Readers Speak Out: Letters to the Editor Our newest column, IT Architect, looks like a winner -- we received more reader response to last month's debut than we could possibly print in this issue. Plus: Chuck Musciano amends his Unix uptime feature with good news from the real world; Rawn Shah takes another look at cable modems; Performance Q&A with Adrian Cockcroft; and more

April  1998

Mail this article to a friend

Connectivity: "Cable network ins and outs" by Rawn Shah

http://www.sunworld.com/sunworldonline/swol-03-1998/swol-03-connectivity.html

In defense of cable modems

Rawn,

Your comparison of cable modem and DSL technology and networks was very interesting. Being a cable modem operator in Argentina I feel I need to respond to some of the performance and security issues you mentioned.

Even though I agree that DSL provides up to 8 megabits per second of dedicated speed per user, it will take phone companies a lot of upgrading to deliver that level of service and, in the end, they will always be limited by that twisted-pair last mile.

Cable networks have a coaxial last mile, which offers a hundred times the bandwidth capacity. There are already 100-megabit-per-second-capable cable modems in the labs, and that's with only one 6-megahertz channel dedicated. Typical HFC nodes have 1500 to 2000 homes passed, and they share the bandwidth. With a 750-megahertz network you could dedicate 10 channels to each node for Internet services and have 1 gigabit per second for that amount of customers. What I'm saying is that having such an enormous last mile will make the difference in the long run.

As for security, cable modems are looking into ways of offering more and more security. For instance the modems we use (Com21) offer 56-bit DES encryption with new keys every 24 hours. It would be very difficult to crack our codes given that limited amount of time. Also there are filters for IPX, NetBEUI, and peer-to-peer networking. With these filters, our network doesn't behave like an Ethernet network at all.

Cable modems have a few other distinct advantages:

• Cable companies charge flat fees (unlike telcos)
• They provide a real online connection (once you try it you'll never go back)
• They exist now!

Daniel Nofal

Daniel,

I've received several messages like yours, pointing out that some cable modem security concerns have been addressed by vendors already.

It's arguable how much a telco would have to upgrade before it could deploy DSL. It's quite different depending on where you go. DSL fees, in many areas of the U.S., are also based on a flat rate, and are not charged by the megabyte. Internationally, there are several countries with flat-rate DSL service.

In many parts of the U.S. almost all the channels on local cable services have been filled even on 750-megahertz networks, and providers now have to sacrifice some channels for others according to demand. Allocating 10 channels is a nice idea, but not every cable service provider can do that.

I have seen a 6-megahertz channel give up to 38 megabits per second; that's approximately 0.019 megabits per second (19 kilobits per second) per home per 6-megahertz channel for 2000 homes. Ten channels bring the ratio up to 190 kilobits per second on average -- and that's just one direction. The math aside, it is true that more often than not you can get rates of up to 400 kilobits per second in your home, dependent upon network usage.

Digital services over cable are excellent for digital TV services -- something that DSL will have to work really hard to provide. HDTV is coming to the U.S., and the two real contenders are cable and LMDS services.

On the other hand, you are right in that cable service is available now. DSL service is barely sparking in some major cities in the U.S., while digital cable has been running for over a year. It's going to be an interesting race.

I'm interested in hearing about digital cable services in countries outside the U.S. I try to keep up with the news, but that doesn't reveal what's really happening with providers.

Rawn Shah

Performance Q&A with Adrian Cockcroft

http://www.sunworld.com/sunworldonline/common/cockcroft.letters.html

Measuring tape I/O

Adrian,

Our environment is very tape intensive (3480,3490, DLT, etc.). We are currently implementing a profitability/charge-back system. The problem is what to measure. We are using a third-party product that uses system accounting. Obviously using CPU stats is a no-brainer. But we would love to get more I/O stats. I know we have disk I/O to look forward to, but the character I/O measurement is (or seems to be) a gray area. What comprises the statistic?

Also, we are very interested in measuring tape I/O specifically. Any ideas?

Alfred W. Somers

Alfred,

Solaris 2.6 includes tape statistics that match disk statistics. They are shown by default in iostat. We do not have per-port character I/O, but iostat -t does give systemwide character I/O counts. Per-process I/O is also available in aggregated form, not per port or per file.

Byte-oriented throughput on network interfaces is available with recent patches to the network interface drivers (le and hme) on Solaris 2.5.1, and is provided in 2.6. A summary of some of the data is shown below using an SE toolkit script:

Name    Ipkt/s Opkt/s InKB/s OuKB/s IErr/s OErr/s Coll%
le0      229.3    0.4  19.26   0.03  0.000  0.000   0.0

IT Architect: "The world of the technical architect" by Kara Kapczynski

http://www.sunworld.com/sunworldonline/swol-03-1998/swol-03-itarchitect.html

Feedback for our newest column, authored last month by Kara Kapczynski

Thank you for your column, it was written well and will be placed in my educational library as a "basic" document. Periodically, I find one like this that rises above all others.

Regrettably, I'm not your audience. Rather, I'm a Web developer who aspires to understand informational technology as an entire system, including the functions of historic (pre Internet) systems and the changes they are undergoing as they integrate with the Web.

My greatest weakness is in the area of corporate IT. Most articles about that subject presume that I, the reader, understand the big view. Terms -- for example, "legacy systems" -- mean much more to me now that I've read your explanation.

Jesse Collins

I welcome the IT Architect column. I am an architect employed by Nortel working on an enterprisewide architecture supporting the implementation of supply chain management worldwide. We have the corporate-suggested guidelines and standards and then we have the specific l.o.b. components and connections.

I am very interested in your ideas and how I might apply them in the next few months. I am especially interested in guidelines or templates of architecture documentation that may be available in the public domain.

Henry Zabokrzycki

I truly enjoyed your opening column, and I'm looking forward to next month's installment.

My most recent roles have been in the design and architecture of systems and software products rather than IT services. These types of architecture embody many of the same issues as IT architecture (i.e., multiple platforms and OSs, third-party and home-grown software, internal and external networks, etc.). The goal in each case is to support a line of products that are sold and supported, rather than to run an enterprise.

As your column progresses, if there are areas where you see the differences and can recommend approaches for this different set of problems, I would appreciate it.

Keith A. Miller

Excellent reading -- really puts to use different processes and thought modes to ensure that the most critical aspects of an enterprise's architecture and vision are taken into account.

I look forward to reading more in this area from you.

Oleg Chaikovsky

"GUI toolkits: What are your options?" by Cameron Laird and Kathryn Soraiz

http://www.sunworld.com/sunworldonline/swol-03-1998/swol-03-gui.html

Sun funded, user approved

Cameron and Kathryn,

I read with much interest your article on freeware. Given that this is in a Sun-related publication I wanted to make sure you are aware of the freeware project Sun has supported for a number of years.

This support began with Sun's major support of the SunSITE project which I proposed to it in the early 1990s and which has now grown to many sites around the world. See http://www.sun.com/sunsite. for a major set of archives for freeware.

The Sun Freeware Project (http://sunfreeware.com) focuses on getting as much freeware as possible ported to Solaris for SPARC and Intel and packaged for easy use. Many hundreds of thousands of users have used this free service, funded by Sun, to save time and find the software they need. This project is mirrored on most of the SunSITES.

Steven M. Christensen

Worth noting: News and notables from our contributors

My March feature, "Is mainframe-class availability possible in a Unix environment?", prompted a bit of e-mail from a colleague of mine. A year ago, we were part of a team that brought a mainframe-class Sun system online to support a financial application running on Oracle. That system -- a Sun Enterprise 6000 running Solaris 2.5 backed by 700 gigabytes of mirrored disk -- services 500 simultaneous users crunching the books for a billion-dollar business. It just celebrated its first birthday.

In its first year online, it had just 30 minutes of unplanned downtime, which translates to 99.994% uptime. That's great uptime for any environment, and an astounding feat for a system in its first year of existence. Don't think it was easy: The people running that box are outstanding sysadmins, and the hardware is redundant from the dual network interfaces to the duplicate disk-controller cards. Still, it's a great example of what good systems design and rigorous production discipline can accomplish in the Unix arena; something we should all strive for.

Chuck Musciano

A note from Bill's Bookshelf columnist Bill Rosenblatt regarding his February column

A couple of readers objected to my use of the word "Jew" in my review, ("Technology (and ego) gone too far", February 1998) of Ellen Ullman's Close to the Machine and Michael Bloomberg's Bloomberg on Bloomberg.

I used the expression "Jewish Ivy League baby boomers" as a way of tying together the two very different authors of these very different books. It is a string of epithets that also happens to describe me.

Bill Rosenblatt

If you have technical problems with this magazine, contact webmaster@sunworld.com

URL: http://www.sunworld.com/swol-04-1998/swol-04-letters.html
Last modified: