How important is "Blackbox"?
Just in case you don't know what blackbox is, here's how Sun describes it on the product marketing page
17.October.2006 - After today, you'll never look at an ordinary shipping container quite the same way again. Project Blackbox is a prototype of the world's first virtualised datacenter--built into a shipping container and optimised to deliver extreme energy, space, and performance efficiencies.Designed to address the needs of customers who are running out of space, power and cooling, Project Blackbox gives customers a glimpse into the fast, cost-effective datacenter deployments coming in the near future--where thinking out of the box means putting an IT infrastructure in a box.
"Just about every CIO and startup I meet says they're crippled by datacenter energy and space constraints -- today's solutions are clearly failing to meet the needs of Web 2.0," said Jonathan Schwartz, CEO and president, Sun Microsystems. "Rather than trying to improve upon today's datacenter, designed for people babysitting computers, Project Blackbox starts from the world's most broadly adopted industry standard, the shipping container, and asks -- how can we most efficiently create modular, lights-out datacenters from this base? The answer...with one-hundredth of the initial cost, one-fifth the cost per square foot and with 20 percent more power efficiency, we can deliver an immense multiple of capacity and capability -- anywhere on earth."
...
The Project Blackbox prototype is a computing powerhouse capable of hosting a configuration that would place it among the top 200 fastest supercomputers globally. The current prototype could support the following capacities:
- A single Project Blackbox could accommodate 250 Sun Fire T1000 servers with the CoolThreads technology with 2000 cores and 8000 simultaneous threads.
- A single Project Blackbox could accommodate 250 x64-based servers with 1000 cores.
- A single Project Blackbox could provide as much as 1.5 petabytes of disk storage or 2 petabytes of energy-efficient tape storage.
- A single Project Blackbox could provide 7 terabytes of memory.
- A single Project Blackbox could handle up to 10,000 simultaneous desktop users.
- A single Project Blackbox currently has sufficient power and cooling to support 200 kilowatts of rackmounted equipment.
That user count, by the way, is seriously in error - sometimes the people at Sun marketing don't believe their own company's results.
The question, however, is whether project blackbox represents anything important?
I think it both does and doesn't. On the positive side:
- It's an interesting strategic reversal in the IBM/Sun competitive relationship and a second shoe in the StorageTek takeover story - a deal that got Sun a bunch of largely obsolete products and several thousand sales and support people with trusted access to large IBM data centers. Those people can now sell "Blackbox" - and that's going to benefit Sun, benefit the customer, and take budget dollars away from IBM.
- It's a breakthrough in Unix system administration because one person can do it all: manage an entire system capable of 5,000 concurrent threads accessing 600TB of data. All enabled by the N1 tools, Solaris 10's fault management technologies, and the simple assumption that failed parts can be left in place if their failure has no effect on system performance.
- It's an obvious precursor to smaller systems aimed at smaller customers or niche jobs -systems down to the "Sun Pod" I discussed in January with integrated power, networking, storage, and one or more Niagara processors in a rack.
On the negative, all of the positive considerations for this thing have sales and marketing opportunities attached to them - not technology opportunities. Whether Sun assembles your Sun Pod or you do affects what it costs to own and operate, not what you can do with it. Similarly whether Sun provides your data center in a blackbox or you set it up on the traditional floating floor affects cost but not what the technology will do.
At best, therefore, the technical impact and therefore value here is indirect: it will strengthen Sun financially and thus enable more R&D work, it will expand Sun's markets into IBM's data centers and thus lead to some intellectual cross-fertilization in administration, and it will ultimately bring more computing power per dollar to smaller businesses and thereby lead to greater innovation. All great, but all indirect.