You rarely have mail stolen...

This should be emphasized. I visited a gold mine in the US once. Was astounded when they told us they mail their raw ingots (that contain gold, silver and platinum all mixed together) to their refiner by USPS. They matter-of-factly told us that only USPS had the kind of government-force-backed security and guarantees that made transporting around >$100K bars every day feasible.

Cisco Telepresence is what I would consider good enough to have virtual walls that give an "as good as being there" experience. I do a lot of remote work (about 80% of my revenue is all remotely-delivered), and believe me, "a few mbps" over a consumer-grade service provider just doesn't deliver a business-quality video conference, much less the kind of quality needed for always-on telepresence that doesn't become fatiguing from eye strain. Even apart from the spendy infrastructure ($300K for the high-end, $80K for the "budget" version), the Cisco kit requires QoS-delivered, low latency, 9-10Mbps symmetric for the three-screen configuration. It seems to me that there are still technical problems to hammer out. If you think "a few mbps" is sufficient, I can tell you haven't tried always-on telepresence for longer than several months.

It is fascinating that business professors are not inundated with requests from programmers clamoring for just an idea person to pair up with to change the world.

That indicates there is a supply-demand imbalance between idea people and programming people, that all these supposedly business-oriented idea people are overlooking. An imbalance that is classically rectified by raising the value of programmers, and/or lowering the value of idea people until the supply-demand curves intersect at a more mutually-acceptable point.

In any case, trying to discuss whether the idea or the perspiration is more important vastly oversimplifies the complexity of a business. I suggest that the more relevant matter to pursue in a technical forum is, why are so many programmers such poor negotiators that as a group, they have come to be perceived as a commodity in the mainstream?

USPS, UPS and FedEx will like this IF they are involved.

Point out how much they will save on payroll, and in the US, payroll and medical and retirement benefits. You'll have them beating a path to your door (provided you solve the engineering and reliability problems others have pointed out in this thread).

I would imagine there are plenty of people who want to know how well it performs - regardless of features - in comparison to other filesystems.

These people are arguably not the population ZFS' requirements are intended to satisfy. Looking to filesystem performance to solve large data set performance issues unnecessarily narrows the scope of the architecture design problem space. Anyone with small data sets would likely be using ext[34], anyone with large data sets, frequently spread across multiple servers or server clusters, would likely be treating filesystem performance as one among a multitude of design points.

Until btrfs is production-grade in business-critical environments, there is literally no other extant filesystem that can fit the same requirements footprint as ZFS. IMHO IBM (especially the storage business unit) missed a golden opportunity by letting Sun go to Oracle, and their services and DB2 arms are really going to regret it in 3-5 years.

Anonymous Coward pretty much answered what a presales engineer does, but didn't explain why it is at the bottom of the list. The list appears to be sorted by salary range or percent increase. Note that presales engineer had the largest percent increase, but the lowest salary range in the list.

Many presales engineers (especially at the big companies) have a compensation plan that is part salary, part commission and part incentive bonus plans, so this table might not be an accurate reflection of what they really take home. There are many people who dislike the road warrior-esque nature of many presales engineering positions, so it is difficult to recruit really good, seasoned staff to this position who have the right mix of technical and sales personas to pull off the role well. The really good ones are pretty much visually indistinguishable from the sales and account execs (ditch the shorts and sandals that are fine when you're coding at the office for coat and tie), engender confidence in the customer when they present the technical solution, and write up the sales proposals from soup to nuts. What usually sets them apart from the pure sales function is they might not necessarily be quite as extroverted as the sales people.

...and Solaris instantly becomes next to worthless, except for Oracle DBs and big Corporation purchases...

We use Solaris for its ZFS, as no one else has continuous integrity checking in a production-grade filesystem; for hundreds of terabytes, we don't feel comfortable with any other filesystem. FreeBSD is coming close, but ACL support is still very lacking.

As long as merchants still pay for credit card fraud, where is the business case and incentive for the card issuers to adopt this technology as they are currently laying off the risk and the benefits for adopting do not accrue to them while they incur all the costs of adoption? As a consumer, I would purchase this just to collapse all my mag-stripe cards (not just credit/debit) to one card that was secured with a PIN that I could change myself, if it could be sold that way.

jfs and ext3/4 fs for me, for now. maybe some md-raid if I need instant recovery from a failed disk.

But for large-scale storage (say, 300 TB and up), how do you address the need for continuous integrity checking? I don't see anything like it under ext3/4 or any Linux/BSD OS variant that is production-grade today. Sure, btrfs has it, but it isn't production-ready for at least a couple years, if not longer.

...with Verizon when they finally decide to roll out their fiber optic service to my neighborhood.

You are going to be waiting a long, long time, as Verizon has stopped their FIOS expansion for the indefinite future. Why anyone in a FIOS-served area would ever choose any competitor for Internet service is beyond me; their sell-through rate (ratio of subscribers to all potential subscribers) on their Internet service should be way, way higher than its current 25% or so. I'm currently waiting for AT&T Uverse service to reach my area.

Were you guys using ZFS, and if so, how are you addressing its gargantuan addressing capacity under AIX? We think we have a way out using FreeBSD, but we're also evaluating AIX as we expect the schizophrenia at Oracle to continue for awhile.

We fully expect OpenSolaris to wind down over time (even with a fork) without explicit support from Oracle, so we don't want to start planning around OpenSolaris only to have to switch again. The path I'm investigating now is moving our mass storage platform to FreeBSD, and using the net/istgt iSCSI target and Samba to replace the ZFS share properties for those protocols. If that works in our torture testing with our controller cards, we're moving off of Solaris except for a lone server to test against the software we sell.

Oracle shut down all the over-the-web purchasing options for Solaris support contracts, re-directing everyone to their local reps. That pretty much screams, "unless you're talking a minimum of 4 figures and preferably 5, don't bother us".

We were willing to put in enough of an effort to adopt sufficient amounts of Solaris to have it manage all our spinning platters for us. We even had plans to integrate ZFS into our backup system (IBM TSM) so that we would automatically restore files from tape that ZFS indicated were damaged. But most of our work takes place on Linux (RHEL where required by the business application, and Ubuntu otherwise) and OS X, where frankly the hassle factor for maintaining an infrastructure component like an operating system is far less than Solaris (pre-Oracle, we had to wait nearly two weeks to get our support contract ID after purchasing it online). Sun hardware only made sense for us if we were planning on running compute services off of it, but Sun kept Solaris bottled up for so long that by the time we could try it out our infrastructure was already built up around Linux. Switching costs are too high for even ZFS to justify. If Oracle kept to the same support licensing terms however, I could see us gradually move over services one by one when it came time to migrate them to new hardware (which is when we usually evaluate whether it makes sense to switch OS platforms for the application).

I have enough on my hands that I don't need schizophrenic support licensing terms for basic infrastructure. Oracle has clearly signaled that unless you intend to make a major commitment to Solaris by willingly locking into their hardware and software NOW, they don't want your business, even if that timing doesn't fit with your business plans and planning horizons. If their support processes weren't stuck in the Stone Age, and the quality of patches weren't so sketchy that we found we must use Live Upgrade to protect ourselves, we probably would have shelled out. But Oracle jacked the premium for ZFS so high and so quickly they made the decision easy for us to start paying the money to test the alternatives.

We're drafting up backup plans to migrate our mass storage architecture off of ZFS and onto an ext4-based distributed filesystem on multiple nodes, in case FreeBSD doesn't work out. It will cost a bit less than what Oracle wants now, though it will still cost more than a basic support contract for a single Solaris server (pre-Oracle) spinning all the same spindles. We're hoping we can get by with FreeBSD 8.0/ZFS/istgt/smb for the next 4-5 years, and hopefully the situation between ZFS and btrfs under the same roof at Oracle is resolved by then.

At this point what's to tell Oracle that Solaris is better than Linux, because, I'm not sure they're convinced?

For my company, one acronym: ZFS. We're going to start clocking into petabytes of storage within a year, and right now we're handling the tens of terabytes of storage under Solaris with a basic support contract so I could pick up the patch updates and email with the odd, once-a-year problem I couldn't solve myself. I'm shudder to think of the supporting the same scale with any other filesystem; ZFS has seriously saved our asses several times now with just its scrubbing feature.

Oracle's new licensing policy has now put us into a bind. We now have to pick up Oracle Sun-branded hardware, plus the hardware support contract, plus the Oracle Premium software service plan. Then re-integrate the hardware with our existing configuration, possibly picking up new controller cards. Our carrying costs per year for choosing a Solaris-based solution just jumped an order of magnitude.

The only reason we haven't started planning a move to FreeBSD 8.x is because FreeBSD ZFS doesn't yet support iSCSI (because FreeBSD doesn't have an iSCSI target yet). ZFS just got hella more expensive.

Considering Apple's silent dropping of ZFS, I take it as a sign that in the future ZFS development will likely clam up to just Oracle Sun Solaris. Thus, we're going to follow Apple's lead and start testing ext4 under Linux (we first came to ZFS from ext3). I like ZFS, but not enough to justify a 10X cost difference unless there is simply no other way to hold petabytes off a single server.