Proof point: the $400 million exabyte data center
Robin Harris calculated the cost of a 1.8 exabyte datacenter to be $400 million. That’s assuming the barest bones Google-style storage architecture and no redundancy.
$200 million for the 2 million hard drives, another $1 million for the racks, $60 million for the facility, and $2 million per month for power.
That’s a lot of dough! Or a very affordable two exabytes, depending on how you look at it.
The relentless growth of capacity
What happens to these numbers over time?
Arguably racks won’t get any cheaper, and real estate and power over the long haul will track with inflation. Given that half the cost is pure storage – the bytes themselves – what’s the impact of the March of Progress on this number over time?
The result is what we’ve already seen play out over the past decade or more. Data center physical costs follow conventional capital budget cost curves, while capacity multiplies every year or two – rain or shine. The result: a relentless decline in cost per exabyte.
With 2.5″ drives and SSD, there is no end in sight to this cost/capacity discount.
Storage is the mother of all web applications
Yes, web technology is cool. But without the economic foundation of ever cheaper exabytes, the internet that we know today – and Google, Facebook and Twitter – would have remained a twinkle in Silicon Valley’s eye.
Agree or disagree? Let’s hear it.
New ideas for storage may be changing how we buy bytes
Curious about the mysterious ISE from Xiotech? A picture paints a thousand words, and a video does it all the better. Here’s a fun peek of the ISE at SNW from Robin Harris, coincidentally generating 48 simultaneous video streams. Oooh! Aaah!
It’s energizing to see conventional disk drive technology turned on its head to provide dramatic new value. It goes to show that innovation is limited not by technology but by its application.
Atrato is another innovative take on ‘black box’ storage – keep an eye on them as well.
Moving from exciting technology to real-world product is not trivial
Robin Harris posted yesterday and last week on flash – both are interesting reading. My takeaway is that the more flash is applied, the more real-world wrinkles bubble to the surface. That’s exactly as it should be – new technology buzz always begins with what’s possible, then moves to “OK, now how exactly will that work in my solution today?”
Borrowing from disk drives to make flash work
Also noticed how most of the issues and workarounds are things that have already been addressed with today’s disk drives. Just goes to show that storage devices are more than the media. Seagate’s in a great position as they enter the flash solutions world because of their depth in experience in making storage devices out of storage media.
What do you think?
Comments welcome! How do you think you will use your first flash storage (beyond thumb drives)? Have you already?
Xserve RAID is out at Apple: Promise Vtrak is in
StorageMojo reports on Apple’s decision to drop the Xserve RAID storage system. Seems like its beauty was only skin deep.
Goes to show that being pretty is a great strategy for consumers, but not so much for business. A smart and pragmatic decision by Apple.
Replaced by Vtrak from Promise
Apple is offering the Promise Vtrak system in its place. Comments on Robin’s blog suggest that Apple is making this move now to get it out before their update to Final Studio Cut software. You can take a look at it on Promise’s site here.
The A/V production space is huge. Apple is wise to do what it takes to maintain a strong position in this market.
Content access time is the performance metric for Web 2.0 enterprise storage
A very cool entrant in the storage space: Atrato. Think “black box” storage: hundreds of 2.5″ drives, sealed in a 3-year maintenance-free containter that’s “fail-in-place” Does that mean it keeps working for 3 years despite losing a few drives along the way? Let us know if you know.
Atrato is focused on instant access. This is the high capacity content-serving equivalent of transactional performance for traditional enterprise applications. Expect to see more solutions for Web 2.0 infrasctructure that are screaming fast, but in a new way.
Robin Harris has a nice analysis of their new idea and product here.
By the way, Atrato is a river in Columbia that Atrato (the company) says is the fastest in the world. Creative name choice!