Getting Oracle Linux on Amazon EC2

Edit 08-Nov-2017:

As per Oracle's instructions - https://linux.oracle.com/switch/centos/ adding the GPG key is no longer neededthe bad news is - the procedure documented by Oracle throws a "Broken pipe" error, and the OS left in an unusable state (all Yum configuration disappears):



How to fix (basically just nohup the script).

Start with the CentOS 6 (x86_64) - with Updates HVM on the AWS Marketplace (provided by CentOS.org). After logging in (as the centos user) run the following commands as root:

1
2
curl -O https://linux.oracle.com/switch/centos2ol.sh
nohup sh centos2ol.sh &

The script does some package updating that cuts the SSH connection to the EC2 instance, and when the connection gets cut, the script dies in the middle with a "Broken pipe" and leaves the OS in an unusable state. By using nohup we avoid the broken pipe issue.

After the conversion script completes, run the following:

1
yum distro-sync

Step 4 will no longer be required, and you can proceed to install the Oracle software prerequisites.

Old Information Below:

If you feel the need to roll your own Oracle Linux install on Amazon EC2 (since Oracle no longer provides an officially-supported AMI, and you may not be too keen on using one of the community AMI's):

(1) launch an EC2 image using the CentOS 6.5 AMI from the Marketplace (which is free..) then log in as root (not ec2-user)

(2) import the Oracle GPG key
1
2
cd /etc/pki/rpm-gpg/
curl -O https://oss.oracle.com/ol6/RPM-GPG-KEY-oracle

(3) use the Oracle specific rules to convert the CentOS 6 to OL from this documentSpecifically, run the following commands as root:
1
2
curl -O https://linux.oracle.com/switch/centos2ol.sh
sh centos2ol.sh

(4) Synchronize the yum repository
1
yum -y upgrade

Once the command completes, the end-user should have a fully-patched Oracle Linux 6.7.

If you intend to install Oracle Database 11gR2, you can apply all the necessary packages and kernel parameters with this command:
1
yum install oracle-rdbms-server-11gR2-preinstall -y

and if you will install Oracle Database 12c R1, use the following:
1
yum install oracle-rdbms-server-12cR1-preinstall -y

Canon EOS 6D + 35mm f/1.4L vs Fuji XE-2 + 23mm f/2

Introduction

I've been a Canon EOS user for almost twenty years, except for a brief foray with the earlier Pentax DSLR's (K10D and K20D) about ten years ago. My favorite camera for the past few years has been the Canon 6D full-frame, with the spectacular 16-35mm f/4L IS ultrawide zoom.

The 6D and 16-35mm f/4L IS is a large chunk of optics and electronics, and while it's the perfect travel combo, I have always been on the lookout for smaller alternatives. I did have a Panasonic GF2 five years ago (with the 14mm f/2.8 prime) and that experience nearly destroyed my opinion of mirrorless cameras.

Recently I was able to obtain a Fujifilm XE-2 (which is a circa 2013 body, but had a significant firmware upgrade in 2016) and the 23mm f/2 prime lens (equivalent to 35mm on full-frame). After overcoming the sticker shock (the Fuji lenses are almost without exception priced similar to Canon L glass) I figured it would be interesting to compare the XE-2 and the 23mm prime with the Canon equivalent - 6D with the famous 35mm f/1.4L.

I used a tripod, base ISO (100 on the 6D, 200 on the XE-2) on a sunny day. This resulted in 1/4000 second shutter on the 6D. For some reason the XE-2 also wanted 1/4000 ISO at the same aperture levels, even with higher ISO. Here's the entire image (the central and corner areas are highlighted):

Center Performance - Canon

I compared the Canon 35mm f/1.4L at f/2 (one step down) and f/4 (three steps down) with the Fuji 23mm at f/2 (wide-open) and f/4 (two stops down). I also threw in the Canon 16-35mm f/4L IS at 35mm and f/4 (wide-open).

Canon 35mm f/1.4L at f/2 - not very sharp, low contrast, and this is one stop down!

Canon 35mm f/1.4L at f/4 - sharper

Canon 16-35mm f/4L IS at 35mm f/4 (wide-open)

 A few takeaways:
  • the 35L is not sharp! no wonder the Sigma ART has soundly thrashed it
  • even the 16-35mm zoom is equal to or sharper then the 35L (at the same aperture)
  • it is possible that I have a bad copy of the 35L, since I bought it used
  • it's also possible that the 6D is mis-focusing with the 35L
  • seems the 35L's only value is for its bokeh wide-open, since the 16-35mm f/4L IS is equally sharp at f/4 and more versatile

Center Performance - Fuji

Fuji 23mm f/2 at f/2 (wide-open)

Fuji 23mm f/2 at f/4 (two stops down)
Maybe my eyes are fooling me, or it's that magic Fuji X-Trans sensor, but the 23mm f/2 even wide-open beats the Canon 35mm f/1.4Lstopped down to f/2, and in the center, where all lenses perform their best. Also, while the Fuji sharpens up at f/4 compared to f/2, there's not much difference (unlike the Canon, where the improvement going from f/2 and f/4 is quite obvious). I would even say that the Fuji 23mm f/2 beats the Canon ultra-wide zoom, which is a very modern design.

What about the corners?

Corner Performance - Canon

Canon 35mm f/1.4 at f/2 (one stop down) - corner

Canon 35mm f/1.4L at f/4 (three stops down) - corner

Canon 16-35mm at 35mm f/4 (wide open)

The Canon 35mm f/1.4L continues to underwhelm. The 16-35mm zoom continues to amaze.

Corner Performance - Fuji

Fuji 23mm f/2 at f/2 (wide open)

Fuji 23mm f/2 at f/4 (two stops down)
Again, Fuji performance is at par or better than the much larger Canon 6D and L lenses. Here closer to the corners however, the 23mm f/2 lens is visibly sharper at f/4 than wide-open. Overall, the Fuji has lower contrast in JPEG's than a Canon.

Conclusion

The Fuji XE-2 and 23mm f/2 prime is optically equal or better than (my copy of) the Canon 35mm f/1.4L (version 1) and Canon 6D at all equivalent apertures. And check out the size comparison:


The XE-2, even if it is quite a old body, is worlds away from the Panasonic GF2 in terms of usability. It turns on in half a second, has an excellent EVF (that is very usable even in the dark), and competently auto-focuses. The main LCD can be turned off with a function button to mimic DSLR behavior. The XE-2 does not AF as fast as the 6D but for most common photographic situations, it will do. The 6D focuses much faster and more effectively when it gets really dark, but in bright light the XE 2 and Canon 6D with the 35L (a ring USM lens) are effectively neck to neck.

An important caveat: all of these tests were done with JPEG, not RAW. And I used autofocus for both bodies; it's possible that the 6D was mis-focusing with the 35L, and manual focus with Live View would fix that. But nobody would use manual focus and Live View with the 35L in real life.

Furthermore, the Canon has automatic lens correction, and this was quite obvious with the 16-35mm f/4L IS, where rectilinear distortion was minimized. I believe the XE-2 also has some form of lens correction built in.

However, I think this use case (autofocus, JPEG) is a very common one for travelers and other casual users.

I must rent the 10-24mm Fuji ultrawide! if its performance matches the Canon 16-35mm f/4L IS, then that lens with the XE-2 would make the perfect travel kit.

If there's any disappointment with the XE-2, it would be
  • it doesn't turn on instantly like any DSLR does
  • the 23mm f/2 lens is huge (larger than a Canon 50mm f/1.8) - this is not a Summicron



















Polar Alignment With A Bubble Level

In low-latitude locations such as Singapore (1.3521° N) or anywhere that Polaris is not visible, polar alignment can be difficult and time-consuming to achieve. One method for doing initial alignment in altitude is to use the mount's latitude scale, or a digital level of some sort.

This method is not sufficiently accurate because the mount's latitude scale normally only has 2° increments, and digital levels (in spite of their supposed high accuracy) actually have a tiny pendulum inside which is insensitive to small angle changes and really only has a resolution of 0.5° which is 30 arc-minutes and insufficient for a good polar alignment.

Here we can see an EQ mount set to zero latitude (as indicated by the bubble level) but the digital inclinometer is claiming 0.50° angle (which can be zero'ed out, but illustrates the inaccuracy of digital inclinometers).


However, we can take advantage of a bubble level that has a 45° vial (or our untrustworthy inclinometer) to achieve more accurate angle measures.

The Starrett bubble level in the photo above has an accuracy of 1mm in 1m, or about 0.045° or 2.7 arc-minutes, hence we can be reasonably sure (within 5.4 arc-minutes or  1/11 of a degree) that the mount is indeed at zero latitude.

We then take note of the position of the altitude adjustment knob:


The next step is to keep rotating the altitude knob (counting rotations as we go) until the mount is at 45° latitude (as confirmed by our bubble level or digital inclinometer):


Again we have an uncertainty of 5.4 arc-minutes in this measurement; adding to the uncertainty in the zero measurement, yields a total potential error of around 11 arc-minutes or 0.2°. On the other hand, if we used the digital inclinometer to zero the mount and also measure the 45° angle, our total uncertainty is 1°.

It is obvious that the bubble level provides lower error than a digital inclinometer.

In the case of the mount pictured, it took 39.60 - 39.75 turns of the knob to reach 45° +/- 0.1° latitude; this means that every turn of the knob yields 1.1295 - 1.1338° degrees (67.77 - 68.028 arc-minutes, or an average of 67.89 arc-minutes) of altitude.

We can see that the mount's latitude scale indicates roughly 45° as well, however since we don't know if the mount and tripod are level, and the latitude scale has no vernier and only 2° increments, the latitude scale alone is insufficient for setting the latitude.

If we had used the digital inclinometer, with its systematic error of 0.5° (total 1° including uncertainty about the zero point) then over 45° +/- 0.5° every turn of the knob yields 1.119 - 1.149° (67.14 - 68.94 arc-minutes or an average of 68.04 arc-minutes) of altitude.

Most EQ mounts use a sort of tangent arm assembly to adjust the altitude; as the angle gets higher, there is an increasing error. At small angles, θ ≈ sin(θ) but as θ gets larger, the error also increases. For example, at 45° (0.785 radians), sin(45°) = 0.707, a difference of 11%.

Hence, we need to reduce our calculated arc-minutes per turn by around 11% - so 68.04 arc-minutes per turn of the knob is reduced to 61 arc-minutes per turn.

The Astro-Physics Mach1 has a very accurate altitude adjustment, and is spec'ed for 62 arc-minutes per rotation with 16 ridges on the altitude knob (3.875 arc-minutes per ridge); this is in very close agreement with our calculated 61 arc-minutes per turn.

It is obvious that by using the 0° and 45° points as reference, we can significantly reduce the effects of systematic error in the bubble level or digital inclinometer.

After resetting the mount to zero latitude (again confirming with the bubble level), we can set it to 1.3521° by rotating the knob (1.3521° * 60 / 3.875 arc-minutes) = 21 knob ridges.  If we had used our derived value of 61 arc-minutes per rotation (3.8125 arc-minutes per turn), (1.3521° * 60 / 3.8125) = 21 knob ridges which is the same as the (known for the Mach1) setting.

On the other hand, if we need to set the latitude to a higher value, say 30°, we would use the 45° angle as the reference.  In the case of 30°, we would zero the mount at 45°, then lower it by 15° (which is 900 arc-minutes) which would require 14.5 turns of the knob (14 turns and an additional 8 knob ridges).

This method would allow accurate altitude alignment to within (5.4 arc-minutes error of the level) + (8 arc-minutes from the knob uncertainty) of about 13 arc-minutes (0.22°) worst-case; actual error may be half that. Such a result is good but not great, a drift alignment can achieve better accuracy. With this amount of altitude error, there is approximately 3.5 arc-seconds of drift per minute, hence an unguided exposure of 1 minute with a typical DSLR will still show round stars.

In contrast, had we set the mount's latitude directly from the digital inclinometer, we could have been in error by 1° which would lead to a 15 arc-second drift in 1 minute, thus limiting maximum exposure times to around 20 seconds before star elongation would be visible.