Also, forgot to mention we seem to be in agreement. I just meant that in the command given above, the 8K sizes were not set and that performance would suffer. I saw the same issue with Gemstone/S OODB.
Yes, I am aware of that. However, even with setting the ARC size in the kernel with parameter for a p size of 256MB (in hex) a c size of 256MB and a c max of 512MB it doesn't adhere to the limitations.
I don't think Novell paid MS to get MS to pay more. I think the plan was so that MS could walk away and tell people Novell paid them. I think this is FUD from MS (and I don't blame Novell too much for the actions of a few execs who though they were doing something smart). I think they are afraid of losing market share and wanted a piece of "That Linux thing"
You forgot to set the blocksize to oracle's preferred 8K. You also forgot to set the read size to 8k so you don't take in 128K pages every time you only want 8K random reads. You also forgot to limit the ARC size (buffers) so that ORACLE will not have to contend with ZFS for memory resources.
With enough spindles you could use it on a DB server if you set the max ARC size to a smaller number to keep the cache from growing to insane sizes. You would also want to create the volumes with the write block size that the database uses for random reads (usually 8K) and also set the read size to the same value. Then you should be able to get performance similar to other file systems like UFS. You would lose some of the features that you would probably not care about. Realize ZFS has not been "out that long" relative to many other stable file systems, and as such a huge undertaking, it is going to have major bugs to squash. I think it was released well before it was primetime, but it is getting better.
After 2 months of trying ZFS out in a non-prod environment, we reverted to UFS because ZFS was not as fast on 8k random reads and appeared to use a ton more RAM. In actuality, it used less than perceived but try explaining that to developers and DBAs. UFS will use system RAM for cache and report it as free RAM, then relase it when another process needs it. ZFS does similar things (with worse memory accounting) though if it uses system RAM for cache it reports it as used. A bug had to be squashed in ZFS regarding purging pages of cache when the system requested it back, because it would page out faster than it could account for doing so and cause massive thrashing in RAM. That has been fixed, however.
It still has memory hogging issues as well as performance issues in certain areas. More kernel tuning will be needed to tame the beast that is ZFS. It is good for many things but it does not replace EVERYTHING just yet.
Ahh, the "workstation" comment is what explains it to me. As of now, our workstations (are not running *nix) are Windows laptops and we have a server base of hundreds of various UNIX and Linux boxes.
Not claiming to be a seasoned veteran admin, but couldn't something like Zenworks allow you to control access in the manner you described for NIS and LDAP? I am experienced and continuing to learn as much as I can, but don't think I am the stink on shit. I realize that there are admins with less experience that I can learn from, as well as those with more that I can teach. I think that the issue is that there are very complex infrastructures and the need for more people to manage them. People tend to gravitate towards the piece that interest them most. I happen to like storage management and may go more towards that route.
Because the law abiding citizens this imposes restrictions on.........THEY are the ones we must stop. Aside from recent issues (VT) there have been only flies among 757s where legal gun owners are the ones who commit violent crimes. Many legal gun owners actually STOP crimes. Why should that matter... Feel good legislation is where it is at, man
I have been fired a couple of times (only once in IT, the rest were in BS jobs in/before college. The IT one was because the company couldn't afford to keep paying me). As for the other part of what you said, one of my bosses was pissed that I was leaving because they just assumed I would be there forever. They felt cheated that they would not have me around 24/7 for when emergencies came up. Now I am contracting with them for when they need me, for training, and on retainer:)
Depending on the needs and configuration, however, I have found UFS to be up to 5x faster than ZFS for reading small blocks of data "
One more thing, ZFS will try to use almost all memory as a cache. This is bad for databases or performance applications. ZFS will release memory when an application, which should have a higher priority, needs it, but there is a bug in some versions of ZFS that, due to the bad memory accounting, releases too many pages of memory, and too fast, causing thrashing and a performance hit. For a file server, ZFS is great, but it still has a ways to go for some applications.
Also, if you are using a controller (as on a SAN) that has read/write order battery backed cache, you will want to disable fsync() for ZIL in the/etc/system file. There is also a nice script that can be used to limit the amount of memory ZFS uses, though, again, due to the poor memory accounting, it doesn't work that great. I have limited ZFS to 512MB of RAM to use and it grabs 3GB. Solaris 10 u4 is supposed to address some of this. Because of ZIL (part of the integrity check/scrubber) performance for multiple writes suffers and it can make using it as an NFS server less than ideal. It has a LOT of potential, and with enough spindles it can overcome some of these issues temporarily. It just has not matured to the level at which they are selling ZFS.
How about Icarus? They are flying too close to the sun on this one. I hope their wings don't get melted and cause them to fall into the sea for their hubris.
It was not a misspelling, but rather a colloquialism of sorts. I chose not to spell it as fucking because that was not the impression I was trying to give.
Frig also is a word unto itself, and not a misspelling of fuck. Go look it up
Funny that you mention RouterOS. My company (actually, I am leaving them very soon) uses routerboard routers with RouterOS on them in place of Cisco stuff because it is cheaper and far more functional (easier to use too). The boxes are small, very cheap and work well. I think we had to reboot ours recently, after almost 350 days of uptime, only because we had to move it.
Split tunnels do that. My company uses a Juniper/netscreen/Neoteris SSL VPN (they were progressively bought out, originating as Neoteris, then Netscreen, then Juniper IVE SA series. They ROCK. You have a "clientless" VPN that can support multiple users without having to configure a client on the remote machine. They log into a web portal and then can launch (or your policy can auto launch) Network Connect. It is unobtrusive, and very friendly (except a minor bug that can cause problems if you lose your connection and the IVE box thinks the connection is active. Then you won't get another connection, but that is rare). It supports Mac OSX, Linux, Windows 2000/XP etc. All you need is a browser with either ActiveX (bleh) or Java. You could even send the user the small binaries that the IVE installs and they don't have to run to the webpage. It is very granular and plugs in with other directory services (Active Directory, etc) or RADIUS, local users, whatever.
SecuRemote/SecureClient sucks hard big balls. My company still uses CP for our firewall, but we have replaced that horrid VPN client (constantly broke remote user's network settings) with a Juniper SSL VPN. However, the Cisco VPN clients we have used sucked harder, so perhaps you were right in saying that they were not worse than others.
Slashdot Mirror
Also, forgot to mention we seem to be in agreement. I just meant that in the command given above, the 8K sizes were not set and that performance would suffer. I saw the same issue with Gemstone /S OODB.
http://www.solarisinternals.com/wiki/index.php/ZFS _Best_Practices_Guide#Memory_and_Dynamic_Reconfigu ration_Recommendations
is a good place to start looking, but as you can see here:
http://www.opensolaris.org/jive/thread.jspa?messag eID=122983
You can use the command located here:- does-zfs-consume.html
http://milek.blogspot.com/2006/09/how-much-memory
to show how much is being consumed by ZIO buffers.
I don't think Novell paid MS to get MS to pay more. I think the plan was so that MS could walk away and tell people Novell paid them. I think this is FUD from MS (and I don't blame Novell too much for the actions of a few execs who though they were doing something smart). I think they are afraid of losing market share and wanted a piece of "That Linux thing"
You fail it
double secret probation?
To be fair, Novell PAID MS a lot of money. However, MS PAID Novell a LOT more :)
With enough spindles you could use it on a DB server if you set the max ARC size to a smaller number to keep the cache from growing to insane sizes. You would also want to create the volumes with the write block size that the database uses for random reads (usually 8K) and also set the read size to the same value. Then you should be able to get performance similar to other file systems like UFS. You would lose some of the features that you would probably not care about. Realize ZFS has not been "out that long" relative to many other stable file systems, and as such a huge undertaking, it is going to have major bugs to squash. I think it was released well before it was primetime, but it is getting better.
After 2 months of trying ZFS out in a non-prod environment, we reverted to UFS because ZFS was not as fast on 8k random reads and appeared to use a ton more RAM. In actuality, it used less than perceived but try explaining that to developers and DBAs. UFS will use system RAM for cache and report it as free RAM, then relase it when another process needs it. ZFS does similar things (with worse memory accounting) though if it uses system RAM for cache it reports it as used. A bug had to be squashed in ZFS regarding purging pages of cache when the system requested it back, because it would page out faster than it could account for doing so and cause massive thrashing in RAM. That has been fixed, however.
It still has memory hogging issues as well as performance issues in certain areas. More kernel tuning will be needed to tame the beast that is ZFS. It is good for many things but it does not replace EVERYTHING just yet.
Ahh, the "workstation" comment is what explains it to me. As of now, our workstations (are not running *nix) are Windows laptops and we have a server base of hundreds of various UNIX and Linux boxes.
Not claiming to be a seasoned veteran admin, but couldn't something like Zenworks allow you to control access in the manner you described for NIS and LDAP? I am experienced and continuing to learn as much as I can, but don't think I am the stink on shit. I realize that there are admins with less experience that I can learn from, as well as those with more that I can teach. I think that the issue is that there are very complex infrastructures and the need for more people to manage them. People tend to gravitate towards the piece that interest them most. I happen to like storage management and may go more towards that route.
I believe there was a | between them, perhaps to differentiate them or assume a choice was to be made
no, he meant NIS, a type of domain system where users are known via an NIS master. You can work without it.
Because the law abiding citizens this imposes restrictions on.........THEY are the ones we must stop. Aside from recent issues (VT) there have been only flies among 757s where legal gun owners are the ones who commit violent crimes. Many legal gun owners actually STOP crimes. Why should that matter... Feel good legislation is where it is at, man
I have been fired a couple of times (only once in IT, the rest were in BS jobs in/before college. The IT one was because the company couldn't afford to keep paying me). As for the other part of what you said, one of my bosses was pissed that I was leaving because they just assumed I would be there forever. They felt cheated that they would not have me around 24/7 for when emergencies came up. Now I am contracting with them for when they need me, for training, and on retainer :)
Also, if you are using a controller (as on a SAN) that has read/write order battery backed cache, you will want to disable fsync() for ZIL in the /etc/system file. There is also a nice script that can be used to limit the amount of memory ZFS uses, though, again, due to the poor memory accounting, it doesn't work that great. I have limited ZFS to 512MB of RAM to use and it grabs 3GB. Solaris 10 u4 is supposed to address some of this. Because of ZIL (part of the integrity check/scrubber) performance for multiple writes suffers and it can make using it as an NFS server less than ideal. It has a LOT of potential, and with enough spindles it can overcome some of these issues temporarily. It just has not matured to the level at which they are selling ZFS.
It will be dog slow no matter HOW MANY "Type R" stickers you put on that turd
How about Icarus? They are flying too close to the sun on this one. I hope their wings don't get melted and cause them to fall into the sea for their hubris.
Frig also is a word unto itself, and not a misspelling of fuck. Go look it up
http://www.answers.com/topic/frig
losing. How frigging hard is it to type LOSING instead of LOOSING?
You would want to make sure you went with something else, completely avoiding openSSH, apache, GNU/Linux, BSD, etc. Nope.....Windows for you, Sir
Funny that you mention RouterOS. My company (actually, I am leaving them very soon) uses routerboard routers with RouterOS on them in place of Cisco stuff because it is cheaper and far more functional (easier to use too). The boxes are small, very cheap and work well. I think we had to reboot ours recently, after almost 350 days of uptime, only because we had to move it.
Split tunnels do that. My company uses a Juniper/netscreen/Neoteris SSL VPN (they were progressively bought out, originating as Neoteris, then Netscreen, then Juniper IVE SA series. They ROCK. You have a "clientless" VPN that can support multiple users without having to configure a client on the remote machine. They log into a web portal and then can launch (or your policy can auto launch) Network Connect. It is unobtrusive, and very friendly (except a minor bug that can cause problems if you lose your connection and the IVE box thinks the connection is active. Then you won't get another connection, but that is rare). It supports Mac OSX, Linux, Windows 2000/XP etc. All you need is a browser with either ActiveX (bleh) or Java. You could even send the user the small binaries that the IVE installs and they don't have to run to the webpage. It is very granular and plugs in with other directory services (Active Directory, etc) or RADIUS, local users, whatever.
SecuRemote/SecureClient sucks hard big balls. My company still uses CP for our firewall, but we have replaced that horrid VPN client (constantly broke remote user's network settings) with a Juniper SSL VPN. However, the Cisco VPN clients we have used sucked harder, so perhaps you were right in saying that they were not worse than others.
Netscreen (By Juniper Networks), Astaro Firewall, and a relative newcomer, ZyWall by ZyXel. They should all work REALLY well.