As I have settled in to my new desktop of choice (Last two posts, here, and here), I have been working to figure out how to put back together my disrupted desktop existence. Evolution with EWS seems to work well most of the time, but there are occasional hangs requiring Evolution to be restarted. Better than the past, but still not as stable as I would like. Calendar invites seem to be the thing that cause hangs the most. If I just wait, they often time out and clear up, but sometimes I just xkill it and restart it rather than waiting.
Fedora 17 seems so far away right now, with its newer version of Evolution. Half the bugs I have reported are fixed upstream there. I'll be Alpha and Beta testing that for sure.
IM Not Sure What is Happening
Pidgin / SIPE started working with MS Communicator again, like it did under Mint. When I was trying to figure out what was wrong I decided I needed to run a Wireshark trace and compare the working version of Pidgin (1.11/Mint) with the non-working version (1.12/Fedora). I set up Mint 12 in a Virtual Machine, and configured Pidgin / SIPE. That came right up and all my 'buddies' appeared. They like to think of themselves as 'co-workers'.
I went back to Fedora 16 and fired up Pidgin / Sipe there, all ready to start the trace... and everything worked.
I do not understand everything I know about that, other than it means I can now IM from Linux, and that can be handy when talking to my MS Windows using co-workers. Buddies. Whatever.
Side note on Mint 12: I looked at Cinnamon while in the Mint VM, and while it really helps make Gnome more usable again, there was not enough there yet to make me leave XFCE. Evolution-EWS was not on Mint 12 at all, so Mint, as popular as it now (I just looked at Distrowatch, and it is number one by a wide margin: 3564 to 2083 for number two Ubuntu, Fedora at 1721) stays a VM for the foreseeable future. Reading the various trades about Cinnamon makes it clear though: the disruptive change that is Gnome-shell (under gnome 3) or Ubuntu's Unity is facing a disruption of its own. There is a Fedora version of Cinnamon, and that sounds like something I may want to watch.
By and large, MS Windows is something I hardly need anymore. I was working with another disruptive technology and came across something where I had to use MS Windows. No choice about it. I was working with... Re Hat Enterprise Virtualization. The new one. 3.0. I thought about that as I stared dumbfounded at the message saying I had to use IE and MS windows to access the RHEV Manager console (Think the V/Sphere console if you are VMware literate).
I knew that the older version of Red Hat Enterprise Virtualization had this requirement. It was the technology from the Qumranet acquisition. The console uses .NET bits. I know RH is moving to port all that to technology that will work from any browser under any OS rather than dictating such a specific OS and browser need. I had hoped that some of it would be in place with 3.0, but clearly not.
Andrew Cathrow is doing a RHEV Roadmap at the upcoming Red Hat Summit in Boston, so hopefully there will be some good news about when this restriction is removed there. Here is the RHEV lifeccyle diagram, and I see RHEV 3.1 and 3.2 there. There is something of a cognitive dissonance having to configure my RHEV environment from Windows 7. Not to mention the fun and games it took to get IE 9 to accept the certificates from the RHEV-M console.
Other than Windows
Other than the oddity around doing all this from Windows, I have to say I really like RHEV. As someone that has been doing virtualization for thirty years, RHEV's design presses all the right technical design buttons for me.
First and foremost, there is KSM. I like everything about the idea of Kernel Samepage Merging. It just makes too much sense. The secret sauce of virtualization is how well memory can be over committed, or even if it can be. In my internal monitoring of our virtual R&D environments, memory is the bottleneck 80% of the time.
Next is the fact that I can set up both appliance like servers, using the RHEV-H, and standard RHEL servers with the RHEV packages installed, if I want the entire Red Hat Linux software stack available for other things. Either plugs in to the management console, RHEV-M. This makes setting up a RHEV environment easy and / or flexible. The RHEV-H appliance model is a boot, configure, and go thing, and I was able to set those up quickly. It took me longer to figure out how to get a remote console than to install it. the first RHEV-under-RHEL took a little longer, but the second one was fast once I knew what to do.
RHEV-M will look familiar to anyone who has used any sort of management interface for a virtual environment. Leaving aside the Win7 / IE part, it has everything required to configure networks, storage, and has a very granular permission model to allow special userids to be set up in large shops that can only do what you want them to do.
Red Hat documentation is amazingly good as well. Very detailed. Fully indexed by Google so answers to questions are easy to find.
Important to performance in the Hypervisor world is how fast a particular Hypervisor adds support for the latest hardware features. AMD had virtualization assists in the microcode for years, and Intel added parity there with Nehalem and now Sandy Bridge. We have seen huge drops in overhead with those more recent chipsets in our virtual infrastructure. None of that does you any good if the hypervisor does not support the assist though, and RHEV has a strong history of early adoption of those kinds of features.
In the discussion about multi-hypervisor clouds, I am firmly on the side of more is better. In the world of Hypervisors, this is designed as/looks like/runs like one of the best.
The Math of Disruption
One of the funny things I often come across is math around disruptive technologies. As it pertains to Linux, one example I often see is around tablet computers. Full disclosure: I have a Motorola Xoom. I had an iPad, which I gave away to get a Xoom.
When the iPad came out, it owned the tablet marketplace. I saw numbers like 93% (and I wondered what the other 7% was, since the category of tablet computers seemed to have sort of died there for a while)
Later articles talked about the iPad was trouncing the market with 80+% of the market. Android tablets were failing. Later articles noted that Android was still an also ran with iPad having over 70% of the market. Later I read how Android tablets were in trouble because they were so fragmented, and now that the Kindle Fire was out, the iPad still had something around 59% of the market (This article has Android at 39% too).
Humm. Is there a trend there? Apple keeps selling record numbers of the iPad. the unit volume is growing, yet the percentage is trending down. I predict that there is a magic number in there: Probably around 50%. Once Linux based tablets have more percentage market share than iPads, we'll start seeing the articles about how Apple has failed in its quest to dominate, and what Apple did wrong and stuff like that. Even though they will still be selling record unit volumes, and tablet computers will be *everywhere* and the real story (for me) will be just how disruptive all this was, and how big a piece of it Linux has.