Message boards :
Number crunching :
Panic Mode On (20) Server problems
Message board moderation
Previous · 1 . . . 12 · 13 · 14 · 15
Author | Message |
---|---|
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
I myself still find that a decent controller and RAID 5 is my favorite for DB access/speed in most applications. Might be of interest- SSD is the way to go. Price is no object? And buggered if i can find a reveiw i read recently- RAID 10, 5 & 6 compared. Grant Darwin NT |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
Software raid doesn't have that limitation. It's better when budgets don't allow you to stock enough spares. And when every scrounged/handmedown/engineering test server comes with a different card! I said 'historical': they may have started accepting that hardware is the way to go - especially when it's pre-built into a complete server. But I think they're still a bit nervous about failure/recovery. Did you read Matt's saga about how he has to reboot Thumper? |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
Reporting does not happen on a schedule, rather it happens at the first of: OK, which WIKI is it missing from. I am pretty certain it is in the unofficial WIKI. BOINC WIKI |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
And buggered if i can find a reveiw i read recently- RAID 10, 5 & 6 compared. A quick Google brought up http://blogs.zdnet.com/storage/?p=162&tag=nl.e539. |
TCP JESUS Send message Joined: 19 Jan 03 Posts: 205 Credit: 1,248,845 RAC: 0 |
I myself still find that a decent controller and RAID 5 is my favorite for DB access/speed in most applications. Once SSD's can multitask a little better and the price comes down a bit, I will jump on that ship. I currently try to limit myself to $200/drive x 3.....otherwise, one can get VERY carried away. $200 is the magic number, or $600 for 3 drives anyways, so that I can run a RAID 5 on my desktop. The VelociRaptors are a decent performance increase over the last batch of Raptors and I am quite happy with their performance 99.9% of the time. SSD is probably another year or more down the road for me. I build a new semi-monster once a year or year-and-a-half, so perhaps 1 to 2 systems down the road for me, as my current box is less than a month old currently (Seti ID: 5000725). Allan. I am TCP JESUS...The Carpenter Phenom Jesus....and HAMMERING is what I do best! formerly known as...MC Hammer. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Software raid doesn't have that limitation. It's better when budgets don't allow you to stock enough spares. Yes, I did. When you stop and think about it, it's pretty amazing how well things work. The Sun V40z SETI had was a prototype, and apparently, one-of-a-kind as the manufactured product was different when it came time to get parts. Most everything else is either a hand-me-down or an engineering model. In an ideal world, all of the servers would be similar (maybe just two motherboards) and as many parts as possible would be 100% interchangeable. That's what bothers me most when someone says "mismanagement" -- they're doing an amazing job holding things together as-is. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
And Thumper is a Sun X4500: anyone not familiar with those specs, read and drool. "Up to forty-eight hot-swappable disk drives for a total of 48 TB of raw storage." The original Thumper was a prototype: the current replacement may be a bit more maintainable, but I wouldn't want to be the one doing the maintaining. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
And buggered if i can find a reveiw i read recently- RAID 10, 5 & 6 compared. Nah, that's not it. This one was very recent. Although i do remember that bit of doom & gloom from a while back. Dan had a nice article debunking the whole thing. Grant Darwin NT |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
Wow, missed a lot of discussion here. It's only been half a day since I last read the posts. anyway, aggregating the uploads into one transmission is not a bad idea, but one thing that could be done to reduce the CPU load on the server is instead of gzip or something that compresses, why not just a tarball, or some other form of concatenation? Wouldn't require much CPU load at all to break it back apart at the other end. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
anyway, aggregating the uploads into one transmission is not a bad idea, but one thing that could be done to reduce the CPU load on the server is instead of gzip or something that compresses, why not just a tarball, or some other form of concatenation? Wouldn't require much CPU load at all to break it back apart at the other end. This is XML(ish): add the result file name to the header, and just concatenate them. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
Once SSD's can multitask a little better I suggest you check those IO benchmark figures. Mechanical HDDs aren't even in the race. Not even 15,000 RPM SATA enterpise drives. and the price comes down a bit That's the biggest hurdle at the moment. But if you need the performance... The VelociRaptors are a decent performance increase over the last batch of Raptors and I am quite happy with their performance 99.9% of the time. Yep, big boost on the previous drives. Will probably be my drive of choice when i get a new system later this year. Maybe, hopefully, with luck. Grant Darwin NT |
Nicolas Send message Joined: 30 Mar 05 Posts: 161 Credit: 12,985 RAC: 0 |
anyway, aggregating the uploads into one transmission is not a bad idea, but one thing that could be done to reduce the CPU load on the server is instead of gzip or something that compresses, why not just a tarball, or some other form of concatenation? Wouldn't require much CPU load at all to break it back apart at the other end. That would be a project-specific solution... SETI by itself cannot cause multiple result output files to get concatenated before uploading, that would need BOINC client changes. And if it's in the BOINC client, it should work for any project, including those without XML-ish output files. Contribute to the Wiki! |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
And Thumper is a Sun X4500: anyone not familiar with those specs, read and drool. "Up to forty-eight hot-swappable disk drives for a total of 48 TB of raw storage." The power consumption is truly impressive. 240v at 10a. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
Wow, missed a lot of discussion here. It's only been half a day since I last read the posts. Sorry, we've been going off-topic: this is supposed to be the 'Panic' thread, and it's been at least an hour since we had a good panic. How about: no new "tapes" were fetched out of storage today, and the lab is now on overtime after the Friday shift ended (as if they care). Will 111 channels last until Monday? |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
anyway, aggregating the uploads into one transmission is not a bad idea, but one thing that could be done to reduce the CPU load on the server is instead of gzip or something that compresses, why not just a tarball, or some other form of concatenation? Wouldn't require much CPU load at all to break it back apart at the other end. If we're talking about a truly BOINC-centric solution, then it has to be generic. That means we don't get it until enough BOINC clients are upgraded, which has been raised before as an objection. In the short term, we're stuck. A SETI-only solution could be to bundle work units, and change the science application to do "bundles" -- crunch more than one WU, add 'em all to one result file, and let BOINC assign "bundles" and think they're single work units. Bundle MB, don't bundle AP, and put a half-dozen work units in each MB bundle. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
How about: no new "tapes" were fetched out of storage today, and the lab is now on overtime after the Friday shift ended (as if they care). I wouldn't be surprised to find out that the BOINC staff is all "salaried-exempt" and therefore don't get overtime, or have precise set hours. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
How about: no new "tapes" were fetched out of storage today, and the lab is now on overtime after the Friday shift ended (as if they care). Well, I was speaking figuaratively: I agree, any 'lab hours' are probably entirely notional, and even if they exist, SETI staff ignore them by working from home (pssst: don't tell Angela). But I think Matt, at least, tries to get basic chores like fetching data done in daylight: and as things stand, I predict a data outage before the weekend is out. |
Nicolas Send message Joined: 30 Mar 05 Posts: 161 Credit: 12,985 RAC: 0 |
But I think Matt, at least, tries to get basic chores like fetching data done in daylight: and as things stand, I predict a data outage before the weekend is out. Remember Matt is even on vacation now... Contribute to the Wiki! |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
But I think Matt, at least, tries to get basic chores like fetching data done in daylight: and as things stand, I predict a data outage before the weekend is out. And so am I - for a few hours, at least: bedtime, UK time. Thanks for a stimulating evening's conversation, everyone. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.