Mad Dog 21/21: Recovering Lost Prophets
February 18, 2008 Hesh Wiener
If you ask computer professionals who they thought were the leaders in data storage, chances are you’ll hear names like IBM, EMC, Hewlett-Packard, Network Appliances, and Hitachi. If you asked these storage companies, you might get a different list that included some outfits that were largely unknown in the end user world. But end users probably don’t have to worry that they will miss out on some important new vendors. The Storage Establishment is buying up many of the emerging suppliers. This high-profile interest in the new crop of vendors makes all the contenders a lot more visible.
Still, the computer business is huge. It’s possible that some of the storage system developers with the very best ideas will still for the most part be overlooked for a considerable time, like the caves of Qumran. Data recorded on parchment and other media sat in these caves for a couple thousand years until 1947, when a shepherd stumbled on some of the documents, beginning a process of archeological discovery that lasted about nine years and yielded several hundred items.
Qumran is near the Dead Sea, and the records found in nearby caves are known as the Dead Sea Scrolls. The material includes bits and pieces from all the books of the Old Testament except for the books of Esther and Nehemiah, local laws observed by the community that owned the scrolls, and material that is not in the Old Testament as defined by, for instance, the King James Bible; this material, such as the Book of Enoch, appears in some other bibles. There is a lot of wisdom in those scrolls, and plenty of prophecy, too.
There is some scholarly dispute about the community that created and guarded the scrolls. The currently prevailing view is that they were Essenes, at least at the end of the period during which the scrolls were created, an era that began a couple hundred years before the birth of Christ and lasted a total of perhaps 250 years. During that period, political control of the region was in flux, but towards the end of the era, political power was pretty solidly in Roman hands. Nonetheless, the Romans did not completely control the data storage business. The Dead Sea Scrolls, decidedly not written by the Romans, got lost; they were rediscovered after the end of World War II. Their disappearance for a couple thousand years is probably the most important reason they survived. Empires aren’t always good about data storage, particularly when the data isn’t in line with imperial policies.
Today, a few big vendors’ empires control most of the storage business, but their control is less complete than that of the Romans two thousand years ago, and territory changes hands all the time. The storage emperors don’t really care about content, only about who gets to sell the media. There is no way to predict which vendors, which systems, and which data will be around for a long haul, whether there will be a one or a few big empires, or whether the storage market will break into many small pieces. That uncertainty is something data storage today has in common with data storage in the centuries surrounding the birth of Christ.
Of course, the goal of the big name storage companies today is not keeping the Bible intact; it’s about keeping a customer base intact for the next quarterly report, or perhaps for the next several reporting periods. As they move towards these unsurprising goals, the big players in storage make a lot of heat and noise, maybe more than is necessary.
Last November, in what has been one of the most widely publicized deals in storage, Dell said it was buying EqualLogic for about $1.4 billion. EqualLogic is a provider of iSCSI disk subsystems that are a hit in the small business segment but not a big factor in the enterprise market. Dell’s move unleashed a hail of commentary from financial analysts and industry observers, and the main topic was the likely impact of this deal on Dell’s role as a big reseller of EMC storage hardware. Dell probably accounts for about 15 percent of EMC’s top line.
As it turns out, Dell says it will continue to offer storage subsystems it buys from EMC. The EMC boxes will generally support different kinds of applications than the EqualLogic equipment. Still, it’s anyone’s guess how it all will ultimately play out. In the meantime, EMC is rolling out new and upgraded software that promises to give its established products some of the attractive features offered by younger rivals.
Network Appliances, which is younger then EMC but older than EqualLogic, is refreshing and enriching its product line to stay competitive and trying to figure out how to avoid getting crushed between giants. It might have to buy some of its competitors to maintain its weight status, which is either middleweight or heavyweight depending on how you look at the business.
Other players that are still independent but which might be acquired by any of the big vendors if a feeding frenzy occurs include Compellent, Lefthand Networks, and 3PAR. All have worthwhile products with lots of innovation, all have to meet some serious challenges.
What makes the acquisition of small fish by larger ones seem more likely right now is IBM’s decision to buy an array developer rather than build a new range of disk arrays. If IBM, which is not always the first out of the gate, doesn’t think it has time to develop its own stuff and instead is buying whole storage companies, it might be time for everyone else in computing to get moving . . . . or risk missing a party in storage that is clearly underway.
IBM’s move was the last thing it did in 2007. On December 31, IBM said it was going to acquire XIV. XIV is Israeli, and its chairman, now an IBM employee, was Moshe Yanai, who is famous in the disk trade for leading the team that created the EMC Symmetrix, a disk array that enabled EMC to rise to the top of the enterprise storage market. Yanai and EMC went their separate ways a few years back. Since then, Yanai has been involved in a couple of storage ventures but nothing with the kind of allure that could get the attention of an IBM or EMC–until now, when Big Blue cut its second major storage deal with roots in the Holy Land.
IBM’s first storage deal connected to the Middle East was its decision to remarket disk arrays from StorageTek (now part of Sun) from 1996 through 1999. In those days, EMC was just killing IBM in enterprise storage and IBM simply did not have a competitive array, while StorageTek did. At the time, StorageTek was struggling to survive in the wake of Chapter 11 bankruptcy, but it had some really excellent products. The superior engineering that was as much a characteristic of StorageTek as dramatic ups and downs in its financial condition stemmed from the character of founder Jesse Aweida.
Aweida came to the USA from Israel (or perhaps less legalistically and more accurately Palestine) as a student. IBM helped him pay for his studies. He became one of the stars of IBM’s magnetic tape engineering group but, driven by irresistible entrepreneurial urges, left to lead the startup that ultimately took a significant piece of the tape market. StorageTek pioneered and made a big success of the enterprise tape library, but, like several other vendors before it, ran out of money trying to beat IBM in the disk market.
Getting IBM out of the data storage business has proved to be about as difficult as driving the Roman Empire out of the Holy Land. The Romans eventually did lose their grip, and the region later became part of the Byzantine Empire. Along the way, the cultural heirs to the Dead Sea Scrolls, including the Torah, transformed the pagan Romans and their Greek-speaking successors into Christians. When, after more than a thousand years, the Ottomans replaced Christian rule in the region, their faith, too, had roots intertwined with the Torah and the Scrolls. It might be argued that all the major battles in the Holy Land during the past couple thousand years have not changed the foundations of the region’s philosophy and beliefs, but rather were fights about the structures erected on those foundations.
So far, IBM has not lost its storage business, although its market share does wax and wane. Even if it IBM was chased out of storage, many of its ideas are so deeply embedded in computing that it can be said that its body pioneering work on drums and disk drives, controllers, and interconnection forms the Old Testament of random access storage. So, if you wish to temporarily put aside the fact the storage is a business and look at it as a process or technological development, you can say that nobody wants to get rid of IBM’s ideas, but a lot of inventors want to keep Big Blue (and other established vendors) from standing in the way of their ambition.
The inhabitants of Qumran and nearby villages weren’t trying to toss the Romans out of Palestine, but also didn’t go along with all the things the Romans wanted. For instance, the religious folk living in the hills near the Dead Sea would not comply with at least one key aspect of Roman Empire law, which under certain circumstances compelled all Rome’s subjects to treat the emperor as a god. None of the Jews of that time and place, nor any of their spin-offs, sects, or whatever you want to call Pharisees, Essenes, Sadducees, Samaritans, and other Torah-toting monotheists of the time and place could reconcile their God with an emperor’s pretensions to divinity. It’s an idea that just plain stuck in their craws. But the Romans couldn’t swallow the locals’ attitudes, either. The two sets of beliefs were so far apart there probably wasn’t even anyone in authority who could state the problem in a way that opened a path to resolution. In the end, it came down to who had the bigger, better army, and that was without a doubt the Romans.
In storage, the big struggles are not about defining the problems, but about providing (and selling) a solution.
The problem that StorageTek solved better than IBM in the mid-1990s, the problem that EMC solved better than anyone at the time, and the problem that all the youngster firms in storage and their older cousins are always trying to solve is this: How can you best use the latest components in a dynamic technical environment and still provide a storage system that can be used by legacy applications on legacy servers?
Today, the cheapest storage arrays use SATA disks similar to the ones included in home PCs. Their controllers use basic X64 server technology, open source operating systems, and economical interconnection schemes based on Ethernet. They try to compensate for the higher error and failure rates and lower performance of cheap disks and, the vendors insist, they do a pretty good job of this.
It seems likely that for many applications the more expensive, higher specification SCSI drives that populate most current arrays may be relegated to a premium corner of the market. But that presumes the new disk arrays really work as promised, and perhaps not all of them will measure up. The movement to a new generation of arrays could take longer than Dell or IBM expect, giving EMC, Hitachi, and Sun ample time to upgrade their offerings . . . or buy into what seems to be an inevitable future.
The details of this future very quite a bit depending on which storage array company you ask, but all the visions seem to have some elements in common. By and large, it looks like X64 chips will lie at the heart of the next generation of controllers. IBM, with its Power chips, could be an exception, but the work done by XIV with commodity chips suggests otherwise. IBM cannot afford to make vanity choices here; there are just too many hard-driving players in the disk array game right now.
Unless disk makers can find a way to reduce the cost of SCSI technology, SATA drives with their lower cost seem destined to dominate, much the way standard 3.5-inch SCSI disks killed off other drive technologies and other form factors. The high performance portion of the market could go to flash drives, but flash remains about an order of magnitude more costly than spinning disks, and that really cuts down the market for the solid state drives. Still, EMC is offering flash disks for it high-end arrays, and it might find there is a bigger market for this luxurious option than skeptics predict.
But the disk array developers are doing a lot more than finding ways to wire cheap disks to cheap controllers. They are also trying to develop related technologies to make their arrays more attractive than any products on offer from the established vendors.
One feature common to new arrays that is not generally available on installed storage subsystems is thin provisioning. What this means is that the user can set up a storage volume without allocating a lot of physical space until it is actually needed to store data. This saves hardware, electricity, and space, but unless it is done very cleverly it can reduce performance and make some functions like data replication a lot more complicated.
Another thing users want from the new vendors is scalability. So, the array makers all have come up with ways to add controllers, disks, and interconnection technology that makes their arrays grow seamlessly. (They are clustered, like most nodes in supercomputers are these days, too.) But no array can grow forever, and it remains to be seen which of the new products really span what range of capacity.
A third aspect of all the new arrays, one that underlies just about every feature, is virtualization technology. Disk virtualization separates the logical and physical storage devices, so an array can more readily incorporate improved drives as they come to market. Processor virtualization lets array makers abstract control functions, supporting among other things new ways for controllers to manage recovery from hardware faults and failures. It also allows for the dedication of virtual controllers that use only a fraction of a physical controller to low-demand tasks. Such controllers might support development work or persistent legacy applications that don’t really need a whole physical controller or, as is the case in some of the emerging systems, a whole collection of physical controllers.
In addition to these core features, all the companies developing storage subsystems try to find new functions and features that their rivals simply do not offer. Some of these ideas will grow to become commonplace and if their originators are clever and lucky, the inventors will prosper as their ideas catch on.
Well, maybe. It might also turn out that most of the notions storage inventors promote in the hope that they can differentiate their products from the offerings of their rivals will in the long run be at most footnotes in some history of the computer business. The Dead Sea Scrolls, even though they include a the oldest known versions of the Torah, also include a lot of material that didn’t make the cut when the body of text called the Old Testament was put into its prevailing form. Some of the material in the scrolls may be older, in literary terms, than Genesis. Other material is about the prevailing rules of the sect that created, copied, and preserved the scrolls. Still other material seems to be stuff that ended up in the Gospels but which might first have been written well before the birth of Christ; this is the kind of literature that gives scholars lots to argue about and gives traditional theologians fits. Does it matter if the New Testament echoes some lost books of what might have once been the Old Testament? It doesn’t to people who simply want the ideas of their faith in language they can understand, but it’s very important to those who take the Bible literally.
If the outfits that use computers just want to store more data for less money, and do so quickly and reliably, they might not care just how their disk systems work or even who invented one or another gadget inside the boxes. That’s not everyone of course, but it is part of the market. There are also users who don’t really care about much except minimizing uncertainty. They will look for arrays from the strongest vendors, whether or not they happen to be the strongest products. And there are always the innovators, companies that know that they will be happiest if they can get some equipment that is at the leading edge. This is probably the smallest part of the market, but it’s the segment that EMC sought when it first offered arrays to compete with IBM’s traditional disks, so its importance vastly exceeds its size.
However the market unfolds, it seems likely that old and new vendors will be pushing larger and less costly arrays to satisfy what for now seems to be an unquenchable appetite for storage capacity across every industry sector and in every geographic region.
It points to what will be a necessary future development, perhaps the next big opportunity for storage system vendors: creating archival technology that can keep pace with services using gobs of online and nearline storage. It would be a mistake to underestimate the importance of tomorrow’s storage systems and the information they will house. Somewhere in Bebo, Facebook, MySpace, or another social networking site there might or might not be some Web pages belonging to prophets, but it’s a safe bet that those sites and their huge file libraries of will generate plenty of profits. And that says nothing about a related business that is expected to become pretty big, pretty fast, storage as a service delivered via the Internet, a business in which the storage device makers might also be the vendors. What is less certain than the prospects for the storage business is whether the information stored on these sites will be around in 2,000 years, and whether it might be a good thing if it’s not.