Hull	Read Replies (1) \| Respond to of 29386

Craig, The fact is that Ethernet wasn't ever designed to be a CHANNEL technology, and it simply doesn't have what it takes to become one. Could you expand on this point please? Is this an issue of media access and higher latency? I assume switched ethernet does not employ CSMA/CD - is this true? Can switched ethernet masquerade as a channel? While I'm at it, on vaguely related issues, is ATM a Layer 2 protocol in the OSI? Since I am now the proud owner of an ADSL connection to my ISP, I'm trying to pigeon hole everything. Would it be fair to say I'm running TCP on IP on ATM from my ISP? Anyone know what the Physical Layer is called? Is it ADSL? My Netspeed 204 is then a router mapping ATM into ethernet at the Data Link Layer? Do ethernet and FE differ only at Layer 1? What the heck, I might as well go for it: are SONET and WDWM Physical Layer protocols? Feel free to pass on all of the OSI stuff. I'm just assuming that someone who reads every networking magazine published on this planet ("there is no such thing as a boring network magazine") might be able to answer them off the top of his head or point me in a direction to shorten my research time. Thank you, Greg

To: Craig Stevenson who wrote (17794)	9/1/1998 8:10:00 AM
From: Neil S	Read Replies (2) \| Respond to of 29386

More from the nwfusion SAN forum re the Paul Desmond Water Cooler article: BTW- Recall this one ? ancor.com <<Reliability was the most important requirement. "We can't have our applications down for even one second," said Ron Howell, network architect for IBM Global Services. "Functionally, Ancor's Fibre Channel solution filled our need for a high-speed and absolutely reliable product. We were also very impressed by how easy it was to set up and test a Fibre Channel solution. Plus, Ancor came through with excellent support and a real willingness to do what it takes to complete the project successfully." ---------------------------------------------------------------- nwfusion.com Ron Howell - 07:48pm Aug 31, 1998 EST (#6 of 6) Dear Fellow Network Engineers: There are 3 clear contenders for the Gigabit Speed LAN marketplace: Fibre Channel, Gigabit Ethernet and ATM 622 Mbps. Our labs have used all three technologies and would like to contribute some first hand experience to the Academic " Ivory Tower " Discussions we are holding here. Here are some "Real Life" experiences that may help the discussion and your networking. 1. EFFICIENCY: -------------- Fibre Channel is the clear winner in this category. The unlimited frame size of Fibre Channel MTU's allow Fibre Channel to transmit much more information with less overhead than either Gigabit Ethernet or ATM. Fibre Channel accomplishes this with a technology called SEQUENCING. The RAW line speeds that technologies are rated for mean little or nothing. IF the efficiency of the transmission is low, you will never get anywhere near the throughput of the advertised line speed. This is only seen in "Real Life" use of the technologies. LANE for example lowers the Efficiency of ATM to a MUCH lower level than either Fibre Channel or Gigabit Ethernet. Gigabit Ethernet's very small frame size is the most important limitation to the efficiency of Gig Ethernet today. Flow Control is a very close second place. 2. FLOW CONTROL: ---------------- The first law of data networks is that flow control is required; otherwise the traffic will quickly OVERLOAD the network, creating data loss and expensive delays. The advantage of having a GIGABIT speed pipe is quickly lost if Flow Control is not well implemented. Your expected speed increase will never be realized !!! You will then have wasted your money. Fibre Channel's Strong Flow Control: ------------------------------------ Fibre Channel utilizes Credit Based Flow Control when transmitting information. This means that you cannot send data unless you have an equal credit available. The BENEFIT of Credit Based Flow Control is that you will NOT congest due to blockage. Gigabit Ethernet utilizes an Xon/Xoff Pause based flow control in most cases. This is no where near as strong as Credit Based Flow Control and the possibility of Blockage is much greater than with Fibre Channel. This factor will directly restrict the actual throughput you will receive with that new Gigabit Ethernet or ATM Pipe you are thinking of building. ATM and Flow Control: An interesting fact about ATM is that " IF you drop a single cell, you must retransmit the entire FRAME again !!! ATM does NOT generally perform Flow Control in LAN implementations, ATM leaves this function up to the TCP/IP layer to control how much data is sent at a time ( MTU ) and when it can send with TCP/IP windowing. In most cases, when an ATM circuit becomes congested it begins dropping cells, (ATM just throws the cells of data away ) which can quickly lead to re-transmission of frames from the source. This is commonly known as the SNOWBALL effect. The more cells you drop; the more of your data frames need to be retransmitted, and on and on until the bandwidth is consumed by retransmitted frames. IF this happens, your efficiency is completly LOST and you will never get anywhere near the expected throughput you thought you were buying from the salesperson. We have demonstrated this effect many times in our labs. You can easily choke off a 155 Mbps ATM line with the SNOWBALL effect of retrans issions. Flow Control Lessons Learned: Without strong Flow Control RAW numbers on technologies mean little or nothing in real life networks. Error Checking: --------------- Without strong Error Checking you are placing your corporate data at RISK !! Fibre Channel and Gigabit Ethernet both use IBM's 8B/10B encoding due to the fact that Gigabit Ethernet uses Fibre Channel at the Data Transmission layer. Where Gigabit Ethernet parts ways with Fibre Channel is at the layers above. Fibre channel checks data transmissions for errors in 3 or more ways. Fibre Channel is a TRUE DATA CENTER class transmission medium, much like IBM's ESCON and FICON. ATM ,on the other hand, suffers 1,000 times the number of transmission errors that either Fibre Channel or Gigabit Ethernet does in the basic transmission encoding proceses. 1000 times the number of errors per second. Without strong error checking , you are putting your corporate data at RISK. You NEED a technology that gives you very strong data transmission encoding and data error checking !!! The three technologies we are discussing here are vastly different when it comes to error rates. The tramsmission of data is what networks are all about. As the speeds increase, the error checking and reliability must INCREASE in order to maintain the same level of Accuracy and Reliability. UNLIMITED SCALEABILITY: This one is very simple, the ONLY network technology that can be grown to unlimited speeds WITHOUT having to tear out and purchase new equipment is FIBRE CHANNEL. FIBRE CHANNEL Networks can grow in Bandwidth with the simple plugging in of another cable. Try this with Either Gigabit Ethernet or ATM, it cannot be done today. Fibre Channel has true LOAD AGGRIGATION that allows this "Growth" of Bandwidth to be possible. Fibre Channel is the only available Gigabit Speed Technology of the three we are mentioning here to offer this advantage of INVESTMENT PROTECTON. Zero spanning tree re-configuration times, Zero circuit switching times... BOTH FACTS that neither ATM or GIG Ethernet can offer you today. Ask your vendor these questions and see for yourself. If you would like more information about how all of this works feel free to call or e-mail us anytime. FRAME SIZE: Fibre Channel clearly WINS here !! Fibre Channel can send Sequences of 64 K or much higher with a SINGLE INTERRUPT. Ethernet of any kind can send 1,500 bytes with a single Interrupt. As Dr. Rauechle points out, this results in roughly 85,000 times the number of Interrupts that your CPU must answer. This KILLS CPU performance. As many large corporations have seen first hand, this will saturate your CPU and waste the expensive new Host you purchased. The CPU spends most of its time answering Interrupts, and little is left for processing. FRAME SIZE is a fundamental problem with Gigabit Ethernet. Alteon, bless their heart, has made a business of attempting to remedy this problem with their 9K size frames. A good step in the right direction, Alteon was the first company to recognize that ETHERNET at GIGABIT speeds was going to have a FRAME SIZE problem. It is very true !!! Each time Ethernet has increased the raw speed from 10 to 100 and now 1000 Mbps; the number of Interrupts per second has increased 10 fold at each intervalbecause the basic FRAME SIZE has remained the same 1500 bytes. This means the EFFICIENCY of Ethernet has gone DOWN by a factor of 10 each time the speed has jumped by the factor of 10. ( 10 - 100 - 1000 Mbps )... future speeds. We recommend that this not be IGNORED !!! INTERRUPTS STEAL CPU cycles !!! EASE OF USE / COMPLEXITY of the Solution: ----------------------------------------- Ease of use should be mentioned here; one factor that we as networkers need to consider is the COMPLEXITY of the solution you build. The more complex the solution, the harder it will be to troubleshoot and maintain. Much to our Surprise and pleasure; Fibre Channel demonstrated itself to be by far the EASIEST technology to implement and use. FIBRE CHANNEL is Even Easier than Fast Ethernet to implement and use. Light Years ahead of building and running even a small ATM network. Setting up and running an ATM network is non-trivial. WHY ??? Everything in Fibre Channel is built into HARDWARE. There is nothing to setup or change. As a 14 year veteran network engineer I can tell you this is something that was VERY Important to our people. Ethernet and ATM are both heavily reliant on Software to get their work done. HARDWARE is far more Reliable and FAST than SOFTWARE based networking methods. Period !! The faster networks become, the more evident this will become to other networkers. KEEP YOUR LIFE SIMPLE, and your network will perform at a much Higher Availability level for your customers. When you get rid of the marketing hype and the marketing "smoke and mirrors" you get right down to what are the important differentiators between these technologies. The fact that only Fibre Channel can transport SCSI protocols over long distances today ( > 25 meters ) should not be overlooked as Dr. Rauechle pointed out. Fibre Channel can easily cover 10 Kilometers right out of the box. Without this capability of carrying SCSI protocol commands , Gigabit Ethernet or ATM are of very little use as a SAN !!! We submit these thoughts to the readers in an attempt to differentiate the HYPE from the REALITY of these 3 major networking technologies. We at FCS Laboratories believe that Networks are built for only one reason: " To support users and the applications they need to do their jobs. " WE BELIEVE Networks must eventually become as Reliable as the Telephone Networks, and making networks un-necessarily complex is going in the wrong direction. TRY this for yourself, ask your vendor these same questions, we are confident that Fibre Channel will be the clear winner in the DATA CENTER Class networks that we must begin building for our future networks. We welcome your comments and/or questions on these matters. We submit this information in an effort to clarify the issues and hype surrounding these technology religious wars. There ARE clear differentiators between the three major contenders for Gigabit Networking into the year 2000 and beyond. We submit this information to help clarify the important points to consider BEFORE you buy a technology. If you have any questions or comments feel free to contact us at anytime for more details. ronhowell@hotmail.com or SMARTcentres@hotmail.com sincerley, Ron Howell MSEE , P. E. Chief Technology Officer FCS Laboratories A Networking Consultancy Group