Henry Newman's Storage Blog Archives for January 2012

Intel Buys QLogic Infiniband, What Now?

I am sure most of you have seen the news that Intel intends to purchase QLogic's InfiniBand (IB) business. For those of you that do not know, IB has been around for over 10 years and is an interconnect for communications and storage used in HPC and a few other environments, such as Oracle RAC. One of the major advantages of IB over Ethernet is low latency. The other big advantage is bandwidth.

While Ethernet is getting just 40 Gb/sec out, IB is already at 56 Gbit/sec. So why would Intel buy QLogic's business, given that Mellanox will be the only IB vendor left? There are a few potential reasons, such as Intel wants access to the router technology as it has a plan for its own interconnect, or maybe Intel wants to prop up the market with this technology.

Whatever the reason, my opinion is this spells the beginning of the end of IB given Intel's track record. The latest example: Intel touted that it would enter the SSD market about three or four years ago. The X-25 line of SSDs was a flop, and do not get me started on all of the issues. Intel is very, very good at its core business, but time and time again throughout Intel's history, products outside of its core business have not been well conceived. If Intel follows its historical course, that will leave only one IB vendor, and that means IB will be dead as an interconnect.

Whether you view this as a good thing or bad thing depends on your point of view. IB definitely helped push some technology issues to reduce latency in Ethernet. Some could argue it pushed Ethernet to improve performance. Whether this would have happened anyway is up for debate, but I think it likely would have. It will be interesting to see how this plays out, but I think what will happen is most things will not change.

Labels: Intel, QLogic, InfiniBand

posted by: Henry Newman

ReFS: Too Little, Too Late, Too Few

In my last blog post, I referenced Microsoft's new file system, ReFS. Yes, the world is getting another file system. Is this a good thing? Does it break any new ground?

The answer is, of course, no way. The other interesting fact is that Microsoft is releasing ReFS in the same way it released NTFS back in 1993 -- to commercial customers first. If the same path is followed, those of us that use NTFS for our home systems will have to wait until 2019 to change file systems.

Surely, things need to go way way faster than that. NTFS became available in 1993 and home users were able to get it in 2000 with the release of Windows XP. By 2019 or 2020, disk drives will likely be using heat-assisted magnetic recording (HAMR), and we will have very very large densities. HAMR is expected around 2016 or 2017. Let's say I have a 12 TB hard drive and I am running NTFS, then I have a crash and I need to run a chkdsk. Are we talking about two days or three days to complete the full run?

This is not a funny joke. As I said in the last entry, ReFS is far too little far to late, but from a feature list perspective, it is far better than NTFS, which I often call the worst file system in the world currently being used. Home users need to demand that Microsoft move up the timeline for ReFS for lower-end systems. I solved the problem at my house: I do not use a Microsoft product for my data storage. Instead, I use XFS Linux on a NAS device. However, most people outside our industry do not have the skills to do this.

Labels: Microsoft, Windows 8, file system, ntfs, ReFS

posted by: Henry Newman

Microsoft Might Get a Modern Stack

Someone just pointed me to this blog post about Windows 8

It looks like Windows 8, at least the server version, is going to get a modern storage stack file, volume management, a file system and some resiliency features. I see the target of what they are doing as the lower end of the small business market with people that today have under 10 TB of data.

This is a big change for Microsoft. As someone who has railed about the poor performance, poor metadata management, poor resiliency and so on of NTFS, this is a small step in the right direction. It, of course, does not solve the home user problems as I quote, "first as a storage system for Windows Server, then as storage for clients, and then ultimately as a boot volume. This is the same approach we have used with new file systems in the past."

Over time, could this change the market for the lower end NAS products as why do have an extra management framework? NTFS should have been replaced at least 5 years ago with something that has some modern features and functions. So it looks like we have another 3 to 4 years from now to have a new file system that runs across the Windows product lines from the home user to the server. Let's see, by that time we should have disk drives supporting bit-patterned media with significantly greater density, SAS channels running at 12 Gbits/sec and PCIe 3.0 running in home PCs.

Microsoft has delivered way below the mark of what is needed -- at least what they have announced. We get Windows 8 in 2012. We got NTFS in 1993, and it lasted for almost 20 years. ReFS is not going to work well 20 years from now.

Labels: Microsoft, Windows 8, file system, ntfs, ReFS

posted by: Henry Newman

Will Storage Software Be the Key Differentiator?

I have been looking at hardware platforms and the value of hardware compared to software. The storage hardware stack and functionality is pretty similar these days from the low end to the high end. Yes, there is differentiation with performance, reliability and of course price, but for the most part the functionality at the user level is about the same.

The big difference to the end user is coming down to software to manage the storage -- not the hardware. Examples of this include software that decides where to put your data or software that moves the data to where it is needed in the environment, whether that be in the United States, Japan or the United Kingdom. These are just two examples; there are many more.

Hardware in my mind has become even more of a commodity. The value to the end user and higher profit margins is going to be for companies involved in software. The winners on the software side are those vendors able to integrate the software into environments to meet the requirements of various markets. Like it or not, standards are not going to be around to allow things to seamlessly move between platforms to implement seemingly complex new data policies.

Today, in most cases this is accomplished by applications running on top of file systems. This is the new frontier to implement these policies, but that might not be the case as we move forward.

So what does all of this have to do with hardware? The industry has been talking about hardware as a commodity for years, but the hardware vendors continue to add a feature here and there that differentiates their products. I think the end of the line is near, and software features are where the market will innovate at least for the next few years.

Labels: storage hardware, storage software

posted by: Henry Newman

Disk Drive Pricing Winners and Losers

In my last blog post I looked at the huge price increases in disk drives. Now, I am thinking about the winners and losers because of the dramatic price rise. The obvious losers are end users who need a drive that is easy, but storage companies large and small are going to lose also, but to different degrees.

Larger companies are going to be under extreme pressure to hold their prices for customers, given the prices they got in 2011 and 2010. Some of these large companies might have gotten forward pricing such that they negotiate a price for drives for x numbers of months in advance. I do not know either way, but it is possible. On the other hand, for the most part, for years disk pricing dropped month by month and year by year, so maybe not. Smaller storage companies buy when they need to and do not have large inventories. If you have a limited supply of drives and you are a large disk manufacturer, who are you going to service first? My guess is you will service the companies you have committed large volumes to monthly.

Again, I do not know, but I am speculating that large storage companies have a monthly minimum order. They add to this order based on additional requirements. That way they always have inventory. My guess on the second commitment would be companies in the first category that then add to their order. The last group would be companies that order irregularly. This is what happens in other industry when there is a shortage. You name the product or commodity, from oil and gas to grain to cars. It is how the market works and I do not think disk drives will be any different.

Labels: SSD, disk drives, commoditization, pricing

posted by: Henry Newman

Disk Drive Prices Soar to New Heights

Back in October 2010 I purchased four Western Digital 320 GB Scorpio Black Drives from Amazon for my home NAS at $59.99 per drive. On Dec. 29, I decided to look at the price, and it was $99.28 per drive. I quickly checked the Western Digital site and found that this drive is still a current product, and therefore the price is not high just because it is no longer made. This is a 65 percent increase in price. YIKES!

We all know about the problems in Thailand, but it seems to me that over a 14 month period the price increase seemed out of line. Thinking Amazon might be out of touch with reality, I checked Froogle and found some even higher prices and sites with no availability. Clearly it is not. I am not sure how long this trend is going to last, but it will surely have an impact on the cost of storage throughout the market, from servers to laptops to anything using a disk drive.

Given the pricing, there must be shortages of drives, as I also looked around at local computer shops and saw much higher prices than I expected.

Labels: disk drives, DAS

posted by: Henry Newman

SSD Consolidation Is Coming

About a month ago I wrote my yearly set of storage industry predictions, and one of them was continued SSD market consolidation.

Recently, I saw an interesting article in The Register about the oldest SSD company that I am aware of being for sale. Yes, Texas Memory is for sale.

The SSD market must consolidate, given the number of players and the cost of marketing and sales. Although total sales of SSDs are going up, most of the products being sold are the commodity versions -- - not the enterprise versions. The development cost of enterprise SSDs is much higher than their commodity counterparts.

Labels: SSD, Texas Memory Systems, Storage Consolidation, Storage

posted by: Henry Newman