Heh,

Windows is great because of all the add-on software available and because if you used *nix you''d have to write your own? That sure is an interesting comment.

You build an architecture in ''the garage'' and you duplicate it when you''ve found the right recipe, providing that you have a reasonable plan for scaling said architecture.

One obvious argument against having things broken down into smaller boxes is that sometimes it makes complete sense to have a few mid-range/upper-range servers because they take up less real estate than the 50 servers it would take to replace their computing power, not to mention that in a lot of cases it would be easier to manage and more cost-effective.

I don''t have anything great to say about support staff, being that I''ve never really been all that impressed with Windows support folks and generally speaking a few good *nix/network people are worth more than 10 cheap Windows folks that have a knack very little other than Windows voodoo ( which a lot of the time I''ve seen as a result of their own ignorance of what is going on under the hood of their beloved OS ). Anyway, not to start a flame war -- if you have a Windows environment you''ll need Windows folks and the rare good ones aren''t any cheaper around here ( Silicon Valley ).


.z

i know this is stupid but what is UDP?

quote:

---

Original post by SomePerson
i know this is stupid but what is UDP?

---

User Datagram Protocol, which is a simple datagram-oriented ( whereas TCP is stream-oriented ) transport layer protocol -- that is in simple terms, each output operation results in a single UDP datagram ( not counting fragmentation mind you ). UDP doesn't provide the reliability that TCP does because it simply packages the application data up and sends it to the IP layer for delivery ( and since IP provides best-effort but no reliability, UDP isn't reliable on its own ).

If you have further interest I suggest you take a look at the RFC ( and possibly some beginning TCP/IP books to get you going ):

http://www.faqs.org/rfcs/rfc768.html


.z

[edited by - zfod on February 7, 2004 5:20:45 PM]

Somewhere up in this thread hplus0603 talks about ''seamless simulation server migration''. Very interesting concept, and im looking for more info on this topic, links anyone?

(google''d allot, but can''t realy find anything satisfying... )

thnx

AP, start a new thread for a new question!

Also, google for "ghosting" which is how most systems do it.

Regarding using a high-end server for space savings WRT computing power: there just isn''t that much more compact computing power to get than Intel iron these days. Sticking them in a Blade server can up your density a bit, but comes with other trade-offs, such as cost and an even worse thermal situation -- not to mention non-standard hardware.

All in all, racks of 1U is Where It''s At (tm).