TerraScale were an innovative bunch that developed some interesting technologies around the xfs file system, and made it scale in a cluster. Some time ago, Rackable bought them.
Now it appears that Rackable is pulling back from this market, and is putting TerraScale … er … RapidScale on the auction block.
Ok, its not quite like this … they have engaged a financial advisor to “seek strategic alternatives” for the group. This is usually corporate-speak for “we don’t want to be in this business any more, and will sell the IP, the group, and the brand to someone … if we don’t get what we are looking for, we will just shutter it.”
John West and the (as usual) excellent InsideHPC blog point to the PR from Rackable.
I think the salient point was that this was discussed during a Q2 earnings call. That is, the financial folks want to know what Rackable was going to do to trim losses/increase share price. What you do in these cases, when losses mount, or earnings slip is trim expenses, and start looking at cutting non-profitable lines.
This actually points to a likely problem with acceptance of this particular type of clustered storage system. The problem is something of a self-reinforcing problem.
John pointed to an analysis at Blocks and Files which talked about the situation. The first sentence points to the issue.
Not that most people knew it was in the clustered storage business but Rackable …
This is the problem.
If people don’t know about your solutions, you aren’t going to sell much. Rackable bought TerraScale when the strategic growth directions and synergy made sense. Unfortunately, it seems not to have gotten traction with the product, could not attract enough customers to offset costs. So while it was investing in development, its marketing efforts do not seem to have paid off.
The self-reinforcing problem is, in short, if a small company develops/builds something of quality, it almost doesn’t matter. To get this into market, you need help, or you need to spend large sums of your own money to get visibility, and the initial sales. In storage, who is going to by from Joes fish, chips and terabytes of clustered storage, even if there is great technology in the mix? I mean, why trust your data to anyone but the big guys, cause they will be there next month …
… of course this argument is complete hogwash, but the perception is that there is safety in the larger vendors. Large vendors have killed off product lines which customers have built business dependencies upon. It is rare that large vendors will develop “risky” products, which are products which have significant potential to shake up the status quo in a market, by offering capability/functionality that customers want/need. Moreover the larger vendors will often claim that the “risky” functionality is not needed … that is … until they bring it out themselves. They will feed the perception, and drive home the FUD. This makes it harder …
… to get traction for innovative (aka “risky”) technologies. Which means that quarter over quarter, these could be loss-makers rather that profit-makers.
So, while TerraScale and subsequently Rackable, had good technology available, it was likely frustrated in its marketing, in part due to perception issues. This is a guess on my part, I don’t know.
The Blocks and Files entry then points to, what it believes to be likely suitors for this technology. Specifically pointing to those “without” their own clustered file system capability. Sun is mentioned, as is HDS, and Dell.
I found it interesting that Sun was mentioned. CFS assets were acquired by Sun. This doesn’t mean that Sun bought CFS, but that CFS was in somewhat dire straights, and that Sun swooped in and got the assets for a bargin. Arguably, similar to their MySQL acquisition.
Remember, Sun pushes zfs, and they push it hard. The TerraScale technology was based upon xfs. Sun is pushing (and has been pushing) zfs to sit underneath Lustre. I don’t think that Sun wants to dilute what they perceive to the value of Lustre in terms of zfs integration, by acquiring something that doesn’t fit into this strategy. That is, I would think that Sun is not a good candidate for this. Obviously I could be wrong about this.
Not sure of HDS, I don’t think I see this meshing well with their kit. Dell is possible, but remember, they sell hardware, and clustered file systems usually have some sort of service component.
But who is going to trust this technology? If it isn’t being pushed by one of the “majors” … you know, the tier 1’s, it probably won’t fly. Someone is going to need to commit marketing $$ into this to make it work. I am not convinced that the engineering is the problem. I am convinced that the lack of sales is.
This actually gets to another interesting and related problem worth talking about. I can count at least one company that has built a storage system with Lustre as its core product. It did this prior to Sun acquiring the assets of CFS. Now here is this small startup, and they are selling hardware and software which directly competes with Sun’s hardware.
I am not sure they are long for this world unless they can adapt, and in this case, it might be something like reselling Sun hardware. Or leaving Lustre behind and adopting other technologies. Why would Sun want a competitor selling hardware against Sun, using a Sun developed bit of software?
FWIW: the Lustre repository is here. There are many subtle ways that Sun can impede a competitor using this technology, including keeping separate trees internally and externally. Anyone can download the tree … from the external respository. What, you want the updated stuff? Well … buy it. From Sun. This statement on the wiki suggests something like this:
The latest version of Lustre is always available from Sun Microsystems, Inc. Public Open Source releases of Lustre are available under the GNU General Public License.
The fundamental suggestion in the text, is that they are not the same. The latest version may not be the Public Open Source release.
And this small startup did build a product upon that. I would imagine they are looking for a new CFS. They may vehemently deny this, and indicate support for Lustre now and forever. I would be at least thinking about an alternative if I were in their shoes.
What is odd about this, is their name. And how similar it is to the business unit now on the block from Rackable. TeraScala.
If I were their CEO, I would likely go back to my VC friends to get the capital for an acquisition. This would be a hedge against Sun doing something that knocks them out of using Lustre. Like changing the licensing.
But they will … and do … have the same issues that Rackable had. As all small companies have. Like ours. Getting the attention of the right people. Getting them to accept that “risks” have upsides and downsides, and you can’t get the upsides without taking the “risks”. Risk-free computing means you are on IBM mainframes which will run forever. System 360. Wouldn’t want to try them-thar newfangled vaxes. Or heaven-forbid … unixen.
The small guys are the innovators. Paraphrasing George Bernard Shaw
“The reasonable company adapts itself to the world; the unreasonable one persists in trying to adapt the world to itself. Therefore all progress depends on the unreasonable company.”
Rackable was simply not successful in getting the world to adapt. Thats the marketing side.
The engineering side is not as much of an issue. If you aren’t aware of GlusterFS or ceph yet, you should become aware of them. The problem isn’t completely solved, but it is working well enough.
Update: It is worth noting that one (possibly very large) reason why Hulk/Maui and others are not being pushed hard is the reluctance of these companies to slaughter their own cash cows … unless it looks like a competitor will. They can prolong investments and make development take as long as they wish. Once a serious competitor, that is, one that is taking noticeable sales away from them, or pressuring them on price, arrives, expect that kit to come out fast.