Agreed. TN panels... such suck. PVA does too. I'll stick with my professional series wide-gamut NEC monitor (LG panel). Great, ACCURATE colors. Very even backlighting. Lots of desk realestate.
Quite aware of that. Thats not a bad application - its basically what VNC was designed for, remote desktop access.
WonderWare is a different beast. Its SCADA system lets you monitor sensors and perform actions based on this sensors, either manually or automatic. Instead of making a client/server system where the 'server' does the control and data storage portion, and the clients connect to the server to view and do setup, wonderware does the odd approach. In order to view it remotely, the correct solution is to use remote desktop to another instance on the remote machine. No monitoring webpage. No actual data connection to the server. Remote Desktop. And its an advertised feature...
And I'm not even going into the rest of the system, which is just as bad.
Too many people are cheaping out in screens. Dell's 20 and 24inchers used to be IPS panels, but now have mostly reverted to the faster but lower quality (viewing angle color and gamma shift primarily) PVA and MVA panels.
Finding good quality IPS panels in consumer products is now mostly impossible. You need to step up into Pro quality and then even some are MVA/PVA panels, such as the Samsung XL30. Luckily NEC still uses IPS panels in the 2690 and 3090 monitors.
Placing your average cabinet over a perf tile will do... nothing.
Servers vent, in their standard setup, front to back. While a vertical system with a hot air plenum is fantastic, its going to require specialized hardware, and since most datacenters are a odd mix of new and old, is never going to happen. Google could afford to do it, but Random Company X or Colo Y can't.
If you're talking about a full cold-aisle containment system, in which a standard cold-air subfloor is much more closely controlled, then by all means will it save energy. There is a lot of momentum in fixing problems, as well as monitoring and control stuff (see HP Dynamic Smart Cooling, or SynapSense stuff.
I've learned to buy Asus and SuperMicro boards only as well. You can't beat the SuperMicro boards when you need a solid but still affordable Dual-Xeon setup with 16GB of RAM:)
Now for what really matters, how do those keyboards do for coding?:-)
I'm always interested in "better" keyboards for large volumes of text entry. It does get minus points for putting the Ctrl key in the wrong spot - who uses capslock anyway?
For someone who has been using OpenSolaris (SXCE) as a server platform for Apache, ZFS, etc for awhile now, I welcome an easy to upgrade and improved userspace Solaris. Will try this one out.
Solaris has had a relatively poor userspace experience for someone used to Linux machines. The kernel is top-notch though.
Intel probably boils down to a performance and shader problem.
You haven't used ATI's Linux drivers have you? The fact that they work at all is a miracle. The current open source 'radeon' driver only supports older (non X----) cards. The fglrx driver is a POS: buggy, unreliable- especially in 3d operations, and nowhere near the polish and functionality of the NVidia driver. Yes, the nvidia driver is closed source, but at least it works.
The efficiency of power supplies is also non-linear. They only peak when under a fair load. If a switching supply is minimally loaded, its efficiency is very bad (50%).
Java dynamic code tends to be pretty well behaved, and is actually heavily used. For instance, EJB 3.0 makes very heavy use of the reflection API to auto generate RMI wrappers, maintain object pools, etc, with no heavy lifting from the developer.
Because there is no 100% portable method for naming UNIX sockets, named pipes, and whatever else is in use out there. Java sometimes picks the lowest common denominator for its utilities in an effort to not introduce headaches. Its really a design compromise. IP is mostly universal though, so IP sockets and addressing was chosen. If jEdit binds to the loopback device only, what difference does it make?
Thats the plus part of Java. There is a Java defined way of doing things (byte order, separators, etc). All you need is a competent JVM (which is not as portable:))
Linux has it too (microcode updates have been available since the Pentium Pro days). The microcode update is soft though, the original microcode is restored on a reboot. The best fix would be a relevant BIOS update which loads the patched microcode much earlier.
May I recommend this setup. Its very stable, the Xeon class server and workstation boards are very solid (using a Supermicro X7DAL-E), and the performance for software development is simply unmatched. Everything on a good Intel 5000X based board is supported by Linux as well. Run a 64-bit distribution and pack it full of RAM (FB-DIMMs are interesting beasts - various ways to deal with ranked memory and exploit the parallel nature of the serial memory bus and trading latency for bandwidth and vice versa)
The market segment is very different. There are no clock tuning widgets in the BIOS (but you can do serial BIOS redirection), no splashy other stuff. Its designed as a very boring very well performing workstation. If LED light fans and overclocking excite you, you're better off with a Core2 Duo/Quad system.
Yes, pretty much. Raidz2 is a type of element of the pool, not 2 raidzs. So you can lose at most two drives from each raidz2 component and be ok (the mean time between data loss is very good for double parity).
Mirror is of course the best in terms of both speed and reliability. ZFS supports N-way mirrors (multiple 3 way mirrors has a very high MTDL)
For the reasons you stated, ZFS doesn't try automagically making things fault tolerant, leaving it up to the human to figure out.
So when you add disks, you add them in mirrors, raidz (raid5ish), or raidz2 (raid6ish) vdevs. The trick is added volumes automatically become part of the underlying storage pool, and data gets striped over these volumes. You can also upgrade capacity by systematically removing drives from a mirror/raidz vdev, waiting for a recover, and repeat. Once the smallest drive in the mirror/raid vdev are replaced, the capacity instantly increases.
It is possible to have a pool of mixed mirrors and raidz sets, but it is not recommended (differing levels of redundancy - as data is striped over all the sets the weakest point can fail everything).
Slashdot Mirror
Agreed. TN panels... such suck. PVA does too. I'll stick with my professional series wide-gamut NEC monitor (LG panel). Great, ACCURATE colors. Very even backlighting. Lots of desk realestate.
WonderWare is a different beast. Its SCADA system lets you monitor sensors and perform actions based on this sensors, either manually or automatic. Instead of making a client/server system where the 'server' does the control and data storage portion, and the clients connect to the server to view and do setup, wonderware does the odd approach. In order to view it remotely, the correct solution is to use remote desktop to another instance on the remote machine. No monitoring webpage. No actual data connection to the server. Remote Desktop. And its an advertised feature...
And I'm not even going into the rest of the system, which is just as bad.
And an obvious reason why a spreadsheet program is not how you deal with that kind of data.
When the 'remote access' and 'server based' system is Remote Desktop Client, you know to run away, very very quickly.
Interestingly enough, the Lightroom database is an SQLite 3 file. :)
FireGPG will provide PGP based public key cryptography in Firefox as well.
Finding good quality IPS panels in consumer products is now mostly impossible. You need to step up into Pro quality and then even some are MVA/PVA panels, such as the Samsung XL30. Luckily NEC still uses IPS panels in the 2690 and 3090 monitors.
Servers vent, in their standard setup, front to back. While a vertical system with a hot air plenum is fantastic, its going to require specialized hardware, and since most datacenters are a odd mix of new and old, is never going to happen. Google could afford to do it, but Random Company X or Colo Y can't.
If you're talking about a full cold-aisle containment system, in which a standard cold-air subfloor is much more closely controlled, then by all means will it save energy. There is a lot of momentum in fixing problems, as well as monitoring and control stuff (see HP Dynamic Smart Cooling, or SynapSense stuff.
I've learned to buy Asus and SuperMicro boards only as well. You can't beat the SuperMicro boards when you need a solid but still affordable Dual-Xeon setup with 16GB of RAM :)
Never liked it. I'm fine with Qwerty, I'm just looking more for key-press-quality and tactile feedback.
I'm always interested in "better" keyboards for large volumes of text entry. It does get minus points for putting the Ctrl key in the wrong spot - who uses capslock anyway?
Solaris has its strengths (ZFS, etc), but it also has its weaknesses (why would you use it on an average desktop system?)
Why they didn't just choose apt/dpkg is probably having to do with licensing.
For someone who has been using OpenSolaris (SXCE) as a server platform for Apache, ZFS, etc for awhile now, I welcome an easy to upgrade and improved userspace Solaris. Will try this one out. Solaris has had a relatively poor userspace experience for someone used to Linux machines. The kernel is top-notch though.
Rush Hour U-Drive-It sucked.
Rush hour had new road types and the intensely useful commute graphs. So it wasn't all bad.
Societies is horrible though.
Intel probably boils down to a performance and shader problem.
You haven't used ATI's Linux drivers have you? The fact that they work at all is a miracle. The current open source 'radeon' driver only supports older (non X----) cards. The fglrx driver is a POS: buggy, unreliable- especially in 3d operations, and nowhere near the polish and functionality of the NVidia driver. Yes, the nvidia driver is closed source, but at least it works.
The efficiency of power supplies is also non-linear. They only peak when under a fair load. If a switching supply is minimally loaded, its efficiency is very bad (50%).
Java dynamic code tends to be pretty well behaved, and is actually heavily used. For instance, EJB 3.0 makes very heavy use of the reflection API to auto generate RMI wrappers, maintain object pools, etc, with no heavy lifting from the developer.
But you lose some power efficiency from the use of power sucking FB-DIMMS.
Because there is no 100% portable method for naming UNIX sockets, named pipes, and whatever else is in use out there. Java sometimes picks the lowest common denominator for its utilities in an effort to not introduce headaches. Its really a design compromise. IP is mostly universal though, so IP sockets and addressing was chosen. If jEdit binds to the loopback device only, what difference does it make?
Thats the plus part of Java. There is a Java defined way of doing things (byte order, separators, etc). All you need is a competent JVM (which is not as portable :))
Linux has it too (microcode updates have been available since the Pentium Pro days). The microcode update is soft though, the original microcode is restored on a reboot. The best fix would be a relevant BIOS update which loads the patched microcode much earlier.
May I recommend this setup. Its very stable, the Xeon class server and workstation boards are very solid (using a Supermicro X7DAL-E), and the performance for software development is simply unmatched. Everything on a good Intel 5000X based board is supported by Linux as well. Run a 64-bit distribution and pack it full of RAM (FB-DIMMs are interesting beasts - various ways to deal with ranked memory and exploit the parallel nature of the serial memory bus and trading latency for bandwidth and vice versa)
The market segment is very different. There are no clock tuning widgets in the BIOS (but you can do serial BIOS redirection), no splashy other stuff. Its designed as a very boring very well performing workstation. If LED light fans and overclocking excite you, you're better off with a Core2 Duo/Quad system.
Yes, pretty much. Raidz2 is a type of element of the pool, not 2 raidzs. So you can lose at most two drives from each raidz2 component and be ok (the mean time between data loss is very good for double parity).
S _Best_Practices_Guide
Mirror is of course the best in terms of both speed and reliability. ZFS supports N-way mirrors (multiple 3 way mirrors has a very high MTDL)
http://www.solarisinternals.com/wiki/index.php/ZF
For the reasons you stated, ZFS doesn't try automagically making things fault tolerant, leaving it up to the human to figure out.
So when you add disks, you add them in mirrors, raidz (raid5ish), or raidz2 (raid6ish) vdevs. The trick is added volumes automatically become part of the underlying storage pool, and data gets striped over these volumes. You can also upgrade capacity by systematically removing drives from a mirror/raidz vdev, waiting for a recover, and repeat. Once the smallest drive in the mirror/raid vdev are replaced, the capacity instantly increases.
It is possible to have a pool of mixed mirrors and raidz sets, but it is not recommended (differing levels of redundancy - as data is striped over all the sets the weakest point can fail everything).