Seti is down again

[msattler]

Unfortunately, despite repeated emails, we have had no response from the top brass at Seti. Trying to help with the current woes, but seemingly unable to.

Sometimes it makes it so hard to keep the faith.

Giz.

I will try to check into this a bit.......

[Gizbar]

Thanks Mark.

You seem to have a good line of communication with the guys at the moment, and it would be a shame to see this kit wasted.

Regards, Giz.

[msattler]

I have sent you an email on the subject.....
Must get some more sleep with the kitties now.

And hopefully our favorite project can get things moving a little bit in the morning.

Meow meow.

[Josef W. Segur]

The first campus shuttle bus should arrive at the lab in about an hour. If Jeff's usual early arrival happens, things which couldn't be handled remotely will be possible. I just hope they get the server status page updating fairly promptly.
                                                                                   Joe

[msattler]

The first campus shuttle bus should arrive at the lab in about an hour. If Jeff's usual early arrival happens, things which couldn't be handled remotely will be possible. I just hope they get the server status page updating fairly promptly.
                                                                                   Joe

Well....
Parts of the server status page have updated, so some activity is taking place.  But some of it is still iffy.  Even though the upload server still shows as online, they are not working yet.

[Cosmic_Ocean]

Giz, if memory serves correctly, I believe they are trying to avoid "whiteboxes" since there is no real support for them.  If you spend US$10K on a server from a major brand, they provide support to keep you happy so that hopefully, you'll buy more from them.  With whiteboxes, you're on your own, basically.

That was discussed in the one forum thread that had basically a wishlist, and one of the servers they wanted was pretty simple, but through a major brand, was more than five times the cost of going to newegg and getting all the parts, and the latter option was turned down.

Now maybe these days, with desperation at its peak, they might be willing to go with a whitebox, but if Oscar is going to be as good as we're all hoping, maybe not.  We'll just have to wait and see.

[Fredericx51]

Aside from from the Server issues, how about the BandWith, which is an isue, too.
UPDate to GigaBit Link,  that also is necessary and has been Discussed many, many times
times before, I presume, the new SERVER, has to have 1, better 2 or more, gigabit
nic's, but also have manage the INcreased DataFlow per second.
(Which puts more strain on the Servers, then ever!)

Quite a change anyway, cause MODEM-ROUTER and everything between this end SERVER,
has to be UPDated, too. In most cases, there isn't much, between these 2, a (cat5) cabel.
(Twisted Pair)
Is there any news, about this, remenbered this?
I was called if I was interrested in VDSL and Fiber?  Still expensive, but incredible fast!
AsynchronousDSL*, is the most used connection, I think, WiFi is gaining strong and has
many advantages, especially in house. Not suitable for longer distances then 1 Mile (Visible, but 0.5 mile is a more realistic distance)
* Has limitations, except ADSL 2 is fater then ADSL (1), but the speed, strongly depends on the Distance between 'Home' and  'Central' or a (MODEM and Amplifiers Central)
Sorry, I'm drifting (away)
On Topic.


Digital Subscriber Line, VDSL.
VDSL.

[Cosmic_Ocean]

Upgrading to gigabit would in fact make data transfers go much faster, and my understanding is that the lab is already gigabit internally, including the router that is hooked up to the line that runs down the hill.. it's the equipment down on campus that needs to be upgraded, as well as the medium itself.

However, making a bigger pipe will only increase the strain on the servers due to allowing more connections to successfully reach through the saturated link.  It's one of those Catch 22 situations, but as others and I have said so far.. we'll have to wait and see just how good Oscar really is before theorizing new things to change/upgrade/replace.

[Fredericx51]

UPDated, my 'previous post' and added the remark on more load on the Down Load SERVER,  the servers end should be able to handle that added traffic!

Seti isn't the only  project with the most people, but also in Serving Hosts

[perryjay]

Okay, now we're testing my memory here. They just ran one fiber cable up the hill but it is for the entire SSL, hence SETI only gets 100MB out of it. The memory check comes in that as I remember, Hurricane Electric is already giving SETI  a Gig for the price of 100GBs. The problem was in that the college IT department wanted something like $30,000 to run the dedicated fiber up the hill to them. My guess is that at the time they ran the new fiber for the lab SETI couldn't afford the price of the fiber so that it could be run up at the same time. Again, if I remember correctly, the college won't allow an outside company to run anything on campus, it must go through the college IT, even though Hurricane would probably do it for SETI at a greatly reduced price.

Somebody please correct me if I'm wrong here.

[RottenMutt]

now i know nothing about the server side, but anyone other then me starting to wonder if this is a software problem?

[msattler]

now i know nothing about how the server side, but anyone other then me starting to wonder if this is a software problem?

I did....I brought it up and Eric was gonna ask Jeff to update reload mysql, which apparently they tried on jocelyn with no improvement.  Don't know if they tried it on mork or thumper.
Eric was going to peruse the mysql bug reports to see if there were any clues there...

Not sure why they were not able to get things running today.....I'll probably hear a bit more in the morning.

[RottenMutt]

not just the database, but in the basic programming/logic of the whole thing.  shouldn't the results out in the field be approaching zero?!?

[Miep]

not just the database, but in the basic programming/logic of the whole thing.  shouldn't the results out in the field be approaching zero?!?

No?

Some 38 hours ago there were about 4 million results in progress, with at least half a million issued in the last reporting/DL window.

That number can only drop if results are reported or time out (and only with running validator/transitioner).

If the processing was running with no UL/DL you would see the decay with timeout [asuming 'out in the field' does not include tasks waiting for resend due to timeout] - this would reach zero after about 6 weeks
If everything was up but no new WU were being handed out you would see a fast reduction from quick hosts followed by slow reduction while slower hosts hand their work back in - possibly asymptotic to the number of tasks not really being crunched. In six week that fallout times out, gets reissued and you get another asymptotic decay to a smaller positive number. Which times out another 6 weeks later and get reiessued. Thankfully we are dealing with natural numbers (and finite resends) or that number couldn't reach zero in your life time With a six week deadline you'd be looking at some 42 weeks of reissues.

Look at the bright side - if a child was concieved the day SETI stopped handing out new work and just waited till everything was back, that child would see the light of day before the last WU was crunched.

[RottenMutt]

so you think the last work unit will need to be reissued 9 times before credit is given!!!???!!!