My father is a retired teacher but loves to work with his hands. I can remember very early on in my up bringing, him teaching me that it is good to measure twice and cut once. Whether it was building a deck or just a bird house the point was it took more time to cut something wrong and then has to re-cut the board shorter or even wastes the old board and cut a whole new one.
When I was preparing for this article I remember having to learn that lesson the hard way and how much effort really is put into that second cut. The problem in the storage industry is the misaligned partitions from a move of a 512 byte sector to a new 4096 byte sector. This has to be one of the bigger performance issues with virtualized systems and new storage.
Disk drives in the past had a limit on the number of sectors to 512 bytes. This was ok when you had a 315 MB drive because the number of 512 byte blocks was not nearly as large as what is in a 3 TB drive of today’s’ systems. Newer versions of Windows and Linux will transfer the 4096 data block that match the native hard disk drive sector size. But during migrations even new systems can have an issue.
There is also something called 512 byte sector emulation which is where a 4k sector on the hard disk is remapped to 8 512 byte sectors. Each read and write would be done in eight 512 byte sectors.
When the older OS is created or migrated, it may or may not align the first block in the eight block group with the beginning of the 4k sector. This causes misalignment of a one block segment. As the reads and writes are laid down on the disks the misalignment of the logical sectors from the physical sectors mean the 8 512 byte blocks now occupy 2 4k sectors.
This now forces the disk to perform an additional read and/or write to two physical 4k sectors. It has been documented that sector misalignment can cause a reduction in write performance of at least 30% for a 7200 RPM hard drive.
This issue is only magnified when adding other file systems on top of this misalignment. When using a hyper visor like VMWare or Hyper-V, the virtual image can be misaligned and cause even further performance degradation.
There are hundreds of articles and blogs written on how to check for you disk alignment. A simple Google search of the words “disk sector alignment” and you will find this has been a very popular topic. Different applications will have different ways of checking and possibly realigning the sectors.
One application that can help you identify and fix these is a tool called the Pargon Alignment tool. This tool is easy to use and will automatically determine if a drive’s partitions are misaligned. If there is misalignment the utility then properly realigns the existing partitions including boot partitions to the 4k sector boundaries.
I came across this tool when looking for something to help N series customers who have misalignment issues in virtual systems. One of the biggest things I saw as an advantage was this tool can align partitions while the OS is running and does not require the snapshots to be removed. It also can align multiple VMDKs within a single virtual machine.
For more information on this tool and alignment check out the Paragon Software Group website.
In the end, your alignment will effect how much disk space you have, how much you can dedupe and the overall performance of your storage system. It pays to check this before you start having issues and if you are already seeing problems I hope this can help.
IBM published a paper this week describing how the scale out NAS product, SONAS, works with a software package in the seismic processing space called Paradigm Epos4. The report goes into detail of both the hardware and software issues surrounding the massive amounts of data associated with finding deposits of fossil fuels in the strata.
The software supports NFS mounts which seems to be the sweet spot of the linux based SONAS system. One of the biggest problems with the oil and gas industry is the tremendous and rich amount of data.
The cost of drilling varies depending on the depth of the well, remoteness of the location and extra services required to get the oil or gas up to the surface. With some of the deepwater rigs the rates for 2010 was around $420,000 per day and could be more on higher performance rigs.
With so much on the line, it is very important to get information accurate and quickly so that companies can avoid costly mistakes. IBM has been working in the oil and gas industry for over 50 years. We have experts not only in the hardware, software and services but we understand the industry and how “big-data” is changing that industry faster than others.
SONAS allows for companies to have a large scale NAS solution that can have a single files system for multiple peta-bytes of data. SONAS also allows data to move from faster pools to other virtualized systems down to a tape archive. This increases the ROI by having the most recent accessed data on the faster drives and customers can expand their buying cycles further because they are not spinning old data.
The other variable in this industry is companies need to scale projects up quickly and not always with a 1:1 ratio of performance to storage space. SONAS is able to scale both of these variables independently of one another. As new systems are brought online, disks can be added and rebalanced non disruptively. The same can be done with the interface nodes.
More information about the testing can be found in the report here.