Re: [Bacula-users] Time for change
On Thu, 18 Dec 2008, Jesper Krogh wrote: I'm multiplexing anything up to 20 jobs at a time. To ensure that small incremental and diff jobs are dumped in one hit and to ensure that full backups are laid in as large chunks as possible, this is the kindof size which is required. Are you having that many clients or is it actually larger volumes that have been split up in smaller filesets? Mostly larger volumes broken into 1Tb filesets. There are about 150 desktop boxes being backed up but their load is minor. Yes, I use spooling even for our +2TB volumes. I hope you've benchmarked restore times. Between the sheer volume of data and the possibilities of several million files, we settled on 1Tb as a workable unit for current technology and even that's had trouble when restoring some filesystems (10 million file restores make Bacula-dir unhappy!) 250Tb startup, 1Pb within 18 months, and growing past that. Filesizes measured in tens of Mb apiece. Still over Gigabit netwok or do you have better infrastructure? Right now the links are Gb, with some LACP overlaying being done, but as this larger fileset comes online most stuff will be moved to 8Gb/s fibre (IP over F/O) and 10Gb/s ethernet. Prices on this gear (especially switches!) are falling fast so it makes sense to not deploy until actually needed. -- ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Hi, 18.12.2008 06:50, Jesper Krogh wrote: Alan Brown wrote: Jesper Krogh wrote: I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I still have got to see a reasonable priced SSD' disk that can deliver around 100MB/s both ways at the same time. There aren't any mechanical disks which can do it either. Which is why I'm not trying to do that - replacing 4 RAID0 mechanical disks with 4 SSDs will provide similar sustained throughput to the mechanical RAID0, but provide _much_ better performance for anything where the mechanical disks had head seeking involved - such as multiple simultaneous input/output streams to LTO drives. http://www.slashgear.com/samsung-64gb-ssd-performance-benchmarks-278717/ Make sure you compare apples with something remotely looking like apples. The ONLY SSds which are suitable fo this kind of use are SLCs, not MLCs Ok. I havent spend enough time in that area. In short, MLC's are cheaper but slower. There are claims they are less reliable, i.e. allow lesser rewrites. I have beefed up my director with sufficient amount of memory and mounted it as a ramdisk for spooling. That doesn't impose any limitations on the 2 LTO3 drives attached. How much do you regard as sufficient? 100-200Gb ram and systems capable of addressing that amount of memory are still far more expensive than a stack of flash drives, else I'd use them. But do you need to spool a complete tape? In order to avoid doing evil stuff to you tape drive, much less is sufficient. Well, Alan probably runs jobs spanning more than one tape. Also, he's running jobs concurrently. Now, if he wants to avoid interleaving of different jobs, he needs to spool complete jobs, and probably several jobs of more than LTO3 capacity - thus the need for a rather large spool space. My concern isn't just backup run time. So you'd like to spool a complete Job? Whats you average job-size? (mine is less than 8GB) if its larger, we just need a period of despooling (I'd love to have concurrent spooling/despooling in bacula). Currenly I use at most 32GB for spooling area, with a Job Concurrency at 4 and 2 tape drives. I believe Alan is handling data sets a bit larger... When doing large backups(full+archive) I mostly have one (or two) drives in action at the same time while spooling to disk with 2(or 3) threads at the same time. The LTO3 drives far outperform our network speed (1gbit). Transfer to tapes are in the range from 60MB/s to 100MB/s (and I unfortunately have no idea why they spread that much). In total numbers we're around 25TB to disk/month with monthly full and daily incrementals. Concerned about job run time, its my impression that spool space only speeds up incremental/differential. Depends - the important thing is that spooling allows multiple concurrent jobs without or with less interleaving. Restore times are also important and having a tape read back 1Gb, then seek, then pull back another 1Gb (or even 10Gb) is a significant penalty over reading larger blocks when worst-case 75Tb+ restores are considered (25-60 days on 2 drive LTO2, dpeending on the directory structures being restored.) Whats the time consuming part in this? Seeking on tapes? Neither SSD's or memory will change that. Moving tapes, loading and unloading times. Seek times add to that, but are not that bad with LTO. AFAIK the spooling area is only used when going TO tape, not FROM tape. Right, but if you manage to have all jobs in one continuos block on tape, you minimize the number of volumes needed per job, and thus save much of the volume handling time. And spooling doesnt need any form for persistence, so its fine that its gone after reboot Indeed. If it was practical I'd use ramdisks. Right now it's not. Apart from the cost factor there is very little hardware which can address more than 128Gb of Dram. There are RAM arrays which are setup to operate as F/O scsi devices, but these are currently silly money as they're marketed at the world of high end, high cost databases. Again, I assume we're talking about spooling space, then try to think about if you need that much. I believe Alan thought about it (and I hope my reasoning above is correct :-) In 12 months time that may change, Ram is always falling in price - but Flash drive pricing is falling faster, performance/durability is rising at the same time and there isn't the same issue with massive address ranges as it just looks like more disk, vs having to change out entire servers at $20k a time if RAM limits are reached. I'm not just looking at the issue of my current setup. Projects are already pencilled onsite which will
Re: [Bacula-users] Time for change
On Thu, 18 Dec 2008, Jesper Krogh wrote: 100-200Gb ram and systems capable of addressing that amount of memory are still far more expensive than a stack of flash drives, else I'd use them. But do you need to spool a complete tape? In order to avoid doing evil stuff to you tape drive, much less is sufficient. I'm multiplexing anything up to 20 jobs at a time. To ensure that small incremental and diff jobs are dumped in one hit and to ensure that full backups are laid in as large chunks as possible, this is the kindof size which is required. My concern isn't just backup run time. So you'd like to spool a complete Job? Whats you average job-size? (mine is less than 8GB) Full backups run 500Gb to 1Tb apiece, the average nightly incremental is about 80-150Gb - multiplied out by 90 filesystems. You get the idea. Concerned about job run time, its my impression that spool space only speeds up incremental/differential. It does at the moment. HOWEVER if spooling isn't used then jobs are interleaved on the tape at record time, resulting in massive shoeshining on restores and restore throughput rates measured in kb/Sec instead of 25-40Mb/sec Whats the time consuming part in this? Seeking on tapes? Neither SSD's or memory will change that. AFAIK the spooling area is only used when going TO tape, not FROM tape. Using large spool areas increases the size of the chunks dropped to tape and thus allows the tapes to stream for longer periods. ONE of my backups might span 5 LTO2s Can you give some numbers, so we have a feeling about the sizes you talk about? 250Tb startup, 1Pb within 18 months, and growing past that. Filesizes measured in tens of Mb apiece. -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Thanks for the elaborate reply. Just a few more querious questions. Alan Brown wrote: On Thu, 18 Dec 2008, Jesper Krogh wrote: 100-200Gb ram and systems capable of addressing that amount of memory are still far more expensive than a stack of flash drives, else I'd use them. But do you need to spool a complete tape? In order to avoid doing evil stuff to you tape drive, much less is sufficient. I'm multiplexing anything up to 20 jobs at a time. To ensure that small incremental and diff jobs are dumped in one hit and to ensure that full backups are laid in as large chunks as possible, this is the kindof size which is required. Are you having that many clients or is it actually larger volumes that have been split up in smaller filesets? My concern isn't just backup run time. So you'd like to spool a complete Job? Whats you average job-size? (mine is less than 8GB) Full backups run 500Gb to 1Tb apiece, the average nightly incremental is about 80-150Gb - multiplied out by 90 filesystems. You get the idea. Absolutely. Concerned about job run time, its my impression that spool space only speeds up incremental/differential. It does at the moment. HOWEVER if spooling isn't used then jobs are interleaved on the tape at record time, resulting in massive shoeshining on restores and restore throughput rates measured in kb/Sec instead of 25-40Mb/sec Yes, I use spooling even for our +2TB volumes. Whats the time consuming part in this? Seeking on tapes? Neither SSD's or memory will change that. AFAIK the spooling area is only used when going TO tape, not FROM tape. Using large spool areas increases the size of the chunks dropped to tape and thus allows the tapes to stream for longer periods. ONE of my backups might span 5 LTO2s I can have the same for LTO3, but that due to a few large jobs running full backups concurrenly. Can you give some numbers, so we have a feeling about the sizes you talk about? 250Tb startup, 1Pb within 18 months, and growing past that. Filesizes measured in tens of Mb apiece. Still over Gigabit netwok or do you have better infrastructure? -- Jesper -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
On Wed, 10 Dec 2008, Josh Fisher wrote: I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I've been wondering about those, but was thinking it would be better used as DB storage for Mysql, where I/Os per second is going to be more important than throughput. I see your point, though, where concurrent jobs are running. There's no way in the world that I'd put any Database on RAID0, for pretty obvious reasons. In the case of Mysql, the biggest win will be simply having enough memory to prevent the system from paging and having the mysql system tuned appropriately to use it. Past that, good fast arrays will help, but not as much as having sufficient memory in the first place (where sufficient may be 16Gb or more for large databases) AB -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Alan Brown wrote: On Tue, 9 Dec 2008, Brian Debelius wrote: John Drescher wrote: In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. So what do you think a reasonable cpu for bacula would be? I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I still have got to see a reasonable priced SSD' disk that can deliver around 100MB/s both ways at the same time. http://www.slashgear.com/samsung-64gb-ssd-performance-benchmarks-278717/ I have beefed up my director with sufficient amount of memory and mounted it as a ramdisk for spooling. That doesn't impose any limitations on the 2 LTO3 drives attached. And spooling doesnt need any form for persistence, so its fine that its gone after reboot. -- Jesper -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Jesper Krogh wrote: I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I still have got to see a reasonable priced SSD' disk that can deliver around 100MB/s both ways at the same time. There aren't any mechanical disks which can do it either. Which is why I'm not trying to do that - replacing 4 RAID0 mechanical disks with 4 SSDs will provide similar sustained throughput to the mechanical RAID0, but provide _much_ better performance for anything where the mechanical disks had head seeking involved - such as multiple simultaneous input/output streams to LTO drives. http://www.slashgear.com/samsung-64gb-ssd-performance-benchmarks-278717/ Make sure you compare apples with something remotely looking like apples. The ONLY SSds which are suitable fo this kind of use are SLCs, not MLCs I have beefed up my director with sufficient amount of memory and mounted it as a ramdisk for spooling. That doesn't impose any limitations on the 2 LTO3 drives attached. How much do you regard as sufficient? 100-200Gb ram and systems capable of addressing that amount of memory are still far more expensive than a stack of flash drives, else I'd use them. My concern isn't just backup run time. Restore times are also important and having a tape read back 1Gb, then seek, then pull back another 1Gb (or even 10Gb) is a significant penalty over reading larger blocks when worst-case 75Tb+ restores are considered (25-60 days on 2 drive LTO2, dpeending on the directory structures being restored.) And spooling doesnt need any form for persistence, so its fine that its gone after reboot Indeed. If it was practical I'd use ramdisks. Right now it's not. Apart from the cost factor there is very little hardware which can address more than 128Gb of Dram. There are RAM arrays which are setup to operate as F/O scsi devices, but these are currently silly money as they're marketed at the world of high end, high cost databases. In 12 months time that may change, Ram is always falling in price - but Flash drive pricing is falling faster, performance/durability is rising at the same time and there isn't the same issue with massive address ranges as it just looks like more disk, vs having to change out entire servers at $20k a time if RAM limits are reached. I'm not just looking at the issue of my current setup. Projects are already pencilled onsite which will increase storage demands by a factor of 20 from the current size within 12 months and I have to try and be ready to back that data up. -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Alan Brown wrote: Jesper Krogh wrote: I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I still have got to see a reasonable priced SSD' disk that can deliver around 100MB/s both ways at the same time. There aren't any mechanical disks which can do it either. Which is why I'm not trying to do that - replacing 4 RAID0 mechanical disks with 4 SSDs will provide similar sustained throughput to the mechanical RAID0, but provide _much_ better performance for anything where the mechanical disks had head seeking involved - such as multiple simultaneous input/output streams to LTO drives. http://www.slashgear.com/samsung-64gb-ssd-performance-benchmarks-278717/ Make sure you compare apples with something remotely looking like apples. The ONLY SSds which are suitable fo this kind of use are SLCs, not MLCs Ok. I havent spend enough time in that area. I have beefed up my director with sufficient amount of memory and mounted it as a ramdisk for spooling. That doesn't impose any limitations on the 2 LTO3 drives attached. How much do you regard as sufficient? 100-200Gb ram and systems capable of addressing that amount of memory are still far more expensive than a stack of flash drives, else I'd use them. But do you need to spool a complete tape? In order to avoid doing evil stuff to you tape drive, much less is sufficient. My concern isn't just backup run time. So you'd like to spool a complete Job? Whats you average job-size? (mine is less than 8GB) if its larger, we just need a period of despooling (I'd love to have concurrent spooling/despooling in bacula). Currenly I use at most 32GB for spooling area, with a Job Concurrency at 4 and 2 tape drives. When doing large backups(full+archive) I mostly have one (or two) drives in action at the same time while spooling to disk with 2(or 3) threads at the same time. The LTO3 drives far outperform our network speed (1gbit). Transfer to tapes are in the range from 60MB/s to 100MB/s (and I unfortunately have no idea why they spread that much). In total numbers we're around 25TB to disk/month with monthly full and daily incrementals. Concerned about job run time, its my impression that spool space only speeds up incremental/differential. Restore times are also important and having a tape read back 1Gb, then seek, then pull back another 1Gb (or even 10Gb) is a significant penalty over reading larger blocks when worst-case 75Tb+ restores are considered (25-60 days on 2 drive LTO2, dpeending on the directory structures being restored.) Whats the time consuming part in this? Seeking on tapes? Neither SSD's or memory will change that. AFAIK the spooling area is only used when going TO tape, not FROM tape. And spooling doesnt need any form for persistence, so its fine that its gone after reboot Indeed. If it was practical I'd use ramdisks. Right now it's not. Apart from the cost factor there is very little hardware which can address more than 128Gb of Dram. There are RAM arrays which are setup to operate as F/O scsi devices, but these are currently silly money as they're marketed at the world of high end, high cost databases. Again, I assume we're talking about spooling space, then try to think about if you need that much. In 12 months time that may change, Ram is always falling in price - but Flash drive pricing is falling faster, performance/durability is rising at the same time and there isn't the same issue with massive address ranges as it just looks like more disk, vs having to change out entire servers at $20k a time if RAM limits are reached. I'm not just looking at the issue of my current setup. Projects are already pencilled onsite which will increase storage demands by a factor of 20 from the current size within 12 months and I have to try and be ready to back that data up. Can you give some numbers, so we have a feeling about the sizes you talk about? -- Jesper -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
On Tue, 9 Dec 2008, Jeff Kalchik wrote: *NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. That hasn't been true in Linux for a number of years. Given a modern machine (less than 2-5 years old) _and sufficient ram_, Linux software raid is a great deal faster than most cheap dedicated hardware controllers without having any noticable CPU load. HOWEVER: as noted by others, Windows software raid performance is very poor AB -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
On Tue, 9 Dec 2008, Lukasz Szybalski wrote: I'm not sure about ubuntu, but if you are installing debian you can setup raid right there, during installation. It was fairly easy. After that bacula setup and you are ready to go. Ubuntu _is_ Debian, more or less. There's plenty of configurability in the drive setup and I setup raid1 on my personal machine in about 5 minutes during the installation procedure (it took longer to navigate the menus than to configure the raid array) Do people have recommendation on that? I actually would like to add a 5th 500gb drive to my raid but I have no space for it. If you have a PCIe x1 connector then there are plenty of dumb dual SATA adaptors available. Quad-SATA can be found with some difficulty (Promise make one, but it can be hard to obtain) If you have PCIe x4, then smarter cards are available, up to 16 SATA/SAS ports. Alternatively (or as well as...), something like 1 or more LaCie Biggest or Quadra external eSATA arrays may be worth looking at. http://www.lacie.com/uk/products/product.htm?pid=5 http://www.lacie.com/uk/products/product.htm?pid=10458 AB -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
On Tue, 9 Dec 2008, Brian Debelius wrote: John Drescher wrote: In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. So what do you think a reasonable cpu for bacula would be? I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. (*) The SD array is more than fast enough to drive a single LTO4 drive for one backup, but when driving 2 LTO2 drives and running up to 6 simultaneous backups, the heads thrash quite hard and throughput slows by 50%. Installing 15krpm drives over the existing 7500rpm ones won't help as that only reduces seek time by about 10-15%. This machine also has the Bacula director and Mysql on it, Mysql is the main memory hog, with the director coming in as the CPU hog. The biggest single problem is memory - due to the database size (40Gb bacula dumps) and the existing desktop-class motherboard is maxed out for that. AB -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Well right now I have an old Asus k8v-se which has an Athlon 64 processor. So from the conversation, it seems that it should be enough for software raid. But the basboard has only 2 sata ports, and 2 raid ports, and 5 PCI slots. If I wanted more disks I would have to add a controller, or two. Would the PCI bus be a bottleneck? -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
On Wed, Dec 10, 2008 at 10:57 AM, Brian Debelius [EMAIL PROTECTED] wrote: Well right now I have an old Asus k8v-se which has an Athlon 64 processor. Some of my raid older servers are very similar to this motherboard with Athlon64 3000 chips. So from the conversation, it seems that it should be enough for software raid. But the basboard has only 2 sata ports, and 2 raid ports, and 5 PCI slots. If I wanted more disks I would have to add a controller, or two. Use the raid ports in sata (probably called jbod) mode. Would the PCI bus be a bottleneck? From my testing with a similar using the PCI bus will limit the total bandwidth to the drives that are hung off that to around 100MB/s. So for a 6 drive array using 4 mobo ports I was getting around 200MB/s reads. I am not sure about writes. I think I have the same box still available, so I probably can run some numbers but that will have to be later as I have a meeting... John -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Lukasz Szybalski wrote: On Tue, Dec 9, 2008 at 1:04 PM, John Drescher [EMAIL PROTECTED] wrote: So what do you think a reasonable cpu for bacula would be? Depends on what level of performance you are looking for. My director is a 2 processor 2GHz opteron machine (circa 2003) with 4 GB of memory and 18 or so x 250 GB SATA 1 drives in raid 6. My main storage daemon is on a second machine with the same cpus and ram but only 5 x 250 GB SATA drives and the storage is a 2 drive x 24 slot LTO 2 autochanger. Right now the bacula database is on that machine but it has been on yet another machine. On top of that a typical SATA fileserver will be a desktop class 2.5GHz Athlon X2 with 4 or 8GB of memory and 5 to 10 750 GB Segate SATA2 drives in raid 6 using mostly the motherboard sata ports. This type of machine will net 300MB/s writes and even higher reads in xfs. At 300 MB/s writes the cpu usage never goes above 7% for the raid under normal conditions. Replacing a drive will make it go to around 20% of 1 cpu until the drive is re synced. I would go with software raid, no need to spend extra money unless you already have one, then go ahead. See/compare some performance statistics and functionality like swap drives: http://lucasmanual.com/mywiki/Bacula#head-3ad60a172345da090bfbaa9488f8f9e6b88722b8 Also if you are going with hardware raid will you be using SAS which supports (SAS + SATA connectors) or just SATA? I'm not sure about ubuntu, but if you are installing debian you can setup raid right there, during installation. It was fairly easy. After that bacula setup and you are ready to go. It would seem to me anything over 2004 pentium 4 should be enough. The biggest question is: does the pc you want to purchase has enough space to fit let say 4 hdd and does the motherboard supports these many hard drives connectors. Do people have recommendation on that? I actually would like to add a 5th 500gb drive to my raid but I have no space for it. Thanks, Lucas I use an Antec 900 case as has 3 x 5.25inch triple bays, so thats 9 disks plus then 3 single bays, so a dvd bay and 2 boot disk bays in which I have placed a dual scsi bay unit...(There is now an antec 1200 case which has an extra triple bay I think). My Motherboard is an ASUS p35 with 6 sata plus 2 esataI run the dvd off the ide channel..one triple bay was removed and a 5 bay scsi tray unit installed, so I have 2 scsi disks as the OS raid1 (in the dual scsi bay) and three scsi R5 as the primary data (so I can add two more but scsi are too expensive)...then another scsi raid card powers the 6 bay external scsi disk pack, that is a R5(4+1) of 146s for more data the 6th is my restore disk..(146gb scsi). I plan to have 2 x 1tB for data as expansion as needed on the dual esata and the other 6 internal internal could be (2 at the moment) 1Tb disks mounted one at a time to backup to...Tape back up is to a scsi powered 2 x DLTiv 75 unit...I dont use it much though too manual...considering using single external 1tb fwire attached disks instead. There is also a Corsair 3 x 5.25inch bay unit that holds 4 disk per unit, so I could run 13 internal disks, 2 external disks and 2 for the OS... :) On the Antec, each bay has a 120mm fan in front of it and the case has a 120mm fan at the back plus a 200mm fan at the top rearI run them all at slow speed and there is little (fan) noise, the scsi disks whine noticably though (10,000rpm). At the moment I have a 520HX Corsair PSU, high efficiency and a 5 year warrantee...it sits at the bottom of the case so air gets drawn over the pci[e] slots, it seems to run at min fan speed so cant be stretched the raid cards have the spare ASUS chipset fans over them attached to the motherboard...looking at the 600HX unit when I add more disks though (the psu will then also be 3 years old)...I run a pci graphics card so the 2 x 16 pci-e are free for sata/sas controllers... The Antec 900 I got on special...pleased with it...should last years with hardware refreshes every so often...oh the CPU is a dual core cerelon at 1.6Ghz and 2.5gb ram about to become 4, I can fit 8 but 2.5 seems enough, I just happen to have 2 more sitting around so it goes in when I next add more disks. regards Steven -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Alan Brown wrote: On Tue, 9 Dec 2008, Brian Debelius wrote: John Drescher wrote: In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. So what do you think a reasonable cpu for bacula would be? I'm running spooling on a 4 drive software raid0 quite happily on a 4Gb 3GHz P4D machine. The limiting factors are disk head seek time(*) when running concurrent backups to 2 LTO2 drives and available SATA ports. Because of that I'm considering dropping in solid state disks. I've been wondering about those, but was thinking it would be better used as DB storage for Mysql, where I/Os per second is going to be more important than throughput. I see your point, though, where concurrent jobs are running. (*) The SD array is more than fast enough to drive a single LTO4 drive for one backup, but when driving 2 LTO2 drives and running up to 6 simultaneous backups, the heads thrash quite hard and throughput slows by 50%. Installing 15krpm drives over the existing 7500rpm ones won't help as that only reduces seek time by about 10-15%. This machine also has the Bacula director and Mysql on it, Mysql is the main memory hog, with the director coming in as the CPU hog. The biggest single problem is memory - due to the database size (40Gb bacula dumps) and the existing desktop-class motherboard is maxed out for that. AB -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Hello, I am currently running Ubuntu. The bacula director and sd are on this box. I have bacula configured to spool to this box, and then it goes directly to tape. I want to change how I am backing up. I would like to have some period of time of disk backups, and have them moved/migrated/copied to tape. Should I consider linux software raid. or use a hardware raid adapter? If software does the cpu choice have much of an impact? I am thinking of configuring the disk as follows: One OS and database disk, maybe a mirror Two disks in raid 0 for spooling, and recovering Three or more disks in raid 5 for the sd storage. Thoughts or comments on this? *NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. Jeff Kalchik -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
*NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. In windows software raid performance sucks so you need a hardware card there. John -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Normally I would say, 1) Performance generally ie cpu hit is not a huge issue IMHO, especially with todays dual and quad core cpus...and of course ram is dirt cheap. 2) For me the big issues is Ive lost software raid sets from power failures for a backup partition that's probably no biggeeunless of course your data partition and your backup partition are both software raided...and you loose both at once... From whats posted the backup system is seperate from the production servers, in which case software raid would be fine IMHO. If its lost, the data should be available on tape and on the production serversthe proviso is that you backup to disk overnight and then immediately spool to tape..which could be during the day. regards Steven Jeff Kalchik wrote: Hello, I am currently running Ubuntu. The bacula director and sd are on this box. I have bacula configured to spool to this box, and then it goes directly to tape. I want to change how I am backing up. I would like to have some period of time of disk backups, and have them moved/migrated/copied to tape. Should I consider linux software raid. or use a hardware raid adapter? If software does the cpu choice have much of an impact? I am thinking of configuring the disk as follows: One OS and database disk, maybe a mirror Two disks in raid 0 for spooling, and recovering Three or more disks in raid 5 for the sd storage. Thoughts or comments on this? *NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. Jeff Kalchik -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
So what do you think a reasonable cpu for bacula would be? Depends on what level of performance you are looking for. My director is a 2 processor 2GHz opteron machine (circa 2003) with 4 GB of memory and 18 or so x 250 GB SATA 1 drives in raid 6. My main storage daemon is on a second machine with the same cpus and ram but only 5 x 250 GB SATA drives and the storage is a 2 drive x 24 slot LTO 2 autochanger. Right now the bacula database is on that machine but it has been on yet another machine. On top of that a typical SATA fileserver will be a desktop class 2.5GHz Athlon X2 with 4 or 8GB of memory and 5 to 10 750 GB Segate SATA2 drives in raid 6 using mostly the motherboard sata ports. This type of machine will net 300MB/s writes and even higher reads in xfs. At 300 MB/s writes the cpu usage never goes above 7% for the raid under normal conditions. Replacing a drive will make it go to around 20% of 1 cpu until the drive is re synced. -- John M. Drescher -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
John Drescher wrote: In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. So what do you think a reasonable cpu for bacula would be? -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Dear Jeff Kalchik, In message [EMAIL PROTECTED] you wrote: *NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. He. So what does me a h/w RAID controller good when I find myself having the system in 90% I/O wait? On a file server (that does not perform other services which would utilize the CPU) s/w RAID gives much better performance - at least in the usage szenarios we see (which means serving a *big* number of relatively small files). Best regards, Wolfgang Denk -- DENX Software Engineering GmbH, MD: Wolfgang Denk Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: [EMAIL PROTECTED] Generally speaking, there are other ways to accomplish whatever it is that you think you need ... - Doug Gwyn -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Brian Debelius wrote: John Drescher wrote: In linux, I find this to be completely wrong. I have 15TB of software raid 6 and the most load that it puts on the cpu is around 7% and these are raid arrays that net over 200MB/s writes on single core systems that are 3 or so years old. So what do you think a reasonable cpu for bacula would be? depends on what you are backing up. If its network backup your network is likely to be the first limit... I have a dual core 1.6Ghz cerelon with 2.5gb of ram, my bottleneck is my scsi disks.so if you have money to spend I'd suggest the disks and disk controllers are the priority. regards Steven -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
Re: [Bacula-users] Time for change
Jeff Kalchik wrote: Hello, I am currently running Ubuntu. The bacula director and sd are on this box. I have bacula configured to spool to this box, and then it goes directly to tape. I want to change how I am backing up. I would like to have some period of time of disk backups, and have them moved/migrated/copied to tape. Should I consider linux software raid. or use a hardware raid adapter? If software does the cpu choice have much of an impact? I am thinking of configuring the disk as follows: One OS and database disk, maybe a mirror Two disks in raid 0 for spooling, and recovering Three or more disks in raid 5 for the sd storage. Thoughts or comments on this? *NEVER* use software RAID if you can avoid it. Software RAID puts a pretty good hit right on your CPU. Perhaps this applies to some versions of RAID and some implementations of software RAID. It certainly does not apply to simple RAID (such as RAID-1) and fine implementations of software such as FreeBSD's gmirror. I use gmirror on all my machines at home. That said, I also have hardware RAID-1. -- SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ ___ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
