GIS workstation setup tips - Revision history

Wiki-Pcreso: /* RAID Drives */

2013-07-20T20:43:53Z

RAID Drives

Wiki-Pcreso: /* CPU */

2013-07-20T20:20:35Z

CPU

Wiki-Pcreso: /* CPU */

2013-07-20T20:19:38Z

CPU

Wiki-HamishBowman: /* RAID Drives */

2013-07-07T01:42:54Z

RAID Drives

Wiki-HamishBowman: /* RAID Drives */

2013-07-07T01:42:12Z

RAID Drives

Wiki-HamishBowman: /* Operating System */

2013-07-07T01:40:47Z

Operating System

Wiki-HamishBowman: /* RAID Drives */ softraid

2013-07-07T01:39:22Z

RAID Drives: softraid

Wiki-HamishBowman: /* Raid Drives */ ...

2013-07-07T01:37:26Z

Raid Drives: ...

Wiki-HamishBowman: /* CPU */

2013-07-07T01:32:41Z

CPU

Wiki-Nikos: /* RAM */

2012-10-25T23:19:33Z

RAM

← Older revision		Revision as of 20:43, 20 July 2013
Line 38:		Line 38:

	RAID 1 offers the same close to double I/O benefits, and full hardware failure redundancy, but at the cost of overall capacity: it will be limited to the size of one drive (as the second is the full mirror).		RAID 1 offers the same close to double I/O benefits, and full hardware failure redundancy, but at the cost of overall capacity: it will be limited to the size of one drive (as the second is the full mirror).
		+
		+	Hardware RAID systems are no longer the province of high end servers, with effective several Tb 4-8 SATA disk external e-SATA/NAS storage units from companies like Arena, Stardom, Raidon etc., around USD300 These are useful for GIS users wanting to manage large vector or raster databases, with good RAID support, hot swap & auto rebuild capabilities. These won't compete with SSD's for sheer throughput, but are much cheaper for larger volumes.

	== Operating System ==		== Operating System ==

@@ Line 7: / Line 7: @@
 The high end Intel i7 cpu's are the fastest (as of 2012 & in the forseeable future) but bang for the buck decreases radically. Given how most GIS related software does not use multiple threads or cores particularly well, getting many cores is not critical. As of late 2012 I'd suggest an AMD FX8150 or Intel i7-3820 as suitable high end cpu's at sensible prices. AMD about 40% cheaper & 20% slower (if 8 seconds vs 10 is worth the $$ is your call :-)
-Note that most FOSS GIS usres are concurrently running multiple applications, such as using QGIS to view Postgis data. This runs both applications concurrently, so in this sort of usage, where multiple programs (processes) are being run concurrently, you can (if cpu is not the bottleneck) fully utilise multiple cores for your GIS work.
+Note that most FOSS GIS users are concurrently running multiple applications, such as using QGIS to view Postgis data. This runs both applications concurrently, so in this sort of usage, where multiple programs (processes) are being run togther, you can (if cpu is not the bottleneck) fully utilise multiple cores for your GIS work.
 === RAM ===

@@ Line 7: / Line 7: @@
 The high end Intel i7 cpu's are the fastest (as of 2012 & in the forseeable future) but bang for the buck decreases radically. Given how most GIS related software does not use multiple threads or cores particularly well, getting many cores is not critical. As of late 2012 I'd suggest an AMD FX8150 or Intel i7-3820 as suitable high end cpu's at sensible prices. AMD about 40% cheaper & 20% slower (if 8 seconds vs 10 is worth the $$ is your call :-)
-Note that using QGIS to view Postgis data will run both applications concurrently, so in this sort of usage, where multiple programs (processes) are being run concurrently, you can (if cpu is not the bottleneck) fully utilise multiple cores for your GIS work.
+Note that most FOSS GIS usres are concurrently running multiple applications, such as using QGIS to view Postgis data. This runs both applications concurrently, so in this sort of usage, where multiple programs (processes) are being run concurrently, you can (if cpu is not the bottleneck) fully utilise multiple cores for your GIS work.
 === RAM ===

@@ Line 33: / Line 33: @@
 # <strike>It's hardware based</strike> (Linux software RAID works fine and trades slightly slower performance for better written code and tighter OS integration than is found in hardware solutions)
 # You run drive health monitoring tools on a regular basis (<tt>smartmontools</tt>) AND
-# You know how to rebuild it if it fails
+# You know how to rebuild it if it fails (<tt>mdadm</tt>)
 My addendum to the advice above: software RAID (MD RAID) on Linux works well. A 2-disk RAID 0 setup will close to double disk I/O on both reads and writes. But you have doubled the chances of a disk failure, either one will lose all the stored data. You decide if the performance boost is worth the risk for your purposes.

@@ Line 32: / Line 32: @@
 RAIDs are only good if:
 # <strike>It's hardware based</strike> (Linux software RAID works fine and trades slightly slower performance for better written code and tighter OS integration than is found in hardware solutions)
-# You run drive health monitoring tools on a regular basis AND
+# You run drive health monitoring tools on a regular basis (<tt>smartmontools</tt>) AND
 # You know how to rebuild it if it fails

@@ Line 46: / Line 46: @@
 When I first was introduced to Linux I tried a few other distributions. This included ArkLinux, Fedora, OpenSuse and Gentoo. After some advice from my local Linux User Group I tried Debian. I don't think I'd ever have been successful with Linux if it hadn't been for Debian.
-## another perspective from another Linux GIS user: Debian is one of the Linux purists distros, not ideal for new users, but a good server distro. OpenSuse & Ubuntu have the best support for GIS applications, I use both, generally prefer OpenSuse, & I find the GIS repository maintainers for OpenSuse are more responsive to questions & have more up to date packages. Just my impression. Both are good choices.
+:: another perspective from another Linux GIS user: Debian is one of the Linux purists distros, not ideal for new users, but a good server distro. OpenSuse & Ubuntu have the best support for GIS applications, I use both, generally prefer OpenSuse, & I find the GIS repository maintainers for OpenSuse are more responsive to questions & have more up to date packages. Just my impression. Both are good choices.
 Linux lets users "get their hands dirty", with much better support for scripting & command line applications generally, which is often useful to a serious GIS user. It is also easier to build a complete GIS workstation on Linux, using totally free software than it is under Windows, due to the GIS package repositories that provide the software ready to go for your system (well, OpenSuse & Ubuntu anyway, & Fedora isn't far behind.).

@@ Line 29: / Line 29: @@
 In particular, an SSD drive to boot from & a large SATA drive to store data files. An SSD for swap is also likely to help performance if swapping is required.
-=== Raid Drives ===
+=== RAID Drives ===
-Raids are only good if:
+RAIDs are only good if:
-# It's hardware based
+# <strike>It's hardware based</strike> (Linux software RAID is fine and trades slower performance for better written code)
 # You run drive health monitoring tools on a regular basis AND
 # You know how to rebuild it if it fails
-My addendum to the advice above: software raid (MD Raid) on Linux works well. A 2-disk RAID 0 setup will close to double disk I/O on both reads and writes. But you have doubled the chances of a disk failure, either one will lose all the stored data. You decide if the performance boost is worth the risk for your purposes.
+My addendum to the advice above: software RAID (MD RAID) on Linux works well. A 2-disk RAID 0 setup will close to double disk I/O on both reads and writes. But you have doubled the chances of a disk failure, either one will lose all the stored data. You decide if the performance boost is worth the risk for your purposes.
 == Operating System ==

@@ Line 5: / Line 5: @@
 Multi-core is in and there are many steps(grades or variants), if you plan to run virtual machines make sure to get something with the virtualization instruction set.
-The high end Intel I7 cpu's are the fastest (as of 2012 & in the forseeable future) but bang for the buck decreases radically. Given how most GIS related software does not use multiple threads or cores particularly well, getting many cores is not critical. As of late 2012 I'd suggest an AMD FX8150 or Intel I7-3820 as suitable high end cpu's at sensible prices. AMD about 40% cheaper & 20% slower (if 8 seconds vs 10 is worth the $$ is your call :-)
+The high end Intel i7 cpu's are the fastest (as of 2012 & in the forseeable future) but bang for the buck decreases radically. Given how most GIS related software does not use multiple threads or cores particularly well, getting many cores is not critical. As of late 2012 I'd suggest an AMD FX8150 or Intel i7-3820 as suitable high end cpu's at sensible prices. AMD about 40% cheaper & 20% slower (if 8 seconds vs 10 is worth the $$ is your call :-)
 Note that using QGIS to view Postgis data will run both applications concurrently, so in this sort of usage, where multiple programs (processes) are being run concurrently, you can (if cpu is not the bottleneck) fully utilise multiple cores for your GIS work.

@@ Line 12: / Line 12: @@
 The more the better, but it does depend on what applications you plan to use. Things like R benefit more from RAM than some applications that are more graphic intensive.
-Use a 64 bit Operating System or anything over 3Gb om memory is largely wasted (unavailable for applications. (Simplistic but realistic). Getat least 8, preferably 16Gb memory.
+Use a 64 bit Operating System or anything over 3Gb om memory is largely wasted (unavailable for applications). (Simplistic but realistic). Get at least 8, preferably 16Gb memory.
 === Video Card ===