Being a VMware guy, I needed a VMware cluster to train and learn on. The old cluster of a DL385 and a NL40 wasn’t really cutting it and was LOUD. Therefore, I decided it was time to upgrade. I was interested to play with the concept of nested VM’s and build a virtual cluster inside my HP ML310e. So in essence, what’s it like? It’s sexy, very quiet and power efficient (as denoted by the e, whereas p means performance, in terms of more powerful CPUs, etc.) The model I have is the base model, Xeon E3-1220-V2 with a single SATA 7,200 RPM disk and maxed out to 32 GB RAM. This is fine as it is just a test lab. It would also make an excellent small office server as well if an additional disk where added. The standard 2 GB of RAM is a bit miserable but you do have another 3 slots for your RAM unless like me you’re a VMware nerd who crams in as much RAM as possible. The hardware is ideal for a nested cluster. The Xeon E3-1220-V2 has both VT-D and EPT (Hardware look aside buffer in other words) onboard. This means it has all it needs to deal with nested VM’s in hardware rather than software, so the performance is still excellent. So, let’s start with the cosmetics and the outsides. The G8 has a face lift and sports a cool metallic grill. One of the things I didn’t like about it was the fact that to actually get into the box you now have to use a key on the side of the unit. That in itself is not so bad, but you once you have the lock in the unlock position you can’t remove the key,...

I’ve got a Western Digital Sentinel DX4000 in the lab to play with.  There were a few minor hiccups during install, but overall I’m quite impressed with the unit.  What captures the imagination, however, isn’t the unit as shipped, but the unit as it can be configured. When you set up the system you are asked to interact with it through a fairly primitive installable client.  Most users will never even know they can RDP into the thing.  What’s amusing is that the installable client is in fact an RDP client with no “chrome.”  The Dashboard is nothing more than an application running on the NAS’ native Windows Storage Server 2008 R2 Essentials install that is presented to the end user over RDP.    The presentation is fairly seamless, it’s easy to miss the Remote Desktop Connection (RDC) client flashing up for a fraction of a second as the dashboard launches.  But it’s RDP nonetheless, and you can simply fire up the traditional RDC client and log in with the administrator username and the password configured during initial setup. Voila: one fully functional Windows Server 2008 R2 install.  Now for a number of reasons Microsoft does not under any circumstances want you in here mucking about.  The most obvious being that you could theoretically install applications or deploy services that it normally reserves for people who buy much more expensive SKUs.  They even give you a cute little warning to that effect. What’s behind port 3389?   I naturally disregarded this entirely and began mucking about.  A few things leap out at me.  The first is that Western Digital has done some customizing.  There’s a Haali media splitter in here and codecs from Xiph.org.  I approve.    They’ve also added some teaming software to the...

One of the main selling features of the Western Digital Sentinel DX4000 is that it is supposed to be easy to set up and install. But how easy is “easy”? Is it really “plug-and-play”? Could your proverbial Aunt Millie do it? And, most importantly, how many different ways can we break it? More than a few, as it turns out. But don’t hold that against Western Digital; for that you blame this guy. You are unlikely to encounter many hiccups during installation unless you are as obsessively interested in voiding warranties as he is. If it is not done perfectly by-the-book, the setup of the Sentinel DX4000 is not as fast or smooth as advertised. If you do everything according to plan it is certainly within acceptable parameters. Let’s take a look at my experience with the device. Out of the box The DX4000 ships with an ethernet cable, two AC adaptors, and a Quick Install Guide, which is the dead tree version of the pdf found here. The adaptors caused some momentary head scratching. Redundant power supply? Sounds good! The Quick Install Guide says to only plug in one of them. Then a little later, in a contradictory paper insert, it depicts plugging in both. (Aunt Millie furrows her brow in confusion.) I plugged in both. Nothing broke. The ethernet cable is a perfectly nice ethernet cable. It’s also kind of short. Instead of using it, I plugged in a longer cable that was handily lying around the office and which reached to the router from my workstation. Cue ominous music. Setup After you power on the Sentinel DX4000 for the first time, it will run its OS install – which takes a few minutes – and then display an IP address. You...

The design concepts underpinning the Supermicro FatTwin are something I would hold up as “game changing.”  The reasons why are explored in a review I wrote for The Register a while back, but the product is so good it’s worth expanding that review a little.  Free from any real editorial constraints, I think it’s time to take a bit of a wander through what this equipment can actually mean to a systems administrator. So what’s a FatTwin anyways?  A FatTwin is a brand name used by SuperMicro to describe their latest in dense-but-powerful server offerings.  Two dual-processor servers per U, 4 U per chassis for a total of 8 2P servers in a 4U enclosure.  There is no shared backplane for networking or other system components.  The only thing shared by the chassis is power.  This drives the cost of the chassis down so that if you are paranoid and want a spare chassis on the shelf you can probably afford it without breaking the bank.    Isn’t a shared power plane a single point of failure? Yep; a shared power plane is a single point of failure.  If you somehow manage to tank the shared power plane in the chassis you will not be able to light up half of the nodes.  One failure and half of the systems go.  The shared power plane is an intermediate step between single-everything as would be found in entirely discrete systems, and shared-many-systems as is found in blade servers.  A typical blade backplane has hundreds of individual traces; a myriad of single points of failure that could go wrong.  When compared to blades, the choice of “shared power only” offers simplicity; it has the bare minimum of traces and just carries power.  The result is that the power backplane on...