Title says it all.

The crowd is a LOT happer, tamer and generally accepting of things when there are updates ... ESPECIALLY when things go wrong.

It SHOULDNT matter, but were all human, and its the little things that count for us mere mortals.

Updates CALM crowds.
Stats and features BRING people to the project.
Jelly/Jello makes people want to wrestle.

Its all part of being human *hehe*. In a perfect world everyone would crunch for the science and not mind when the project is down, but thats never going to happen.


Anway. Thanks again. You are doing marvellous work.

---
<p align="center"> </p>Click on the counter to visit our team site including Extended Team Stats with daily totals, Member signatures and more!

I'll second that ...

But I don't see anythin in the latest note to explain why I keep seeing messages that the scheduler is failing and my cahce limit has been hit. I'd like to know what's behind those messages and what is happening to change that for the better.

>
> Updates CALM crowds.
> Stats and features BRING people to the project.
> Jelly/Jello makes people want to wrestle.
>

The more I think about it, the more I think I'd better just stay quiet and not add the 4th line that came to mind... :)

Bravo to the team for the updated news!

As I suggested elsewhere in this forum, why not get a Communications major to do the updates for you? That way, the coders can code and the communicators can communicate.

RedHats off to the Berkeley team, thank you
for the communique'..much appreciation.

Regards,

Xi3
<a> [/url]

Id like to secound (or rather, pentum..) this..
I still can't receive any WU, but updated msg's that shows things are being done, is easing my pain easily!

Until things work, back to 'classic'.

-Antwelm

I looked up some info on the snap appliance they are going to be moving the db to is it just me or do they appear to be lowballing some of their needs if your db is io bound why would you use an appliance that only using sata drives rather than scsi???Other than that it looks like a nice unit. well hopefully im wrong and we wont be hearing in two weeks that the new snap cant keep up with the read write actions on the db

heres the info

Accommodates up to 7 SD 30 Expansion Arrays for scalability up to 30TB

· Hot-swappable Serial ATA (SATA) disk drives

· Dual Intel Xeon processors with Symmetric Multi-Processing (SMP) capabilities

· 2GB Double Data Rate (DDR) system memory

· 512MB of battery-backed non-volatile memory

· Industry-standard iSCSI support

· Dual-Gigabit Ethernet ports

· Redundant power and cooling

· Pre-configured RAID 5

· Integrated Computer Associates eTrust antivirus software

· Backup agent support for BakBone, CA, VERITAS and Legato

· Available Ultra2 SCSI port for local tape backup

I personaly don't read chinese so I would like
to know if the last news post bring good or bad
news. What about all the WU we have crunched
during the down period and uploaded/reported today ?
Are these WU also lost in space ?

Thanks
Marc

from what the update said..it appears they arent lost they just have to find a way to inject them back to the database ...at least thats what i got out of it

Perhaps they have a RAID box that will hook to the iSCSI?

Just use the SATA for OS and such.

> Title says it all.
>
> The crowd is a LOT happer, tamer and generally accepting of things when there
> are updates ... ESPECIALLY when things go wrong.
>
> It SHOULDNT matter, but were all human, and its the little things that count
> for us mere mortals.
>
> Updates CALM crowds.
> Stats and features BRING people to the project.
> Jelly/Jello makes people want to wrestle.
>
> Its all part of being human *hehe*. In a perfect world everyone would crunch
> for the science and not mind when the project is down, but thats never going
> to happen.
>
>
> Anway. Thanks again. You are doing marvellous work.
>
And whats more, if you post with some detail of the problem, you might find that someone else has faced the same problem, and knows a solution.

I have faced the same problem, and what we did was to split the file names and create a directory structure out of them. The problem is the absoloute number of files in a directory, not the number of files in a tree.

from what the update said..it appears they arent lost they just have to find a way to inject them back to the database ...at least thats what i got out of it
==========

I think anything before the 5 day shut down is lost due to the bug they said it had about reporting returned results.

Anything run during the shut down and since they have been back online has a good chance of being good and receiving credit once they can get them back into the data base, at least thats my take on it all ...

I tend to agree PoorBoy that by the way it was written that all before the shutdown are gone. At least I hope that is the case. Over the last two days I have uploaded 73 WU's and had the work reported. Well if all is lost, it will just mean starting over again. Happy crunching to all.
Oh and by the way , I saw a post about why they were using SATA instead of SCSI, and if I am not mistaken, that really does not matter when the basis of the problem is the speed information that can be read off of the drives, or am I mistaken?

Rocky

If its speed they want thats simple - don't use SUN hardware (*ever*) - and never ever use NFS - use SAN hardware and get some decent processors in - surely it must be possible to say to some company like HP - look at this stuff sun gave us - can't you do better!!! Then get some decent Alpha or I64 boxes in and a SAN and watch things fly. Whilst we're on the subject - talk to Oracle as well :D

M.

I also applaud the information. More, please!

The flow of news the past couple of days has been wonderfully refreshing!

Kudos to the S@H/BOINC team and, please, keep up the good work!

Although I haven't been getting many units (predictor always gets its nose in there first!!), I feel like I actaully am sort of aware of what's happening Good work!!!!

Once the schedulers and verifiers and all that science stuff is figured out, PLEASE try to separate the forum systems from the distribution system: When distribution is down, it's REALLY important that you are able to communicate with us, and us with you!!!

Thank you for keeping us at least reasonably informed during this Extended Beta Testing =-)

> Once the schedulers and verifiers and all that science stuff is figured out,
> PLEASE try to separate the forum systems from the distribution system: When
> distribution is down, it's REALLY important that you are able to communicate
> with us, and us with you!!!

Good point... I wonder what kind of hardware the team feels they need for "extras" like that... Perhaps we could help find some donated or low cost equipment.

> Oh and by the way , I saw a post about why they were using SATA instead of
> SCSI, and if I am not mistaken, that really does not matter when the basis of
> the problem is the speed information that can be read off of the drives, or am
> I mistaken?

SATA drives right now are fairly inexpensive compared to a like sized SCSI unit. Right now SATA is 150 GB/sec rated and this is roughly the speed of the most common SCSI speed. As far as read speed, that is partially a function of the rotational speed of the drive. SCSI does have an advantage here in that you can get 15K RPM drives where the current top end for SATA drives is 10K ( and the drives are not that large ~70GB ). Yes, you can make an array with SCSI and have better performance than SATA, but not at the same cost. Even similar size and speed arrays with SCSI will cost more.

However, with striping and a large enough array of disks you could configue stripes (for transfer speed) with mirroring (for safety) and maximze the performance of the array (at the cost storage space). Most importantly, you can achieve comparable performances with much lower costs. I don't know where they got the SNAP device, but if it is not a purchase but a donation; then you cannot expect them to look the gift horse in the mouth.

The place where I expect that most of the performance will come is in the transition from software RAID to hardware RAID. You can see one or two orders of magnitude of performance improvement.

As far as Oracle, that might have been a better choice from a single project perspective. But Oracle would have to commit to supplying all now and future BOINC projects, in perpetuity. Not likely. Using Open Source with the licence terms makes a lot of sense in that almost all research is done on a shoestring budget and with donations.

Tipping Redhats again to Admin/Dev. Keep
the updates coming, musch, much appreciation

Xi3
<a> [/url]

Tipping Redhats again to Admin/Dev. Keep
the updates coming, much, much appreciation

Xi3
<a> [/url]

Sorry for the double post...