The company I work for does some neat things with backup and failover using the Cloud. But… I am not here to sell or promote that. That just goes against our little community here that Tim has built.
We use two main Cloud vendors – Amazon and Terremark with a 3rd – RackSpace. First off, all 3 are work fine. You can send and retrieve data. You can erect a Windows server within an hour.
One of the earliest things I’ve learned as I jumped into system management was the importance of standards; not only standards for equipment but also standards surrounding naming conventions. This blog will cover the process around development and implementation of a naming convention standard.
As we begin, I am not going to tell you what kind of naming conventions to use, just a guideline to think about when you are developing them. I have been in many large organizations where there have been many heated debates about xyz names and so on. It’s extremely important that once a naming convention has been selected, that everyone adheres to it as if it were gospel. This will end up being the starting point for your infrastructure organization to ensure equipment can be quickly identified for support, financial, and audit abilities.
In this section we will discuss the need for proper labeling of cables, servers, and storage hardware. I have been in many data-centers over time, and have two extremes of this process. From no labeling to everything being labeled, what I will try to assist you with in this section is the ability to find a happy medium between the two.
Cable Labeling - The importance here is to ensure that each end of the cable is labeled, as well as any junction box interconnects that may exist for the same portion of glass or copper. This will ensure that when there is a physical plant problem of some kind, it will make it extremely easy for you to locate and run down the issue. Another important aspect of cable labeling will ensure that when you are moving to newer switched electronics, all your cables will be extremely easy to locate. The same process will hold true when it comes to labeling your power cords as well, source and destination for the device and the PDU’s. This will ensure you have the same physical tracing abilities as you would have for networking cables. There are many misconceptions that you need to spend a slew of money on an expensive cable labeler system (1k and more), however, all you need is a laser printer, and some printable sheets. These printable labels will work on both copper and glass. I have attached a link below, which shows many of the options available, and are all reasonably priced.
You will all have already seen IBM's announcement last week on it's Long Term File System (LTFS). In short, it's a software feature that enables you to mount a sequential media (only LTO gen 5 or newer) as if it were a local file system. In itself this is not a real innovation. Seagate had this for years, although it never kicked off. In my opinion there has never been a real use case for this type op I/O.
So a small update this time...
I've been testing GPFS as far as installing it on different OS:es. Trying to get a feel of if we can use it for something(/most things?). And also if we can handle it ourselfs. Installed it on AIX (P6) - no problem. Linux - no problem. Windows server 2008 R2 (not supported, we later found out) - problems. Windows server 2003 R2 (got problems with installation .msi not beeing accepted). Tried Windows server 2008 SP2 (and got the same issues as R2). The issue we're having is about trying to create a root user as an local administrator. By changing ownership of the whole SUA recursively we got as far as nodes communicating. But the windows node was never able to mount the gpfs. Next thing I will try before handing the installation over to IBM is renaming the Administrator to root. Since the Administrator user didn't have any problems with permissions from the beginning. The documents clearly states that we should create a Domain Administrator root user in our AD. But we don't want that, and the windows guys can't figure out what difference that would make. I guess we are doing something wrong. Will try the renaming thing tomorrow though.
One thing that always irritated me, as a end-user is none of the vendors every really had any real answers around cable and label management, at least not something you didn’t have to pay for a professional services engagement on. It was something I had to develop primarily on my own based on what worked for my company and the data-centers I was working in. Someone once told me “Anything worth doing is worth doing well”, this holds just as true for the cabling aspect of any data-center.
Over the last few years, I have worked in many data-centers of all shape and sizes, and the one thing I’ve noticed is that people tend to focus on the state of the equipment servers and storage more than the physical layers of the facility itself. Which, in my mind, is even more important than any of those servers or storage. For example, if you build a house and use cheap cement, but buy really expensive furniture, and the house falls in, then there was probably a huge error in your project plan. I realize this may be a tad overstated, however, I just want to convey to you the need to have an efficient, well managed data center.
I am here to declare that I am first to coin the term "JBOGs" - Just a Bunch Of Gigabytes for Cloud-based Storage. This is the same vain as "JBOD" - Just a Bunch of Disks which is a known term to us storage folk.
Now, Just a Bunch of Disks implies a few things - no intelligence, several to dozens of hard drives huddled in the same enclosure, an interface to something with smarts (storage controller or server raid controller or SAS/iSCSI connector). When you say JBOD, you are thinking 12 to 15 drives in an enclosure.
The woes of a storage monkey trying to decide which vendor technology solution is the most appropriate seem to go on for ever.
EMC, NetApp, Equalogic, HP Eva, HP Lefthand, Pillar, and so on and so on............(trying to stay awake here.....)
As an HDS customer I have had enough of the customer facing HDS Support Portal. What about you?
It's broken - slow, unreliable and very hard to find anything.
