But we are not talking about a "lets make the IP subsystem understand X traffic in order to generate desktop popups" sort of layer transition. This is purely talking about the storage allocator in ZFS that simply does a number of things allready present in some of the exising storage areas (namely md devices and snapshots). However the reason it does this is to acheive a number of very positive benefits to the upper filesystem layer.
The reason ZFS duplicates functionality is because it *has* to do achieve the sorts of features it is able to provide (read the whitepaper or the OpenSolaris ZFS community materials to see what I mean). I thourghly recommend the to small ZFS tutorial flash movies on the Open Solaris ZFS community - they show the basics of what ZFS can do very concisely.
Traditional filesystems view the disk as a slab of blocks do what they please with. Traditional LVMs view disks as places to abstract into a single block of devices. There is no communication between other than "write ok, yes/no? Read ok, yes/no?". By breaking that model you allow point in time snapshots of the filesystem, not the device. You allow writable clones that share common blocks, you allow the storage subsystem to tune its performance to the filesystem work, rather than model the underlying storage on your idea of what the filesystem is doing.
And so what if the storage allocator is an amorphous blob. Its open source, it does have interfaces (zvols amongst others). Why would you want to get involved in its internals unless it is not doing its job, and if it isnt the source is there to improve it.
So many of the arguments on this topic seem to be of the line "linux allready does this, so why change?" or "I dont get it, so it must suck". But there is nothing in ZFS which says it *must* replace these functions - you can make a zpool of some devices and md/lvm etc of others - it doesnt care. You can make any other filesystem that wants a block device on top of zvols - it will happily do it. If it proves to not compete or do as well, it will fade away with time and die. If it shows to be an improvement (which in my opinion it certainly is, coming from years of SDS+UFS and VXFS+VXVM admin on Solaris) the over time it will probably replace those other features.
Why must there be a line in the sand instead of a productive, usable alternative?
I think you need to look at ZFS for what it is achieving rather than the way it does things differently.
The arguments flying around here seem to be "it is different, is therefore bad" and "I isnt something I understand, so its bad".
The problem with the first argument is that it *has* to be different in order to do what it is aiming for. By making the filesystem talk to the storage, and making the storage talk with the filesystem you remove all the baggage of using RAID to pretend disks are something they are not. I think the simplest case of this is resilvering the disks - if you pull out and replace a disk, because the block subsystem knows what the filesystem has used, it only copies what is needed - try that with a multi-terabyte RAID stripe where with large enough disks you can have a second failure before your first resync is finished.
And I just dont understand the second argument - I throught geeks liked playing with new toys.
Besides, as I understand it the ZFS design allows for certain plug in modules above the block allocator level which is where you really want customisation anyway (I have heard talk of database storage engine plugins and the like)? Letting ZFS take care of all the grunt work underneath makes sence, assuming it does it well and is reliable.
Dtrace is the exact opposite of error vomit, and I dont recall ever hearing it called that anyway. The entire principle is that you dont need to go inserting metric shitloads of debugging and printf("we got here") statements all through your code, recompile it and then see that the error doesnt occur because all your debugging has now slowed your code enough to prevent the race condition that caused the original error.
True - its a L3 and developer tool for the most part, but there are plenty of scripts out there to show what it can do for an admin. Take a look at http://users.tpg.com.au/adsln4yb/dtrace.html for starters. Stuff like iosnoop, iotop, opensnoop and kill.d can be used quite regularly by admins without the need for putting debugging into active applications.
I dont know about the Alan Cox comment, but for IDE this is a common thing. Simply put IDE disks struggle enough for performance, so by default have write caching enabled.
I work for a major server vendor who creates their own firmware for their disks. By default all SCSI and FCAL disks are configured to have write cache disabled because data integrity is valued over performance. For ATA apparently the disk vendors dont give any option for it, so we are unable to work around that.
This is actually quite a pain when it comes to benchmarks, because for SOME tests it makes OSes which enable the write cache to look really fast. Its not until you suffer a catastrophe that you find out the data never made the platter.
And RAID devices dont lie about completing I/O - the device presents an "disk" interface to a slab of battery backed (hopefully) cache to disk which allows write performance to be massivly better. The RAID card itself takes care of syncing its cache to the disks, it just takes the data in cache and responds to the transaction immediately, flushing later. As far as the OS needs to know, the IO is complete - firmware bugs and battery failures aside, the RAID card handles it internally.
And the author seems to be lacking clue about what he is testing - if anything all it is testing is the the OSes ability to get data down to the disks consistantly - all fsync() knows is that the calls it made to send data to the storage devices returned success, its totally dependant on the volume manger, disk and controller drivers as well as the actual physical storage to get the job done.
For all he knows his drivers might be returning immediately just to make performance look better, but actually scheduling the I/O in some manner which causes it to be lost before commitment to storage.
If he wants to complain about the disks, I think he is going to need a much lower level test that a perl script calling sync.
I do picture this because I have been through several Australian elections. Every single advertisement is along those lines i.e blatent bunch of misrepresentation, small disclaimer - vote X.
Sure, a good portion of the populace swallows it hook line and sinker, but if you leave only the "educated" to vote you are also most likely only leaving the wealthy and those with an agenda who allready enjoy great power. You also spend your entire election listening to grand promises to get people just to vote, instead of proposals for a better future.
At least with the unwashed masses voting, the government is less likely to come out with a "let them eat cake" type platform, because you can only push an idiot so far before they fight back. You never always get the best, but you dont get the worst of what a government can do.
The current Australian government would love nothing more than to change the voting system so that it was non-compulsory and harder for young people to vote. Why? Because they know that in that case the only people voting would be the party faithful, and the opposition would never get back in without the support of blue collar workers. But still, for now they have power - they are getting ever so cocky with it too, until the point they go too far and the entire country agrees it is time for a change.
It isnt always what you want, but its better than a constant state pseudo-dictatorship run by the Liberal party (Aust equivalent of the Republicans) because only the rich vote to protect themselves.
Yes - because otherwise only those motivated (i.e with an agenda) vote. Sure, probably 90% of those who dont want to vote will either:
a) donkey vote (number down the ballot) b) vote whoever is the most sexually attractive c) vote based off hearsay.
At any rate, it is still a truer indications of a nations feelings than if only those with a vested interest force their choice of government on the rest. At least with a compulsory vote a goverment cant risk being overtly bad, because you cant always assume the people will be complacent enough to forget.
Sure - and 49% of Americans didnt want Bush either.
Its called democracy, and the problem with it is unless more than 50% of people think like you, you are probably going to be disappointed.
Also I doubt that a majority of voters were considering copyright issues when they cast the ballot, especially when most propaganda/lies/election material focused on interest rates, Iraq and some song and dance about a first time challenger.
Have you stopped to consider that it is the GPL that is incompatible with the other licenses, not the other way around?
For example, you can mix and match code from CDDL, BSD and MPL all you want, there are no restrictions in that regard. When you throw GPL in, you have to license the entire kit and kaboodle under GPL.
Now, if that is a good or bad thing is an entirely seperate debate - while I dont agree in general with the locking up of code, I do understand there are reasons why some entities may not want that to be the case.
Example - a company works with an OSS project and contributes code back on a regular basis. They have an opportunity to provide the project with support for a new deice, but the information they can do this with is covered under an agreement that prevents them releasing the code for this device. Under BSD or CDDL, no problem, they are not required to provide the code. Under the GPL, they cannot do this - development is stifled. Of course, the root of the problem is the locking down of the required information - but often we are talking buisness and competition issues i.e the real world.
Not a bash of the GPL, just pointing out that one size doesnt fit all.
Gee - it took you 3 attempts to answer basic questions like hostname, IP address and nameserver, select what type of install you wanted and then hit install?
Can you please point me to where there is such a credential as a certified Solaris Installation Engineer?
Considering how many people got it going on thousands of machines allready, and your lack of any real detail as to why it failed, the only conclusion to make is you are unable to grasp the concept of using a mouse and keyboard.
"Look like linux" - what exactly does "linux" look like? Oh, you mean it looks like GNOME, which is available on Solaris and Linux and probably a host of other UNIX operating systems....
"know about the licensing issues" - what is that supposed to mean? That because it doesnt use GPL but another OSI approved license it is an "issue"?
"Have something going on here" - well, if that aint flamebait I dont know what is. Yes, Sun have a high quality OS that integrates GNOME and a host of other FOSS software with appropriate licensing and acknoledgements and because you think it looks like _your_ "linux" desktop (and not KDE or blackbox or fvwm or tvm) they are supposedly doing something dastardly?
And really, if OS install snapshots were news worthy, whatever you do dont look at docs.sun.com, there are just too many consipiricies there to report!
Well, because a FILE is defined as a char in the 32-bit ABI (Application Binary Interface) as the man page tells you - you cant change that without breaking the ABI compatibility.
Unlink some other OSes, Solaris has standards and backwards compatability to maintain, which is one of its main strengths.
Whats wrong with just using open() or making a 64 bit binary?
The reason for re-installation is that you can go and verify every file your package database knows about, but not the ones it doesnt.
Plenty of rootkits go and hide themselves in/dev or out of the way places that your packages never would have touched, so you will fix up your packaged files but I doubt there is a r00tkit-1.1337.i386.rpm you can check against.
Sure, it might just leave some stale binaries or scripts around, but unless you go and validate every inode in your filesystem you cant be sure it isnt just going to just open you up to another r00ting again.
And that, kiddies, is why we have backups. (Or at least with Solaris you can jumpstart install/flash it exactly how you want every time).
I can understand it completely. My point is when say you want to admin a single box or consolidate services, and its well within its capabilties to have:
1 CPU for general processes 1 CPU for a database 2 CPUs for applications
And bind the appropriate processes into that sort of a configuration. You are only using Oracle on one CPU, but you pay for 4.
But its not like anyone who is paying Oracle licensing does just buy 4 boxes anyway to do the job of one.
There are much easier ways of offlining resources in Solaris, unfortunately I believe Oracle dont care - if you phyically have X procs in a box, you pay for X (not the X-Y you assign only to Oracle).
For reference, some options: 1. At the OBP, disable the extra procs with asr-disable cpuX 2. In Solaris, use psradm -n to offline the extra CPUs dynamically 3. Create a processor set with psrset, then pbind your oracle processes into that - it will only ever be scheduled within that processor group.
I belive there are other options (like more fine grained SRM), but the above are the simple enough to do "right now".
"The ITU has approved a global industry standard for full-rate ADSL, known as G.992.1, or G.dmt. This specification calls for operation rates of up to 10 Mbps downstream and up to 768 kbps upstream when operating over telephone lines at distances of up to 18,000 feet."
So basically the specification for the signalling has allocated a slab of the frequencies to upload, and the rest to download. Your provider can mix and match within that, but I expect that due to signal attenuation and overall bandwidth demand they limit it to what is currently provided.
Symmetrical DSL (512/512) is designed to run over a shorter loop (so limited customer range), and as most users dont use much upload capacity there is little point in providing it generally - the users that do want it will pay for a service that delivers it.
Blastwave.org is a collection of open source apps/tools/utilities built for Solaris x86 and SPARC.
You install the packages using a pkg-get utility that was modelled off apt-get in Debian I believe - it works great and the software is typcially more current and more integrated than the Sun freeware stuff.
Not quite, Zones do need to copy a fair whack of their own setup into the zone directory, but you can minimise the ammount of duplication with lofs mounts of common applications (like/opt or/apps or something).
You dont need to do anything with LiveUpdate, you install your Solaris 10 machine, then use zonecfg and zoneadm to create and boot/reboot/halt the zone.
The graphical installer is pretty memory intensive now, mainly due to a poxy netscape that runs while you do things. 256M is pretty much what you need if you want to see it, but the text based (or jumpstart) is easy enought to drive.
It should install SSH and Telnet by default, but they may not be enabled - check svcs -a and look for a/network/ssh or/network/telnet service, if you need to start it run svcadm enable/network/ssh and it will be permantently enabled./usr/ucb/cc is a stub link and useless, Solaris 10 should come with gcc in/usr/sfw/bin/gcc for your immediate compiling needs.
The/home problem is because/home is an automount point, not a real directory. You can create a home account in/export/home/user and then add them to/etc/auto_home to automount the home directory in/home. If you want the cheap way you can just add this entry and any/home/blah lookup will result in a mount attempt:
Slashdot Mirror
But we are not talking about a "lets make the IP subsystem understand X traffic in order to generate desktop popups" sort of layer transition. This is purely talking about the storage allocator in ZFS that simply does a number of things allready present in some of the exising storage areas (namely md devices and snapshots). However the reason it does this is to acheive a number of very positive benefits to the upper filesystem layer.
The reason ZFS duplicates functionality is because it *has* to do achieve the sorts of features it is able to provide (read the whitepaper or the OpenSolaris ZFS community materials to see what I mean). I thourghly recommend the to small ZFS tutorial flash movies on the Open Solaris ZFS community - they show the basics of what ZFS can do very concisely.
Traditional filesystems view the disk as a slab of blocks do what they please with. Traditional LVMs view disks as places to abstract into a single block of devices. There is no communication between other than "write ok, yes/no? Read ok, yes/no?". By breaking that model you allow point in time snapshots of the filesystem, not the device. You allow writable clones that share common blocks, you allow the storage subsystem to tune its performance to the filesystem work, rather than model the underlying storage on your idea of what the filesystem is doing.
And so what if the storage allocator is an amorphous blob. Its open source, it does have interfaces (zvols amongst others). Why would you want to get involved in its internals unless it is not doing its job, and if it isnt the source is there to improve it.
So many of the arguments on this topic seem to be of the line "linux allready does this, so why change?" or "I dont get it, so it must suck". But there is nothing in ZFS which says it *must* replace these functions - you can make a zpool of some devices and md/lvm etc of others - it doesnt care. You can make any other filesystem that wants a block device on top of zvols - it will happily do it. If it proves to not compete or do as well, it will fade away with time and die. If it shows to be an improvement (which in my opinion it certainly is, coming from years of SDS+UFS and VXFS+VXVM admin on Solaris) the over time it will probably replace those other features.
Why must there be a line in the sand instead of a productive, usable alternative?
I think you need to look at ZFS for what it is achieving rather than the way it does things differently.
The arguments flying around here seem to be "it is different, is therefore bad" and "I isnt something I understand, so its bad".
The problem with the first argument is that it *has* to be different in order to do what it is aiming for. By making the filesystem talk to the storage, and making the storage talk with the filesystem you remove all the baggage of using RAID to pretend disks are something they are not. I think the simplest case of this is resilvering the disks - if you pull out and replace a disk, because the block subsystem knows what the filesystem has used, it only copies what is needed - try that with a multi-terabyte RAID stripe where with large enough disks you can have a second failure before your first resync is finished.
And I just dont understand the second argument - I throught geeks liked playing with new toys.
Besides, as I understand it the ZFS design allows for certain plug in modules above the block allocator level which is where you really want customisation anyway (I have heard talk of database storage engine plugins and the like)? Letting ZFS take care of all the grunt work underneath makes sence, assuming it does it well and is reliable.
Dtrace is the exact opposite of error vomit, and I dont recall ever hearing it called that anyway. The entire principle is that you dont need to go inserting metric shitloads of debugging and printf("we got here") statements all through your code, recompile it and then see that the error doesnt occur because all your debugging has now slowed your code enough to prevent the race condition that caused the original error.
True - its a L3 and developer tool for the most part, but there are plenty of scripts out there to show what it can do for an admin. Take a look at http://users.tpg.com.au/adsln4yb/dtrace.html for starters. Stuff like iosnoop, iotop, opensnoop and kill.d can be used quite regularly by admins without the need for putting debugging into active applications.
I dont know about the Alan Cox comment, but for IDE this is a common thing. Simply put IDE disks struggle enough for performance, so by default have write caching enabled.
I work for a major server vendor who creates their own firmware for their disks. By default all SCSI and FCAL disks are configured to have write cache disabled because data integrity is valued over performance. For ATA apparently the disk vendors dont give any option for it, so we are unable to work around that.
This is actually quite a pain when it comes to benchmarks, because for SOME tests it makes OSes which enable the write cache to look really fast. Its not until you suffer a catastrophe that you find out the data never made the platter.
And RAID devices dont lie about completing I/O - the device presents an "disk" interface to a slab of battery backed (hopefully) cache to disk which allows write performance to be massivly better. The RAID card itself takes care of syncing its cache to the disks, it just takes the data in cache and responds to the transaction immediately, flushing later. As far as the OS needs to know, the IO is complete - firmware bugs and battery failures aside, the RAID card handles it internally.
And the author seems to be lacking clue about what he is testing - if anything all it is testing is the the OSes ability to get data down to the disks consistantly - all fsync() knows is that the calls it made to send data to the storage devices returned success, its totally dependant on the volume manger, disk and controller drivers as well as the actual physical storage to get the job done.
For all he knows his drivers might be returning immediately just to make performance look better, but actually scheduling the I/O in some manner which causes it to be lost before commitment to storage.
If he wants to complain about the disks, I think he is going to need a much lower level test that a perl script calling sync.
h, and how is Pauline Hanson and One Nation doing these days?
Great - she has been doing ballroom dancing on one of the most popular prime time shows around.
So really, what any normal half brained ex-leader of a extreme right wing racist based politcal party does.
I do picture this because I have been through several Australian elections. Every single advertisement is along those lines i.e blatent bunch of misrepresentation, small disclaimer - vote X.
Sure, a good portion of the populace swallows it hook line and sinker, but if you leave only the "educated" to vote you are also most likely only leaving the wealthy and those with an agenda who allready enjoy great power. You also spend your entire election listening to grand promises to get people just to vote, instead of proposals for a better future.
At least with the unwashed masses voting, the government is less likely to come out with a "let them eat cake" type platform, because you can only push an idiot so far before they fight back. You never always get the best, but you dont get the worst of what a government can do.
The current Australian government would love nothing more than to change the voting system so that it was non-compulsory and harder for young people to vote. Why? Because they know that in that case the only people voting would be the party faithful, and the opposition would never get back in without the support of blue collar workers. But still, for now they have power - they are getting ever so cocky with it too, until the point they go too far and the entire country agrees it is time for a change.
It isnt always what you want, but its better than a constant state pseudo-dictatorship run by the Liberal party (Aust equivalent of the Republicans) because only the rich vote to protect themselves.
Yes - because otherwise only those motivated (i.e with an agenda) vote. Sure, probably 90% of those who dont want to vote will either:
a) donkey vote (number down the ballot)
b) vote whoever is the most sexually attractive
c) vote based off hearsay.
At any rate, it is still a truer indications of a nations feelings than if only those with a vested interest force their choice of government on the rest. At least with a compulsory vote a goverment cant risk being overtly bad, because you cant always assume the people will be complacent enough to forget.
Election time bribes help though.
Sure - and 49% of Americans didnt want Bush either.
Its called democracy, and the problem with it is unless more than 50% of people think like you, you are probably going to be disappointed.
Also I doubt that a majority of voters were considering copyright issues when they cast the ballot, especially when most propaganda/lies/election material focused on interest rates, Iraq and some song and dance about a first time challenger.
I saw a screenshot yesterday of a 8 way dual core Opteron box showing 16 graphs.
Cant say where though.
Sun, HP and IBM have allready got machines ready, just waiting for launch.
p teron_dualcore/
http://www.channelregister.co.uk/2005/04/08/amd_o
Have you stopped to consider that it is the GPL that is incompatible with the other licenses, not the other way around?
For example, you can mix and match code from CDDL, BSD and MPL all you want, there are no restrictions in that regard. When you throw GPL in, you have to license the entire kit and kaboodle under GPL.
Now, if that is a good or bad thing is an entirely seperate debate - while I dont agree in general with the locking up of code, I do understand there are reasons why some entities may not want that to be the case.
Example - a company works with an OSS project and contributes code back on a regular basis. They have an opportunity to provide the project with support for a new deice, but the information they can do this with is covered under an agreement that prevents them releasing the code for this device. Under BSD or CDDL, no problem, they are not required to provide the code. Under the GPL, they cannot do this - development is stifled. Of course, the root of the problem is the locking down of the required information - but often we are talking buisness and competition issues i.e the real world.
Not a bash of the GPL, just pointing out that one size doesnt fit all.
Sun Mangement Console has replaced admintool - its a little heavy on the Java but it does what you want - admin in a GUI: /usr/sbin/smc
Gee - it took you 3 attempts to answer basic questions like hostname, IP address and nameserver, select what type of install you wanted and then hit install?
Can you please point me to where there is such a credential as a certified Solaris Installation Engineer?
Considering how many people got it going on thousands of machines allready, and your lack of any real detail as to why it failed, the only conclusion to make is you are unable to grasp the concept of using a mouse and keyboard.
Is someone posting for you?
Wow - such artistic expression, you captured it beautifully....
Try this:
h tm l
http://www.docbert.org/Solaris/Jumpstart/linux.
The author knows what he is talking about - I havent tried it myself but this ought to be what you need.
"Look like linux" - what exactly does "linux" look like? Oh, you mean it looks like GNOME, which is available on Solaris and Linux and probably a host of other UNIX operating systems....
"know about the licensing issues" - what is that supposed to mean? That because it doesnt use GPL but another OSI approved license it is an "issue"?
"Have something going on here" - well, if that aint flamebait I dont know what is. Yes, Sun have a high quality OS that integrates GNOME and a host of other FOSS software with appropriate licensing and acknoledgements and because you think it looks like _your_ "linux" desktop (and not KDE or blackbox or fvwm or tvm) they are supposedly doing something dastardly?
And really, if OS install snapshots were news worthy, whatever you do dont look at docs.sun.com, there are just too many consipiricies there to report!
Well, because a FILE is defined as a char in the 32-bit ABI (Application Binary Interface) as the man page tells you - you cant change that without breaking the ABI compatibility.
Unlink some other OSes, Solaris has standards and backwards compatability to maintain, which is one of its main strengths.
Whats wrong with just using open() or making a 64 bit binary?
The reason for re-installation is that you can go and verify every file your package database knows about, but not the ones it doesnt.
/dev or out of the way places that your packages never would have touched, so you will fix up your packaged files but I doubt there is a r00tkit-1.1337.i386.rpm you can check against.
Plenty of rootkits go and hide themselves in
Sure, it might just leave some stale binaries or scripts around, but unless you go and validate every inode in your filesystem you cant be sure it isnt just going to just open you up to another r00ting again.
And that, kiddies, is why we have backups. (Or at least with Solaris you can jumpstart install/flash it exactly how you want every time).
I can understand it completely. My point is when say you want to admin a single box or consolidate services, and its well within its capabilties to have:
1 CPU for general processes
1 CPU for a database
2 CPUs for applications
And bind the appropriate processes into that sort of a configuration. You are only using Oracle on one CPU, but you pay for 4.
But its not like anyone who is paying Oracle licensing does just buy 4 boxes anyway to do the job of one.
There are much easier ways of offlining resources in Solaris, unfortunately I believe Oracle dont care - if you phyically have X procs in a box, you pay for X (not the X-Y you assign only to Oracle).
For reference, some options:
1. At the OBP, disable the extra procs with asr-disable cpuX
2. In Solaris, use psradm -n to offline the extra CPUs dynamically
3. Create a processor set with psrset, then pbind your oracle processes into that - it will only ever be scheduled within that processor group.
I belive there are other options (like more fine grained SRM), but the above are the simple enough to do "right now".
This page has some of the gory details:
"The ITU has approved a global industry standard for full-rate ADSL, known as G.992.1, or G.dmt. This specification calls for operation rates of up to 10 Mbps downstream and up to 768 kbps upstream when operating over telephone lines at distances of up to 18,000 feet."
So basically the specification for the signalling has allocated a slab of the frequencies to upload, and the rest to download. Your provider can mix and match within that, but I expect that due to signal attenuation and overall bandwidth demand they limit it to what is currently provided.
Symmetrical DSL (512/512) is designed to run over a shorter loop (so limited customer range), and as most users dont use much upload capacity there is little point in providing it generally - the users that do want it will pay for a service that delivers it.
Blastwave.org is a collection of open source apps/tools/utilities built for Solaris x86 and SPARC.
You install the packages using a pkg-get utility that was modelled off apt-get in Debian I believe - it works great and the software is typcially more current and more integrated than the Sun freeware stuff.
Not quite, Zones do need to copy a fair whack of their own setup into the zone directory, but you can minimise the ammount of duplication with lofs mounts of common applications (like /opt or /apps or something).
You dont need to do anything with LiveUpdate, you install your Solaris 10 machine, then use zonecfg and zoneadm to create and boot/reboot/halt the zone.
The graphical installer is pretty memory intensive now, mainly due to a poxy netscape that runs while you do things. 256M is pretty much what you need if you want to see it, but the text based (or jumpstart) is easy enought to drive.
/network/ssh or /network/telnet service, if you need to start it run svcadm enable /network/ssh and it will be permantently enabled. /usr/ucb/cc is a stub link and useless, Solaris 10 should come with gcc in /usr/sfw/bin/gcc for your immediate compiling needs.
/home problem is because /home is an automount point, not a real directory. You can create a home account in /export/home/user and then add them to /etc/auto_home to automount the home directory in /home. If you want the cheap way you can just add this entry and any /home/blah lookup will result in a mount attempt:
It should install SSH and Telnet by default, but they may not be enabled - check svcs -a and look for a
The
* localhost:/export/home/&
Good luck and have fun!