Re: How git performs when you throw all of Debian at it
On Mon, Sep 02, 2013 at 11:27:33AM +0200, Tollef Fog Heen wrote: > ]] Luca Filipozzi > > On Sat, Aug 31, 2013 at 12:40:08AM +0200, Michael Stapelberg wrote: > > > Luca Filipozzi writes: > > > > Why do you say that when you haven't even asked? > > > Because I thought the answer was going to be “not in the Linux kernel, > > > no chance”. > > We also run kfreebsd, with some challenge, but we have it. > It's not really an option for codesearch since codesearch needs systemd, > though. Also I'm not sure how ZFS would solve the basic constraints of hardware and be a magic bullet. The index can already be compressed in the application layer. Deduplication adds the need to use a lot of RAM for the dedup tables. The L2ARC of ZFS is commonly put on SSDs we don't have. So it'd basically yield the need to add a lot of RAM to the VM for a reasonably sized ZFS in-memory cache and the dedup tables for probably not much gain in avoided seeks (if git delta dedup doesn't even find that much, why would block-level debug work better). It's not that it wasn't proposed to instead throw RAM onto more page cache (which could yield similar savings), which was equally considered a bad choice. Kind regards Philipp Kern signature.asc Description: Digital signature
Re: How git performs when you throw all of Debian at it
On Sun, Sep 01, 2013 at 09:07:47PM +0100, Stephen Gran wrote: > luca is not asking, "why aren't you using the new shiny", he's asking, > "why hasn't a proposal for a project that does string searching used one > of the already available, off the shelf string searching programs". > > Asking someone why they didn't choose off the shelf tools to do a job > that is largely a solved problem is a reasonable thing to do, especially > when being asked to support new, custom code that is certain to come > with its own quirks, bugs, and security issues. > > I think the answer here is, "more code could be reused by borrowing work > from google" - the NIH syndrome appears to be google's, rather than > Michael's, if I understand correctly. That's probably a fine answer, > but the question also needed to be asked. Fair point. My personal guess is that Solr's support for regex search was not up to par at that point. codesearch is not about simple string searching, but about regex as well, which was long something that wasn't very performant because it couldn't be indexed efficiently. The few bits I read up about Solr + regex seem to imply that your indices somewhat explode in size (because it has to look at more context than for simple strings), but that may or may not be true anymore. Kind regards Philipp Kern -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130902125357.ga10...@hub.kern.lc
Re: DSA, kFreeBSD, and the Singularity (was: How git performs when you throw all of Debian at it)
On Sun, 01 Sep 2013, Steven Chamberlain wrote: > I can only recall one wishlist bug from DSA at the moment which is > #711247 requesting pflogd. I'd love to hear more wishlist kfreebsd > ideas from DSA. syslog-ng on kfreebsd doesn't properly reconnect to logservers after they went a way for a while or the network dropped. Would be nice if that could be fixed. -- | .''`. ** Debian ** Peter Palfrader | : :' : The universal http://www.palfrader.org/ | `. `' Operating System | `-http://www.debian.org/ -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130902121637.gc14...@anguilla.noreply.org
Re: How git performs when you throw all of Debian at it
Tollef Fog Heen writes ("Re: How git performs when you throw all of Debian at it"): > ]] Luca Filipozzi > > We also run kfreebsd, with some challenge, but we have it. > > It's not really an option for codesearch since codesearch needs systemd, > though. Why is that, JOOI ? Ian. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/21028.26486.458404.471...@chiark.greenend.org.uk
Re: How git performs when you throw all of Debian at it
]] Luca Filipozzi > On Sat, Aug 31, 2013 at 12:40:08AM +0200, Michael Stapelberg wrote: > > Luca Filipozzi writes: > > > Why do you say that when you haven't even asked? > > Because I thought the answer was going to be “not in the Linux kernel, > > no chance”. > > We also run kfreebsd, with some challenge, but we have it. It's not really an option for codesearch since codesearch needs systemd, though. -- Tollef Fog Heen UNIX is user friendly, it's just picky about who its friends are -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/877gezefp6@qurzaw.varnish-software.com
Re: DSA, kFreeBSD, and the Singularity (was: How git performs when you throw all of Debian at it)
On Sun, Sep 01, 2013 at 09:49:30PM +0100, Steven Chamberlain wrote: > Hi Luca! > > On Sat, 31 Aug 2013 16:12:11 +, Luca Filipozzi wrote: > > We also run kfreebsd, with some challenge, but we have it. > > How can we help with that? http://dsa.debian.org/ports/kfreebsd/ > e.g. Do you need it to better suit running as a virtualised guest, on > KVM or Xen for example? virtio drivers should be forthcoming for jessie > and a XENHVM flavour is possible if it seems worth it. We use ganeti (kvm) for virtualization. Yes, better guest support would be great. More equivalency between linux and kfreebsd userland, at least as far as puppet is concerned, would help. For example, we like ferm, but ferm is iptables only. > I can only recall one wishlist bug from DSA at the moment which is > #711247 requesting pflogd. I'd love to hear more wishlist kfreebsd > ideas from DSA. See above url with incomplete list. -- Luca Filipozzi http://www.crowdrise.com/SupportDebian -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130902071431.ga1...@emyr.net
Re: DSA, kFreeBSD, and the Singularity (was: How git performs when you throw all of Debian at it)
Hi Luca! On Sat, 31 Aug 2013 16:12:11 +, Luca Filipozzi wrote: > We also run kfreebsd, with some challenge, but we have it. How can we help with that? e.g. Do you need it to better suit running as a virtualised guest, on KVM or Xen for example? virtio drivers should be forthcoming for jessie and a XENHVM flavour is possible if it seems worth it. I can only recall one wishlist bug from DSA at the moment which is #711247 requesting pflogd. I'd love to hear more wishlist kfreebsd ideas from DSA. I'd like it to be readily available whenever it might be a good fit for some internal project. Having as much Debian software as possible feeding back into Debian development systems, makes the project seem something of an unstoppable machine approaching the Singularity... Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/5223a85...@pyro.eu.org
Re: How git performs when you throw all of Debian at it
This one time, at band camp, Philipp Kern said: > On Sat, Aug 31, 2013 at 04:12:11PM +, Luca Filipozzi wrote: > > I'm curious why there's no apparent appetite for hdfs / solr / etc. > > I don't know how far regex matching is with solr these days. This > implementation is AFAIK based on [1]. But then the tool exists and > would need to be thrown away completely for no obvious gain. How would > you store the solr index? How would lookups be faster than a custom > built index where it's basically known how many seeks you need per > request? > > It's akin asking you to port everything to Chef just for the sake of > it, except that the target language would actually be Java, which > likely consumes even more resources than the Go binary. That's looking at it upside down, I think. chef was the new shiny after we'd already started down the puppet road. Michael's implementation did not, in fact, exist before solr. luca is not asking, "why aren't you using the new shiny", he's asking, "why hasn't a proposal for a project that does string searching used one of the already available, off the shelf string searching programs". Asking someone why they didn't choose off the shelf tools to do a job that is largely a solved problem is a reasonable thing to do, especially when being asked to support new, custom code that is certain to come with its own quirks, bugs, and security issues. I think the answer here is, "more code could be reused by borrowing work from google" - the NIH syndrome appears to be google's, rather than Michael's, if I understand correctly. That's probably a fine answer, but the question also needed to be asked. Cheers, -- - | ,''`.Stephen Gran | | : :' :sg...@debian.org | | `. `'Debian user, admin, and developer | |`- http://www.debian.org | - signature.asc Description: Digital signature
Re: How git performs when you throw all of Debian at it
On Sat, 2013-08-31 at 22:22 +0200, Adam Borowski wrote: > On Sat, Aug 31, 2013 at 12:44:21AM +0100, Ben Hutchings wrote: > > > Funny that you ask. What's the usual competitor for ZFS? > > > btrfs is included in stock kernels, doesn't take massive amounts of > > > memory, > > > and has a different approach to deduplication. > > > > and is slower than any of its competitors. > > time (tar xfa /usr/src/linux-source-3.11-rc4.tar.xz;sync) > ext4: 32.002s > btrfs: 18.770s > (an old spinning disk, just had a failure of my primary one) > > As Ceph backend: > http://ceph.com/uncategorized/argonaut-vs-bobtail-performance-preview/ On the other hand: http://article.gmane.org/gmane.comp.file-systems.xfs.general/54140 http://www.phoronix.com/scan.php?page=article&item=linux_310_10fs&num=2 > On the other hand, unless you deal with the fsync madness somehow, dpkg > performance is, before kernel 3.5, worse than abysmal, and on 3.5+, merely > bad. Yet even on ext4, getting rid of fsync gives a massive speedup, so > you want to wrap apt/dpkg with eatmydata where possible[1]. With no fsyncs, > btrfs slightly wins. > > > Thus: it depends on your particular usage pattern. With filesystems of > so different design principles it's hard to tell which is faster: there > are cases when btrfs beats competition by a lot, there are cases where > it gets beaten. > > > > > Recent kernels are needed only for race-free deduplication, "cp --reflink" > > > works in oldstable. > > > > Please don't suggest using the squeeze or wheezy version of btrfs in > > production. > > 2.6.32 (squeeze) has serious problems in ENOSPC conditions (up to a panic, > but no data loss) and works otherwise, what's the problem with 3.2? ~/src/d-k/dists/wheezy/linux$ grep -r 'BUG_ON(ret)' fs/btrfs | wc -l 290 So there are still 290 instances where an error will crash the system. Some will be cases where the caller has established a precondition that means the called function can't fail. Surely not all of them, though. (And in 3.11, there are still 100 of these...) Many of the many bug fixes I see cc'd to stable appear to be fixing bugs that exist in 3.2, but due to intervening changes they can't be applied without backporting. No-one's providing the backported versions, so the bugs don't get fixed. > Sure, fsync speed-ups from 3.5 and send/receive from 3.6 are good to have, > but by no means necessary. > > Distributions whose commercial support specifically mentions btrfs (Oracle, > SUSE) are using 3.0 and 3.1. [...] Bet they have lots of backported fixes, though. Ben. -- Ben Hutchings The most exhausting thing in life is being insincere. - Anne Morrow Lindberg signature.asc Description: This is a digitally signed message part
Re: How git performs when you throw all of Debian at it
On Sat, Aug 31, 2013 at 12:44:21AM +0100, Ben Hutchings wrote: > > Funny that you ask. What's the usual competitor for ZFS? > > btrfs is included in stock kernels, doesn't take massive amounts of memory, > > and has a different approach to deduplication. > > and is slower than any of its competitors. time (tar xfa /usr/src/linux-source-3.11-rc4.tar.xz;sync) ext4: 32.002s btrfs: 18.770s (an old spinning disk, just had a failure of my primary one) As Ceph backend: http://ceph.com/uncategorized/argonaut-vs-bobtail-performance-preview/ On the other hand, unless you deal with the fsync madness somehow, dpkg performance is, before kernel 3.5, worse than abysmal, and on 3.5+, merely bad. Yet even on ext4, getting rid of fsync gives a massive speedup, so you want to wrap apt/dpkg with eatmydata where possible[1]. With no fsyncs, btrfs slightly wins. Thus: it depends on your particular usage pattern. With filesystems of so different design principles it's hard to tell which is faster: there are cases when btrfs beats competition by a lot, there are cases where it gets beaten. > > Recent kernels are needed only for race-free deduplication, "cp --reflink" > > works in oldstable. > > Please don't suggest using the squeeze or wheezy version of btrfs in > production. 2.6.32 (squeeze) has serious problems in ENOSPC conditions (up to a panic, but no data loss) and works otherwise, what's the problem with 3.2? Sure, fsync speed-ups from 3.5 and send/receive from 3.6 are good to have, but by no means necessary. Distributions whose commercial support specifically mentions btrfs (Oracle, SUSE) are using 3.0 and 3.1. I second the recommendation to use newer kernels, but I see no reason to discourage using it with wheezy. [Disclaimer: while I use btrfs on a bunch of machines, I never touched its internal RAID, so I can vouch only for single block device use (including hw and md RAID).] [1]. On ext4, this should be at least debootstrap (no data to lose yet) and pbuilder/piuparts (a throw-away chroot). On btrfs using eatmydata on apt is safe even on the host system, albeit not in the default configuration: you'd need snapshots you can roll back to. -- ᛊᚨᚾᛁᛏᚣ᛫ᛁᛊ᛫ᚠᛟᚱ᛫ᚦᛖ᛫ᚹᛖᚨᚲ -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130831202231.ga26...@angband.pl
Re: How git performs when you throw all of Debian at it
On Sat, Aug 31, 2013 at 04:12:11PM +, Luca Filipozzi wrote: > I'm curious why there's no apparent appetite for hdfs / solr / etc. I don't know how far regex matching is with solr these days. This implementation is AFAIK based on [1]. But then the tool exists and would need to be thrown away completely for no obvious gain. How would you store the solr index? How would lookups be faster than a custom built index where it's basically known how many seeks you need per request? It's akin asking you to port everything to Chef just for the sake of it, except that the target language would actually be Java, which likely consumes even more resources than the Go binary. Michael was never opposed to sharding. HDFS is apparently a high-throughput FS, so not particularly suitable for random access if you try to avoid SSDs at all cost anyway. Kind regards Philipp Kern [1] http://swtch.com/~rsc/regexp/regexp4.html signature.asc Description: Digital signature
Re: How git performs when you throw all of Debian at it
On Sat, Aug 31, 2013 at 12:40:08AM +0200, Michael Stapelberg wrote: > Luca Filipozzi writes: > > Why do you say that when you haven't even asked? > Because I thought the answer was going to be “not in the Linux kernel, > no chance”. We also run kfreebsd, with some challenge, but we have it. > > To address this specific thread, the challenge with ZFS is not that we don't > > like the idea (I'm keen on it, actually) but that it's not in the > > Linux kernel. > See. > > I don’t have any energy to spend on this right now. Don’t expect > answers, I am killing this thread in my mailclient. Okay. I'm curious why there's no apparent appetite for hdfs / solr / etc. -- Luca Filipozzi http://www.crowdrise.com/SupportDebian -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130831161211.ga30...@emyr.net
Re: How git performs when you throw all of Debian at it
On Sat, Aug 31, 2013 at 12:32:47AM +0100, Dmitrijs Ledkovs wrote: > On 30 August 2013 20:55, Steven Chamberlain wrote: > > Hi, > > > >> [...] using git instead of the file system for storing the contents > >> of Debian Code Search. The hope was that it would lead to fewer disk > >> seeks and less data due to gits delta-encoding > > > > Wouldn't ZFS be a more natural way to do something like this? > > > > A choice of gzip, lzjb and more recently lz4 compression; snapshots > > and/or deduplication both reduce the amount of disk blocks and cache > > memory needed. > > > > I've pondered before at this overlap in functionality between packing by > > Git, and those features of the ZFS filesystem. They are doing much the > > same thing but with different granularity. It would be neat if they > > could work together better. > > I haven't finished packaging bedup - btrfs deduplication tool. bedup is only an userspace tool that calculates per-file hashes, then uses chattr tricks to avoid a race condition if some other process tried to write to the file. git renders the first part not needed: hashes are already known. If you're the only writer, you don't need to care about write races either. > Anybody have benchmarked that, if that's any good and/or comparable to zfs > deduplication? It's an apples to microsofts comparison: zfs takes a massive amount of memory to store block hashes. This has an upside: duplicated data never hits the actual disk, and a downside: the memory cannot be used for anything else, and if hashes hit the disk things become really slow. With btrfs, unless you know the hash beforehand (git), deduplication works after a write. This might be months later (one-shot), during the night (cron) or, if *notify is used, a small fraction of second later. btrfs can enumerate recently changed blocks for you so there's no need to read the whole disk in that cron job. -- ᛊᚨᚾᛁᛏᚣ᛫ᛁᛊ᛫ᚠᛟᚱ᛫ᚦᛖ᛫ᚹᛖᚨᚲ -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130831001654.ga16...@angband.pl
Re: How git performs when you throw all of Debian at it
On Sat, 2013-08-31 at 01:21 +0200, Adam Borowski wrote: > On Fri, Aug 30, 2013 at 10:14:25PM +, Luca Filipozzi wrote: > > On Fri, Aug 30, 2013 at 10:49:48PM +0200, Michael Stapelberg wrote: > > > Steven Chamberlain writes: > > > > Wouldn't ZFS be a more natural way to do something like this? > > > Possibly, but I have zero hopes of getting it set up and supported by > > > DSA, so we can’t use it for this service. > > > > To address this specific thread, the challenge with ZFS is not that we don't > > like the idea (I'm keen on it, actually) but that it's not in the Linux > > kernel. > > We prefer to use stock Debian kernels than custom-built kernels or modules > > for > > our machines. > > > > Is there another filesystem (or another approach) that would improve > > performance? > > Funny that you ask. What's the usual competitor for ZFS? > btrfs is included in stock kernels, doesn't take massive amounts of memory, > and has a different approach to deduplication. and is slower than any of its competitors. (But at least it doesn't have an incompatible licence.) > Recent kernels are needed only for race-free deduplication, "cp --reflink" > works in oldstable. Please don't suggest using the squeeze or wheezy version of btrfs in production. Ben. -- Ben Hutchings If God had intended Man to program, we'd have been born with serial I/O ports. signature.asc Description: This is a digitally signed message part
Re: How git performs when you throw all of Debian at it
On 30 August 2013 20:55, Steven Chamberlain wrote: > Hi, > >> [...] using git instead of the file system for storing the contents >> of Debian Code Search. The hope was that it would lead to fewer disk >> seeks and less data due to gits delta-encoding > > Wouldn't ZFS be a more natural way to do something like this? > > A choice of gzip, lzjb and more recently lz4 compression; snapshots > and/or deduplication both reduce the amount of disk blocks and cache > memory needed. > > I've pondered before at this overlap in functionality between packing by > Git, and those features of the ZFS filesystem. They are doing much the > same thing but with different granularity. It would be neat if they > could work together better. I haven't finished packaging bedup - btrfs deduplication tool. Anybody have benchmarked that, if that's any good and/or comparable to zfs deduplication? lzo compression is also available. And well available in linux kernel. Regards, Dmitrijs. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/canbhluibdvtshc5dbtwobf67xdexnjpumafcpvx1auj37te...@mail.gmail.com
Re: How git performs when you throw all of Debian at it
On Fri, Aug 30, 2013 at 10:14:25PM +, Luca Filipozzi wrote: > On Fri, Aug 30, 2013 at 10:49:48PM +0200, Michael Stapelberg wrote: > > Steven Chamberlain writes: > > > Wouldn't ZFS be a more natural way to do something like this? > > Possibly, but I have zero hopes of getting it set up and supported by > > DSA, so we can’t use it for this service. > > To address this specific thread, the challenge with ZFS is not that we don't > like the idea (I'm keen on it, actually) but that it's not in the Linux > kernel. > We prefer to use stock Debian kernels than custom-built kernels or modules for > our machines. > > Is there another filesystem (or another approach) that would improve > performance? Funny that you ask. What's the usual competitor for ZFS? btrfs is included in stock kernels, doesn't take massive amounts of memory, and has a different approach to deduplication. Recent kernels are needed only for race-free deduplication, "cp --reflink" works in oldstable. -- ᛊᚨᚾᛁᛏᚣ᛫ᛁᛊ᛫ᚠᛟᚱ᛫ᚦᛖ᛫ᚹᛖᚨᚲ -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130830232112.gb16...@angband.pl
Re: How git performs when you throw all of Debian at it
Hi, On Samstag, 31. August 2013, Luca Filipozzi wrote: > We are much more amenable to "how could I improve X so that it can run > faster" than we are to... I'd like to chime in and say a big thanks to DSA! piuparts.debian.org is running very very smoothly (and fast) now, thanks to the awesome ganeti cluster. And requests are usually answered+done very fast if they are sensible and if not, I usually learn nicely. /me bows. Thanks a lot & keep up the good work! (And these new schroot chroots are the awesome too!) cheers, Holger P.S.: they also take patches, why else would there be http://anonscm.debian.org/gitweb/?p=users/holger/debian.org.git ? :) signature.asc Description: This is a digitally signed message part.
Re: How git performs when you throw all of Debian at it
Hi Luca, Luca Filipozzi writes: > Why do you say that when you haven't even asked? Because I thought the answer was going to be “not in the Linux kernel, no chance”. > To address this specific thread, the challenge with ZFS is not that we don't > like the idea (I'm keen on it, actually) but that it's not in the > Linux kernel. See. I don’t have any energy to spend on this right now. Don’t expect answers, I am killing this thread in my mailclient. I am sorry for what I wrote and did not mean to offend anybody. Have a nice weekend. -- Best regards, Michael -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/x6y57izttj@midna.lan
Re: How git performs when you throw all of Debian at it
On Fri, Aug 30, 2013 at 10:49:48PM +0200, Michael Stapelberg wrote: > Steven Chamberlain writes: > > Wouldn't ZFS be a more natural way to do something like this? > Possibly, but I have zero hopes of getting it set up and supported by > DSA, so we can’t use it for this service. Why do you say that when you haven't even asked? Significant issues for DSA are: - service owners saying "I need to unpack the entire archive and i need to search it fast so get more hardware like SSDs" or equivalent without understanding our constraints - service owners not engaging with us early in their implementation to understand what constraints we might need to impose or to sound us out on architecture to improve performance or to reduce resource usage or to reduce software complexity We are much more amenable to "how could I improve X so that it can run faster" than we are to "zero hopes that DSA will support it". Disparaging us certainly does not draw bees to honey, as they say. To address this specific thread, the challenge with ZFS is not that we don't like the idea (I'm keen on it, actually) but that it's not in the Linux kernel. We prefer to use stock Debian kernels than custom-built kernels or modules for our machines. If we can overcome these challenges for ZFS (which are significant, I admit), we would be amenable to a discussion regarding ZFS (although it isn't a great match to our current hardware configuration). Is there another filesystem (or another approach) that would improve performance? Are there other things we could consider (hdfs & solr, say)? You know where we live, Luca -- Luca Filipozzi http://www.crowdrise.com/SupportDebian -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130830221425.ga19...@emyr.net
Re: How git performs when you throw all of Debian at it
On 30/08/13 21:49, Michael Stapelberg wrote: > Steven Chamberlain writes: >> Wouldn't ZFS be a more natural way to do something like this? > Possibly, but I have zero hopes of getting it set up and supported by > DSA, so we can’t use it for this service. Oh I see. That's fair enough, but there is some hope that could change someday: ZFS-on-Linux is making good progress recently, and we (GNU/kFreeBSD team and upstream developers of ZFS) can try to better educate folks on ZFS generally. Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/52211337.7050...@pyro.eu.org
Re: How git performs when you throw all of Debian at it
Hi Steven, Steven Chamberlain writes: > Wouldn't ZFS be a more natural way to do something like this? Possibly, but I have zero hopes of getting it set up and supported by DSA, so we can’t use it for this service. -- Best regards, Michael -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/x61u5a29ar@midna.lan
Re: How git performs when you throw all of Debian at it
Hi, > [...] using git instead of the file system for storing the contents > of Debian Code Search. The hope was that it would lead to fewer disk > seeks and less data due to gits delta-encoding Wouldn't ZFS be a more natural way to do something like this? A choice of gzip, lzjb and more recently lz4 compression; snapshots and/or deduplication both reduce the amount of disk blocks and cache memory needed. I've pondered before at this overlap in functionality between packing by Git, and those features of the ZFS filesystem. They are doing much the same thing but with different granularity. It would be neat if they could work together better. Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/5220f898.6000...@pyro.eu.org