anything happen while i was pass...
err.. while i was traveling?

Bandwidth saturation? here?
and no pron? say it ain't so!?

I don't know, I just waited It out and I've been rewarded with this view of the Canyon.

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

What you left out was:

Actually, I left out a heck of a lot more than that. Things such as geographically diverse redundant data centers, UPS, back up generators, 24/7 staff (redundantly trained so that if a fire kills half your employees, you still can run the business), and so forth. It wasn't meant to be a complete list.

It's not easy to do, and it's hideously expensive. For something like SETI, the word "overkill" doesn't even come close. Downtime is tolerable in a system such as this.

I think the idea of redundant connections with diverse routing is a key component, and one BOINC is designed to get along without.

As you pointed out (and I've said more than once) if the SETI@Home servers are down, it's the BOINC clients that are mildly inconvenienced -- and they don't really care, they just retry.

The people don't even know, unless they're watching closely.

---

I think sometimes people forget what distributed computing is all about.

Let's say you need to build a computer system for Wall Street. You're going to make a lot of money, so you have a reasonable budget to work with.

1) You get some good people to work on the project.
2) You go buy some shiny new servers.
3) You go buy some fancy new software (databases and such).
4) You go buy some expensive service contracts for hardware maintenance with guaranteed response times when there's a problem.
5) You go buy some expensive support contracts so that, 24/7, if your database hickups, you can get Oracle on the line and have their best and brightest diagnosing your problem within minutes.

Sounds a lot like SETI? Er, no. Well, #1 does, but the rest?

What you left out was:

6) Multiple internet connections, from multiple providers so that everything can be multihomed -- and delivered over different paths to different entry points into the building.

In other words, if one provider goes down, or one idiot digs up a fiber with a backhoe, you want the other three connections to keep running at speed.

That is also expensive -- and not that easy to do either.

7) The remote backup server in a different city so if a [insert disaster] wipes out the building there is no hiccup.

---

I think sometimes people forget what distributed computing is all about.

Let's say you need to build a computer system for Wall Street. You're going to make a lot of money, so you have a reasonable budget to work with.

1) You get some good people to work on the project.
2) You go buy some shiny new servers.
3) You go buy some fancy new software (databases and such).
4) You go buy some expensive service contracts for hardware maintenance with guaranteed response times when there's a problem.
5) You go buy some expensive support contracts so that, 24/7, if your database hickups, you can get Oracle on the line and have their best and brightest diagnosing your problem within minutes.

Sounds a lot like SETI? Er, no. Well, #1 does, but the rest?

What you left out was:

6) Multiple internet connections, from multiple providers so that everything can be multihomed -- and delivered over different paths to different entry points into the building.

In other words, if one provider goes down, or one idiot digs up a fiber with a backhoe, you want the other three connections to keep running at speed.

That is also expensive -- and not that easy to do either.

7) The remote backup server in a different city so if a [insert disaster] wipes out the building there is no hiccup.

Actually, what Michael was going after was multiple, multiple redundancy, so it wouldn't be a remote backup, but multiple sites all working together.

Point being that all of that costs a lot of money, and if BOINC is supposed to make lots of processing available at a very low cost, you can't spend a ton of money and still do "low cost."

---