July 2007 | Main | October 2007
August 23, 2007
The Burton Group on Fragmentation and Virtualization
The Burton Group is one of IT's leading analyst firms. They research IT architecture, infrastructure, governance, new technologies and other IT-relevant strategies and create detailed advisory reports.
In a recent free report, the Burton Group covered the popular topic of Server Virtualization. In a particular section they covered disk performance and correctly granted that subject significant importance.
Here are some quotes from the paper:
"Over time, the fragmentation will continually degrade performance of the virtual disk."
"File writes and deletions will fragment a virtual hard disk, like any physical hard disk, over time. Running scheduled disk defragmentation operations at periodic intervals or using a disk defragmenter application that defragments disks in real time is the best approach to prevent fragmentation-related disk performance degradation."
Of course Diskeeper Corp strongly recommends the real-time approach as the most effective and modern method for solving fragmentation.
You can sign up to access Burton Group's free papers including the one mentioned above here.
Posted by Michael at 07:26 PM | Comments (2)
August 22, 2007
Are you interested in testing software?
We're looking for a few good software testers. At this time, we are specifically looking for IT professionals who manage 20 or more computers, and can test Diskeeper Corporation software in production environments (or lab replications of production environments).
If you meet this criteria please email our Quality Control group at qtesting@diskeeper.com and let them know you are interested.
We may expand our test group to include home users next month. If we do, I'll provide a new specific contact email address here in the blog.
Posted by Michael at 12:27 AM | Comments (0)
August 21, 2007
Diskeeper HomeServer - our plans
We've been receiving strong interest in Diskeeper for the new Windows Home Server (WHS) platform. One of the questions we are regularly asked is "how much will Diskeeper cost?"
Given that our current Server edition (which already works on WHS) goes for $299, that's a fair question. We certainly recognize that that price is high for a consumer, and you will find that when we release our official version, the price will be far more attractive.
The current plans are that we will have Diskeeper HomeServer available for small and mid-size OEMs in the next week or two. It is already available for the major OEMs signed up with Microsoft to deliver Home Server boxes.
When WHS hits the streets, Diskeeper HomeServer will also be available for users to buy. Expect to see it available on the Diskeeper online store, Windows Marketplace and other online resellers.
The current WHS release date (as much as Microsoft will divulge) indicates that you should be able to get your hands on one in early October. Because WHS is a full package of hardware and software, the final release date is contingent on the readiness of the major hardware vendors producing WHS computers.
Several of our developers and I have been experimenting with WHS and we love it. If you have 3 or more PCs in your home, this product is definitely worth checking out!
Posted by Michael at 12:04 AM | Comments (2)
August 14, 2007
CNET upgrades to Diskeeper 2007
CNET Labs is one of the leading independent reviewers of technology. As is important in any scientific test, CNET needed to eliminate variables that could skew comparisons of computer systems, and addressing fragmentation was a key consideration for them. They know that a fragmented desktop or laptop will perform worse than a defragmented one, potentially invalidating a test.
You can view the process methodology for desktops and laptops.
Earlier this year I had a long conversation with the Director of their test labs, and while I can't disclose the nature of their evaluation of various defragmentation programs, their product needs are similar to what your defrag software requirements would be.
CNET Labs has been a member of the Diskeeper family for years, and we are certainly pleased to support their upgrade to Diskeeper 2007!
Posted by Michael at 09:30 PM | Comments (0)
August 03, 2007
Why Cornerstone Bank Choose Diskeeper To Maintain Their Computer Systems
The Let's Talk Computers radio show interviewed Jason Amick, Network Administrator for Cornerstone Bank of York, Nebraska. The conversation starts out with a description of their IT infrastructure and explores why and how they implemented Diskeeper. It concludes with details on the benefits Cornerstone has realized.
You can listen to it here.
The transcript can be read here.
Posted by Michael at 09:05 PM | Comments (0)
August 01, 2007
SATA, NCQ and Disk Fragmentation
A DiskeeperBlog reader recently had this question, and I felt it deserved it's own thread:
"Does NCQ (Native Command Queuing) on the newer SATA drives help in reducing fragmentation, or the impact of fragmentation on HDD performance? I have run across this subject recently on a few sites, and I'm thoroughly confused about what NCQ does to fragmentation." - Durian
---
A very good question and, I agree, a very complicated one. As it turns out, I'm in the midst of a very long white page paper on various data storage technologies and how they relate to fragmentation. In it is a semi-completed "chapter" on SATA, SCSI, FC (Fibre Channel), and SAS (Serial Attached SCSI) disks. I've taken a small excerpt from that section to address your immediate question. I apologize in advance if it reads poorly as it still needs some polish.
I recommend chapter 1 of the "Shortcut Guide to Managing Disk Fragmentation" by Microsoft Product Manager Mike Danseglio, for technical references. Wikipedia can help define some of the technologies described if you need more info.
As seek optimization is interelated with the technology of queuing , I'll describe both and explain the relationship.
An important concept to keep in mind when reading this is that only the file system understands the 1's and 0's as separate files (e.g. Report.Doc). Physical disks operate as block-level devices, unaware of whether an incoming set of data it is asked to read or write is originating from one single file or from a number of different files.
Disk/File defragmentation allows the operating system to only have to pass along a single, or small number, of requests for data to the disk subsystem, rather than a great many.
Seek Optimization :
There are a number of disk-level algorithms to minimize the impact of physical limitations such as rotational latency (waiting for the disk to spin back around). They include variants of elevator-seeking and shortest-seek-first. These algorithms leverage the disk buffer, prioritizing retrieval of data physically closest, by measure of the data's cylindrical location and/or how close the requested data is to the current location of the disk head. These technologies have already been implemented in hard drives by drive manufacturers for many years now. Seek optimization does not eliminate the need to defragment. While it improves on rote data retrieval methods of the past, it cannot and does not account for file fragmentation because it is a "block" based technology. Seek Optimization organizes and prioritizes data retrieval based on physical location of data blocks on the disk platter. It does not operate on a "file object" level. On that note, defragmenters solve "file" fragmentation, and do not get involved in block-level allocation.
NCQ:
Native Command Queuing (NCQ) is a technology that allows a SATA drive to intelligently re-prioritize and queue disk requests while completing others (up to 32). You could say that it is like dynamic multi-tasking at the disk level. SCSI disks have long supported command queuing (TCQ - up to 256) before SATA and NCQ were created. While queuing will improve prioritization strategies and overall seek time for small asynchronous I/O (random/non-linear - in other words, not a large MPEG file), it does not address "file" fragmentation. Command queuing is also "block" based, essentially providing buffer space for seek strategies to be more effective.
In summary, for all the I/O the disk is requested to process at a given time, it cannot optimize its activity for a particular file because it does not know what of that requested I/O is from one file or another. The disk will simply retrieve the blocks of data in the fastest way that they can be. As a example, you may get half of two different files open before you get all of one, and then the other, open.
To answer the original question, NCQ does not solve fragmentation. While the performance benefit that NCQ does provide is irrespective of file fragmentation, they are complimentary.
Posted by Michael at 01:28 AM | Comments (0)