Hi everyone,
My mom used to say...and still says at 90. "If wishes were horses, beggars would ride!" Here are 3 wishes for this list.
1. I wish I was big and smart and everyone knew I was market savvy:
I haven't signed up as a partner...yet, but I was looking at the pre-release Microsoft Internet Security & Acceleration (ISA) servers and noted their claim
http://www.microsoft.com/presspass/press/2000/Oct00/CachePR.asp on the Web Polygraph 'Cache-off Event'. If I remember right they turned off the firewall rules. Here are the results for all entrants:
http://www.measurement-factory.com/results/public/cacheoff/N03/report.by-meas.html#Sec:Results When I look at the charts, squid did not appear to compare well. I doubt the Event has relevance for a network with 10 users connected via 56K dialup or ADSL...how much traffic can only 10 users generate through slow links? The Event was organized for high bandwidth connections with hundreds of users; the results won't scale to my little world.
Although Microsoft took full marketing advantage of the event, it looks expensive for little guys to enter the annual competition both in time and money. The Web Polygraph benchmark software is available as a i386 binary:
http://rpmfind.net/linux/rpm2html/search.php?query=Polygraph 2. I wish I was confident:
I do not yet have the confidence to run and interpret Web Polygraph all by myself. Has anyone at e-smith or in the forum run squid benchmarks for various category e-smith configurations against competing software?
If not... then,
If I can convince my partner (a MSCE) to give up his old hopped up 450mhz generic computer for a while AND if there are others who can help guide the setup and discuss results, I would like to run benchmark tests on a Category III e-smith server compared to an evaluation edition of ISA near the minimum required hardware. Using the same hardware seems a fair comparison for the e-smith target market if it can be repeated on several machines within the category.
http://www.microsoft.com/isaserver/productinfo/pricing.htm http://www.microsoft.com/isaserver/productinfo/sysreq.htmFor my market I want to compare price/performance of squid against ISA server on slower, more realistic connections; small office networks served by e-smith category III or IV servers. Since price/performance is my issue, obsolete or hand-me-down PCs (as low as 10% of original cost) should be an advantage for e-smith/squid. From my interpretation of the 'Cache-Off' results, squid will increase performance about 30 per cent where other caching software increased it up to 55 per cent. The performance probably varies with hardware tuning...which I can't afford to do on low end systems.
3. I wish I knew a simple, effective way to accurately compare (diagram) performance with competitors to business clients in my target market:
We all want to avoid spending money on performance that the customer would never notice given bandwidth limitations. It should build sales to be able to reference engineering rules of thumb or test methods (beyond what is already on e-smith pages:-) that help explain or pinpoint IP network bottlenecks. Obviously when one bottleneck is removed, others pop up. I want to estimate performance margins in order to make judgements as to what effect various hardware replacement choices (motherboard, cpu, memory, disk space, NIC, modem, power supply, fan) have on reliability and performance so I can to estimate the lifecycle cost of the server. I plan to lease mostly 8x5 workday servers that will last 2 years before I have to visit them; then they will be replaced, reformatted and given away.
For example, what are the times involved to fully recover from a power outage for the cache and what performance impact does the recovery have?
Because I live in California, it would be nice to advise users on complete recovery times (24 hours?) when long term power failures occur. I want to make diagrams that intelligent but Linux naive business persons would feel comfortable using to convince their peers that my systems will clearly meet their needs and have favorable price/performance over more established alternatives.
Caveat: I realize that it may be difficult to fairly compare Microsoft servers with e-smith servers because I probably will not be able to get all the functionality of e-smith to run reliably on one Windows box. I hope there is there a way around that problem that would be fair as seen by the market. In other words, Microsoft may not be able to economically scale dedicated server(s) to the smaller networks I plan to support.
Anybody have a favorite tool, technique or URL to use to compare software components for sales/marketing efforts?
Best Wishes!-)
Paul Miller
4 Replies
3667 Views