Read up on how ZFS and RAID-Z works. All data is checksummed - if you get checksum error, it will attempt to re-read the data. I am assuming your error rate is a random error, so therefore it will be read correctly on re-read and it will go on with life.
But yes, replacing one disk at a time does open you up to a failure window while it resilvers the new one. Of course, you can avoid that with RAIDZ2.
Name one filesystem that can deal with silent data corruption - sure in a lot of cases they just keep going because they have no way of knowing the data they just read from disk contained half of a corrupted write. Just because years of trial and error has made fsck tools to *attempt* repairs post-mortem does not mean ZFS is doing something wrong when it tells you "your data is fucked, now".
ZFS works fine on cheap hardware too, I have seen some brilliant performance figures on cheap home SATA setups. I would support the ECC statement though from personal experience. Its no good having any sort of decent storage setup when single bit flips in memory crap on your data. With UFS I might not have noticed until it got bad enough to completely stuff the filesystem, but with ZFS I picked it up straight away before I lost any more files.
We had a large java application that was running out of capacity on a fully stocked E6900 (24 dual core 1.35Ghz US-IV cpus). We had a demo T5240 handy to try it on. The component we moved off used about 35% of the CPU on the E6900. On the T5240 it used about 15%. The drop in CPU load on the E6900 was about 50% (scheduling etc - it was Solaris 8 so not the best for 48 cores).
The T2+ CPUs absolutely tear it up for Java - we figured moving the entire app from the constrained E6900 would only use about 40-50% of the T5240. Not bad for a 2RU box vs a full cabinet machine that is a fraction of the price.
I did some database testing with the machine while we had it for a demo - Oracle 11G on ZFS on the T5240 basically performed 1:1 with a POWER5 LPAR, when you take into consideration the difference in clock speeds (1.4Ghz on T2, 1.9Ghz on P5). The issue was the query in question was CPU bound - so the machine was only showing like 2% utilisation. If you have highly concurrent small queries, a T2 would be ideal - but if you have CPU bound larger stuff it just wont keep up with its single threaded performance.
I am assuming you have some sort of redundancy in your pool configuration, this being your uber always running data center machine that in no way is using USB disks.
So as you have either RAID-Z or mirrored vdevs, in both cases you an offline one of the devices, replace it with a larger one, then resilver the pool. Once you have replaced the devices in the vdev with larger ones, the pool will automatically resize.
But your point is taken - the ability to remove devices permanently is a basic administration feature which is lacking currently, but the work is underway. From what I understand the model used to do it opens up a number of useful methods for playing with zpools
No - but you are reading it the way you want - lets go over this again.
Sun wanted to Open Source Solaris because it was the cool thing to do at the time. However there was non-Sun code which could not be released. Also Sun as a OS provider had vendors who wrote kernel modules for Solaris. According to the interpretations of kernel modules under GPL, putting Solaris or any of Sun's kernel code under the GPL required the rest of the kernel to be GPL - not suitable for other vendors who did not wish to GPL their code. Argue that if you want, but find a corporate lawyer who wont read the GPL conservatively.
So - Sun wrote the CDDL in order to have a license which met their needs and their vendors and partners needs, and then released their code.
So now in the minds of Linux fanbois everywhere, Solaris goes from being a closed evil proprietary operating system trying to kill linux with lockin, to a open source evil operating system trying to undermine Linux by tempting it with code that it cannot integrate.
So - why not dual license Linux if you want the code so bad?
What I find funny is the number of posts that on one hand flog ZFS for being inferior to any Linux filesystem for so many ill informed reasons, then decry it for not being GPL and hence not available in Linux.
I dunno - one thing IBM does consistently is show they are willing to drop their pants to make a sale. If that means holding onto SPARC, Coolthreads, Solaris etc in order to maintain that customer base and keep HP out of the picture I bet they will do it - you don't spend $7B in order to lose 30% market share.
That said, the Sun software portfolio will finally get the trim down it needed. I for one would be happy to see a Solaris on Power version - would be intersting for them to say they have one OS that runs on all their platforms - x64, SPARC, Power and Z series.
The difference between adults and teenagers isn't the mental capacity - its the experience to make the right decision, instead of the decision that suits you at the time.
Right - treating someone with a position of legal authority like a "asshole murderer" will make him reconsider his position in life, not push him to become like the thug police you are ranting against. I think most people in that sort of line of work would tend to shove back when shoved.
Anecdotally, after they got their MMR shot their behavior went downhill rapidly and not just in a "they are cranky because the shot made them a little ill" - their entire behaviour changed.
This was before we had any real clue they may be Autistic - so it wasn't a case of seeing something that wasn't there.
That said, I don't think it is a cause as I am fairly sure its genetic - something happened but I don't know how to explain it - maybe something in it did amplify or upset their existing condition. Looking back their newborn behaviour and other clues early on to me lead me to believe it was genetic.
Looking at myself and my own childhood I can see traits in my own personality and thought processes that correlate with the way autistic people describe themselves, although in my case it was probably more to a degree that I wouldn't be diagnosed, I was just "different".
I don't think they are really just wanting to show off their breasts - the majority are proud that they breastfed their children and want to demonstrate that.
No - because one has the life of a child depending on them, and the others have some ideas contrary to common public thinking. One is essential to life, the other is a life choice.
Not to say the arguments of nudists are wrong - just that the two are not in the same ballpark.
Why is the need to feed a baby considered a "right" in the same breath as public photography, rather than being acknowledged as a key function necessary for parenting? It is feeding a baby - not a display of public exhibitionism. All in all that is what boobs are for.
I think you will find many women are shy and discreet about feeding in public for many of the reasons coming up in this debate. It does attract attention, it is uncomfortable, it isn't always easy to keep covered up with a wriggling infant.
But they *have* to do it and it is not always convenient to relocate to do so. Would you enjoy sitting in a public toilet to eat your lunch? Many baby change rooms are not that great a place for it.
My view is that as long as the woman is being discreet and not just getting them out like its mardi gras, then she should be afforded some respect and "privacy" and even legal protection to do so. And as others point out - if others don't like it, they don't have to spectate.
No - It is an arms race - they will just arm themselves better to counter that threat. And most likely as they are already immoral or irrational people they will shoot at the first sign of a threat.
You wont get rid of theives or idiots and you wont get rid of weapons. But going in the direction of every man for himself and fuck you as long as I am all right will just lead to disaster.
Is an insurance claim worth someones, even a criminals, life?
Drivers support is good - Solaris 10 has come a long way in the last few years. I dont recall having to do many hacks on any currect hardware to get things going. To answer your queries:
Graphics - Fully supported driver from Nvidia covering pretty much anything past a GeForce 4. Some older Xorg drivers cover the rest ok. The 3D support is limited to the Nvidia cards at the moment I believe.
SATA - Marvell and Silicon Image SATA2 chipsets supported natively and now with NCQ (with more coming - Nvidia I believe). I have two 4 port SI-3124 cards running ZFS without a hickup and my performance is limited by the PCI bus. I have a friend with PCI-e cards getting 400Mb/s throughput with RAID-Z on a home grade opteron board.
Wifi is coming on board as are many other facets.
Serverwise it natively supports Qlogic and Emulex fibre cards and most current SCSI/SAS interfaces on Sun/Dell/HP/IBM kit.
Intel have allready put back kernel and libc optimisations specific for their processors (Solaris dynamiclly loads customised performance libraries depending on what platform it is on).
While its true Solaris is not as bleeding edge as Linux for new hardware, it does tend to do things right when they are released i.e properly supported by cfgadm or other interfaces, and tend to work out of the box etc.
I am not anti-gun. I grew up in a rural area enjoying being able to take pot shots at stuff in the backyard with my Dad's rifles, I was in the AIRTC and loved the opportunity I got to shoot a SLR, handle Styers as well take part in competitive.22 shooting.
But at the end of the day I agree with the short fuse fuckwit angle. As much as I think Howard is a knob, the gun ban was a good move on his behalf. It did suck for our AIRTC unit though, we had to hand in our welded barrel SLR drill rifles (The instructor who carried in the 10 SLRs to the cop shop got comments that he was lucky they didnt get arrested for having that lot:)).
Of course there will be illegal guns, most now days seem to be stolen off security guards (why the carry them when they practically cant use them is beyond me, its obvious the crims know they wont use them). But by limiting supply to reasonable rural use and sporting type weapons you constrain the supply of opportunistic weapons to be used for crime or violence.
It is unfortunate that collectors and genuine sports shooters get affected by this, but I think thats a reasonable impact, as it limits weapons that can be stolen and go into the above illegal trade.
Any talk of arms for militia groups is just crap now days, sure it worked in 1785 or whatever, but they didn't have attack choppers and thermal imaging cameras back then.
My personal example - I have my kids photos stored on a ZFS mirror pool on a cheap little Athlon PC. Worked fine, until one day I stopped being able to read some files - some checking and I found ZFS had picked up checksum errors on them on *both* sides of the mirror. The strange part was the more I scrubbed, the more errors I found. I noted that new files were more prone to it, and not all of them either.
After some consideration I realised the only way both mirrors could be corrupted was if the data was written that way. If it was disk or controller it would be random, this was perfectly consistant on both sides. For that to happen and ZFS to detect it the checksum or data would have to have changed in memory after ZFS calculated the checksum and then be written to disk that way.
Long story short - I had non-ECC memory and the motherboard had blown a capacitor. It seems that occasionally a flakey area of memory would corrupt some writes and later on ZFS would pick it up.
With any other filesystem I would never have known.
Why do you believe there is a penalty for ZFS doing checksums on CPU and not on some open-to-corruption-and-complexity accelerator card?
I can quite easily push my home RAID-Z setup (4x500G SATA-II) to 120MB/s - limited by PCI bandwidth - the CPU usage for that is fairly minimal on an Athlon64 3400. I have a friend using PCI-e who got benchmarks of nearly 200Mb/s write speeds and 480Mb/s read, on not reasonable home PC motherboards. It flogs $1200 Hardware RAID cards with cheap $30 SATA cards.
Granted it uses more CPU than UFS would, but in most cases multi-core boxes make that negligible and so far for me it has never been an issue.
Because so far ZFS is the only LVM+Filesystem I am aware of that can only resilver the data you use, rather than the entire disk.
In the case of the 7 2Tb drive array, if you only have 2Tb of data on the array, you *have* to read all of the other 6 disks in a failure to repair the RAID - so 12Tb of read and 2Tb of writes to repair the array.
With ZFS RAID-Z you would read ~ 2Tb of data and write 300M. That means you have far less chance of encountering another error while the data is repairing.
Of course, if you filesystem is full you have to do the lot, but for anything less than that ZFS wins.
I really hate that this particular bit of FUD keeps floating around. Seriously - look at the ZFS design, it has so much integrity and self checking built into it that the only two cases for errors are:
1. Data corruption on disk due to a corrupted write or bit rot. Depending on the case it may be recoverable from another location in the pool or not. If the fault is in the uberblock you are screwed, but same as any other filesystem where its root data is stuffed. In the not repairable case it simply gives you an IO error and flags an FMA event that this file/object is screwed. In any filesystem that is a blow away and restore event. Same as any other filesytem you might be able to rm and save some data, but generally its gone and recover time.
2. A ZFS software failure of some sort - bad metadata written to disk with valid checksums for example. In this case the actual fix is to the ZFS module and application of a patch, and the short term fix once the bug is discovered is probably a repair tool from Sun, unless the fix involves ZFS being updated so it is able to self correct this fault.
Which brings us to fsck. Everything that fsck does is based on previously encountered behaviour with a known pathology and possible method for repair. Typically this is addressing known forms of corruption caused by the filesystem itself and on disk inconsistancy (writes lost at poweroff etc). Sure XFS had no tool initially, but once the problems were found one was written. UFS has 20 years to experience to put into fsck.
Many of these cases are allready handled by the inherent design of ZFS - self checking data structures, checksums throughout, copy-on-write disk updates.
So, unless you can find an inherent flaw in the design of ZFS, or a bug in the code you have no case to need a fsck utility more advanced than zfs scrub. It may be the case that one day there will be a zfs repair command - but until such a need arises why be so paranoid?
Most ZFS "oh fuck" events I have heard of have either been hardware issues (expensive RAID arrays doing silent data corruption) or some extreme cases of testing abuse. In my professional experience with it over the last 2 years I have only see one issue caused by bad non-ECC memory on my home PC.
There is plenty of Compaq/HP gear on there, not to mention IBM BladeChassis machines etc. Probably not that interesting for those on a budget or the slasddot "my whitebox is teh leet" crowd, but for real IT shops it is a reasonable list.
When we looked at this a V445 was much more suitable - multiple PCI controllers, more slots and better bandwidth - and enough CPU to keep up with the lot.
We looked at a T2000 for a similar role, IIRC the problem with a setup like you are proposing is that they only have one PCI controller, so that will be your weakpoint as all your SAN and Network traffic will be hammering that poor puppy. Keep in mind everything coming in on the network, will likely go out the SAN, so your real "work" throughput will be roughly half of the capacity of the PCI controller. The CPU wont enter into it too much - there is more than enough to handle it.
The crypto is exposed by kernel interfaces and OS library interfaces. Basically if you link to the/usr/sfw/lib OpenSSL built into Solaris 10, you get the benefit of CPU core crypto acceleration.
We have a few of these boxes - while the straight line speed is noticeably lacking on some tasks, the box runs everything thrown at it quite well. The odd part for me was realizing that a process using 3% cpu in prstat was actually consuming an entire CPU thread.
Slashdot Mirror
Read up on how ZFS and RAID-Z works. All data is checksummed - if you get checksum error, it will attempt to re-read the data. I am assuming your error rate is a random error, so therefore it will be read correctly on re-read and it will go on with life.
But yes, replacing one disk at a time does open you up to a failure window while it resilvers the new one. Of course, you can avoid that with RAIDZ2.
Name one filesystem that can deal with silent data corruption - sure in a lot of cases they just keep going because they have no way of knowing the data they just read from disk contained half of a corrupted write. Just because years of trial and error has made fsck tools to *attempt* repairs post-mortem does not mean ZFS is doing something wrong when it tells you "your data is fucked, now".
ZFS works fine on cheap hardware too, I have seen some brilliant performance figures on cheap home SATA setups. I would support the ECC statement though from personal experience. Its no good having any sort of decent storage setup when single bit flips in memory crap on your data. With UFS I might not have noticed until it got bad enough to completely stuff the filesystem, but with ZFS I picked it up straight away before I lost any more files.
We had a large java application that was running out of capacity on a fully stocked E6900 (24 dual core 1.35Ghz US-IV cpus). We had a demo T5240 handy to try it on. The component we moved off used about 35% of the CPU on the E6900. On the T5240 it used about 15%. The drop in CPU load on the E6900 was about 50% (scheduling etc - it was Solaris 8 so not the best for 48 cores).
The T2+ CPUs absolutely tear it up for Java - we figured moving the entire app from the constrained E6900 would only use about 40-50% of the T5240. Not bad for a 2RU box vs a full cabinet machine that is a fraction of the price.
I did some database testing with the machine while we had it for a demo - Oracle 11G on ZFS on the T5240 basically performed 1:1 with a POWER5 LPAR, when you take into consideration the difference in clock speeds (1.4Ghz on T2, 1.9Ghz on P5). The issue was the query in question was CPU bound - so the machine was only showing like 2% utilisation. If you have highly concurrent small queries, a T2 would be ideal - but if you have CPU bound larger stuff it just wont keep up with its single threaded performance.
I am assuming you have some sort of redundancy in your pool configuration, this being your uber always running data center machine that in no way is using USB disks.
So as you have either RAID-Z or mirrored vdevs, in both cases you an offline one of the devices, replace it with a larger one, then resilver the pool. Once you have replaced the devices in the vdev with larger ones, the pool will automatically resize.
But your point is taken - the ability to remove devices permanently is a basic administration feature which is lacking currently, but the work is underway. From what I understand the model used to do it opens up a number of useful methods for playing with zpools
No - but you are reading it the way you want - lets go over this again.
Sun wanted to Open Source Solaris because it was the cool thing to do at the time.
However there was non-Sun code which could not be released. Also Sun as a OS provider had vendors who wrote kernel modules for Solaris. According to the interpretations of kernel modules under GPL, putting Solaris or any of Sun's kernel code under the GPL required the rest of the kernel to be GPL - not suitable for other vendors who did not wish to GPL their code. Argue that if you want, but find a corporate lawyer who wont read the GPL conservatively.
So - Sun wrote the CDDL in order to have a license which met their needs and their vendors and partners needs, and then released their code.
So now in the minds of Linux fanbois everywhere, Solaris goes from being a closed evil proprietary operating system trying to kill linux with lockin, to a open source evil operating system trying to undermine Linux by tempting it with code that it cannot integrate.
So - why not dual license Linux if you want the code so bad?
What I find funny is the number of posts that on one hand flog ZFS for being inferior to any Linux filesystem for so many ill informed reasons, then decry it for not being GPL and hence not available in Linux.
I dunno - one thing IBM does consistently is show they are willing to drop their pants to make a sale. If that means holding onto SPARC, Coolthreads, Solaris etc in order to maintain that customer base and keep HP out of the picture I bet they will do it - you don't spend $7B in order to lose 30% market share.
That said, the Sun software portfolio will finally get the trim down it needed. I for one would be happy to see a Solaris on Power version - would be intersting for them to say they have one OS that runs on all their platforms - x64, SPARC, Power and Z series.
The difference between adults and teenagers isn't the mental capacity - its the experience to make the right decision, instead of the decision that suits you at the time.
Right - treating someone with a position of legal authority like a "asshole murderer" will make him reconsider his position in life, not push him to become like the thug police you are ranting against. I think most people in that sort of line of work would tend to shove back when shoved.
I have mildly autistic twin boys.
Anecdotally, after they got their MMR shot their behavior went downhill rapidly and not just in a "they are cranky because the shot made them a little ill" - their entire behaviour changed.
This was before we had any real clue they may be Autistic - so it wasn't a case of seeing something that wasn't there.
That said, I don't think it is a cause as I am fairly sure its genetic - something happened but I don't know how to explain it - maybe something in it did amplify or upset their existing condition. Looking back their newborn behaviour and other clues early on to me lead me to believe it was genetic.
Looking at myself and my own childhood I can see traits in my own personality and thought processes that correlate with the way autistic people describe themselves, although in my case it was probably more to a degree that I wouldn't be diagnosed, I was just "different".
Sounds like the mind test games from Ender - it probes you and comes back with shit to fuck with your head.
No - feeding an infant is not a matter of choice - its a required natural function essential to the life of that child.
I don't think they are really just wanting to show off their breasts - the majority are proud that they breastfed their children and want to demonstrate that.
No - because one has the life of a child depending on them, and the others have some ideas contrary to common public thinking. One is essential to life, the other is a life choice.
Not to say the arguments of nudists are wrong - just that the two are not in the same ballpark.
Why is the need to feed a baby considered a "right" in the same breath as public photography, rather than being acknowledged as a key function necessary for parenting? It is feeding a baby - not a display of public exhibitionism. All in all that is what boobs are for.
I think you will find many women are shy and discreet about feeding in public for many of the reasons coming up in this debate. It does attract attention, it is uncomfortable, it isn't always easy to keep covered up with a wriggling infant.
But they *have* to do it and it is not always convenient to relocate to do so. Would you enjoy sitting in a public toilet to eat your lunch? Many baby change rooms are not that great a place for it.
My view is that as long as the woman is being discreet and not just getting them out like its mardi gras, then she should be afforded some respect and "privacy" and even legal protection to do so. And as others point out - if others don't like it, they don't have to spectate.
No - It is an arms race - they will just arm themselves better to counter that threat. And most likely as they are already immoral or irrational people they will shoot at the first sign of a threat.
You wont get rid of theives or idiots and you wont get rid of weapons. But going in the direction of every man for himself and fuck you as long as I am all right will just lead to disaster.
Is an insurance claim worth someones, even a criminals, life?
Drivers support is good - Solaris 10 has come a long way in the last few years. I dont recall having to do many hacks on any currect hardware to get things going. To answer your queries:
Graphics - Fully supported driver from Nvidia covering pretty much anything past a GeForce 4. Some older Xorg drivers cover the rest ok. The 3D support is limited to the Nvidia cards at the moment I believe.
SATA - Marvell and Silicon Image SATA2 chipsets supported natively and now with NCQ (with more coming - Nvidia I believe). I have two 4 port SI-3124 cards running ZFS without a hickup and my performance is limited by the PCI bus. I have a friend with PCI-e cards getting 400Mb/s throughput with RAID-Z on a home grade opteron board.
Wifi is coming on board as are many other facets.
Serverwise it natively supports Qlogic and Emulex fibre cards and most current SCSI/SAS interfaces on Sun/Dell/HP/IBM kit.
Intel have allready put back kernel and libc optimisations specific for their processors (Solaris dynamiclly loads customised performance libraries depending on what platform it is on).
While its true Solaris is not as bleeding edge as Linux for new hardware, it does tend to do things right when they are released i.e properly supported by cfgadm or other interfaces, and tend to work out of the box etc.
I am not anti-gun. I grew up in a rural area enjoying being able to take pot shots at stuff in the backyard with my Dad's rifles, I was in the AIRTC and loved the opportunity I got to shoot a SLR, handle Styers as well take part in competitive .22 shooting.
But at the end of the day I agree with the short fuse fuckwit angle. As much as I think Howard is a knob, the gun ban was a good move on his behalf. It did suck for our AIRTC unit though, we had to hand in our welded barrel SLR drill rifles (The instructor who carried in the 10 SLRs to the cop shop got comments that he was lucky they didnt get arrested for having that lot :)).
Of course there will be illegal guns, most now days seem to be stolen off security guards (why the carry them when they practically cant use them is beyond me, its obvious the crims know they wont use them). But by limiting supply to reasonable rural use and sporting type weapons you constrain the supply of opportunistic weapons to be used for crime or violence.
It is unfortunate that collectors and genuine sports shooters get affected by this, but I think thats a reasonable impact, as it limits weapons that can be stolen and go into the above illegal trade.
Any talk of arms for militia groups is just crap now days, sure it worked in 1785 or whatever, but they didn't have attack choppers and thermal imaging cameras back then.
My personal example - I have my kids photos stored on a ZFS mirror pool on a cheap little Athlon PC. Worked fine, until one day I stopped being able to read some files - some checking and I found ZFS had picked up checksum errors on them on *both* sides of the mirror. The strange part was the more I scrubbed, the more errors I found. I noted that new files were more prone to it, and not all of them either.
After some consideration I realised the only way both mirrors could be corrupted was if the data was written that way. If it was disk or controller it would be random, this was perfectly consistant on both sides. For that to happen and ZFS to detect it the checksum or data would have to have changed in memory after ZFS calculated the checksum and then be written to disk that way.
Long story short - I had non-ECC memory and the motherboard had blown a capacitor. It seems that occasionally a flakey area of memory would corrupt some writes and later on ZFS would pick it up.
With any other filesystem I would never have known.
Why do you believe there is a penalty for ZFS doing checksums on CPU and not on some open-to-corruption-and-complexity accelerator card?
I can quite easily push my home RAID-Z setup (4x500G SATA-II) to 120MB/s - limited by PCI bandwidth - the CPU usage for that is fairly minimal on an Athlon64 3400. I have a friend using PCI-e who got benchmarks of nearly 200Mb/s write speeds and 480Mb/s read, on not reasonable home PC motherboards. It flogs $1200 Hardware RAID cards with cheap $30 SATA cards.
Granted it uses more CPU than UFS would, but in most cases multi-core boxes make that negligible and so far for me it has never been an issue.
Because so far ZFS is the only LVM+Filesystem I am aware of that can only resilver the data you use, rather than the entire disk.
In the case of the 7 2Tb drive array, if you only have 2Tb of data on the array, you *have* to read all of the other 6 disks in a failure to repair the RAID - so 12Tb of read and 2Tb of writes to repair the array.
With ZFS RAID-Z you would read ~ 2Tb of data and write 300M. That means you have far less chance of encountering another error while the data is repairing.
Of course, if you filesystem is full you have to do the lot, but for anything less than that ZFS wins.
I really hate that this particular bit of FUD keeps floating around. Seriously - look at the ZFS design, it has so much integrity and self checking built into it that the only two cases for errors are:
1. Data corruption on disk due to a corrupted write or bit rot. Depending on the case it may be recoverable from another location in the pool or not. If the fault is in the uberblock you are screwed, but same as any other filesystem where its root data is stuffed. In the not repairable case it simply gives you an IO error and flags an FMA event that this file/object is screwed. In any filesystem that is a blow away and restore event. Same as any other filesytem you might be able to rm and save some data, but generally its gone and recover time.
2. A ZFS software failure of some sort - bad metadata written to disk with valid checksums for example. In this case the actual fix is to the ZFS module and application of a patch, and the short term fix once the bug is discovered is probably a repair tool from Sun, unless the fix involves ZFS being updated so it is able to self correct this fault.
Which brings us to fsck. Everything that fsck does is based on previously encountered behaviour with a known pathology and possible method for repair. Typically this is addressing known forms of corruption caused by the filesystem itself and on disk inconsistancy (writes lost at poweroff etc). Sure XFS had no tool initially, but once the problems were found one was written. UFS has 20 years to experience to put into fsck.
Many of these cases are allready handled by the inherent design of ZFS - self checking data structures, checksums throughout, copy-on-write disk updates.
So, unless you can find an inherent flaw in the design of ZFS, or a bug in the code you have no case to need a fsck utility more advanced than zfs scrub. It may be the case that one day there will be a zfs repair command - but until such a need arises why be so paranoid?
Most ZFS "oh fuck" events I have heard of have either been hardware issues (expensive RAID arrays doing silent data corruption) or some extreme cases of testing abuse. In my professional experience with it over the last 2 years I have only see one issue caused by bad non-ECC memory on my home PC.
Never mind the fact Novell sold Sun the rights to do what it wants with its code some 20 years ago, so it cant force Sun to do anything.
Why not relicense the parts of Linux that you see are missing from Solaris? Same result.
OpenSolaris is more about getting community involvement into the direction and shape of Solaris than it is about making a free donation to Linux.
See the Solaris 10 HCL - in particular the OEM vendors page:
v iews/oem_and_system_vendor_products_all_results.pa ge1.html
http://www.sun.com/bigadmin/hcl/data/sol/systems/
There is plenty of Compaq/HP gear on there, not to mention IBM BladeChassis machines etc. Probably not that interesting for
those on a budget or the slasddot "my whitebox is teh leet" crowd, but for real IT shops it is a reasonable list.
When we looked at this a V445 was much more suitable - multiple PCI controllers, more slots and better bandwidth - and enough CPU to keep up with the lot.
We looked at a T2000 for a similar role, IIRC the problem with a setup like you are proposing is that they only have one PCI controller, so that will be your weakpoint as all your SAN and Network traffic will be hammering that poor puppy. Keep in mind everything coming in on the network, will likely go out the SAN, so your real "work" throughput will be roughly half of the capacity of the PCI controller. The CPU wont enter into it too much - there is more than enough to handle it.
The crypto is exposed by kernel interfaces and OS library interfaces. Basically if you link to the /usr/sfw/lib OpenSSL built into Solaris 10, you get the benefit of CPU core crypto acceleration.
We have a few of these boxes - while the straight line speed is noticeably lacking on some tasks, the box runs everything thrown at it quite well. The odd part for me was realizing that a process using 3% cpu in prstat was actually consuming an entire CPU thread.