Kelly Lipp - Vice President, Manufacturing and CTO

Kelly Lipp graduated with a degree in Electrical Engineering and Computer Science from the University of Colorado. As a System Engineer at Ford Aerospace, he was responsible for delivery of hardware and systems design and documentation. Moving to Digital Equipment Corporation as a systems administrator in the storage division, Kelly became an expert on the OpenVMS operating system, storage and networking. In 1995 he, along with three others, formed Storage Solutions Specialists, Inc., to work with IBM to develop the Tivoli Storage Manager client for OpenVMS. The company discovered a need to make TSM simpler and the focus changed from software to appliance development. STORServer, Inc was formed to further these efforts. Currently as Chief Technical Officer of STORServer, Inc. he is responsible for developing appliance solutions meeting diverse customer needs in the backup and storage environment.

Final Jeopardy Answer: Holds 1.6TB of data.  Has data transfer rates of up to 150MB/sec.  Consumes zero power and generates zero BTUs.

Green is the new buzzword in the storage industry.  Practically every advertisement of every product has a green component to it.  But what does it mean to be green?  Specifically, what does it mean to have green backup?

The backup industry is less green now than it was even two years ago.  The proliferation of disk-to-disk backup is directly responsible.  Even with all the data deduplication technologies available, the growth in the number of disk drives because of backup applications has been tremendous.  And what did we really gain:  faster access to data IF that data is required.  But if we aren’t accessing that data, why are we keeping an ever increasing amount of it on an energy inefficient resource?

What is your green backup quotient?

“I insist that all of my reference data have an access time of less than one second.”  Not Green.

“I insist that all of my reference data have an access time of less than one second in two locations.”  Clearly Not Green.

“I insist on having an access time to all of my archived data of less than one second on the off chance that regulators will want access to it.”  Again, Not Green.

If you consider that at a minimum your reference data will be approximately six times your active data, that’s a lot of extra disk required to provide value of a dubious nature.  Must you really have sub second access times to any, much less all, of your reference data?

Let’s look at the numbers.  A typical disk drive consumes between 8 and 12 watts depending on what it’s doing.  If it’s just spinning then it uses 8 watts.  Reading and writing, it uses 12.  Let’s choose ten watts to make all subsequent math easier!  So, ten disks consume the same amount of power and generate the same amount of heat as a 100 watt light bulb.   An average disk enclosure housing 10 drives consumes about 300 watts total.  So let’s agree that each disk drive, including the ancillary hardware required to make it useful, consumes on the order of 30 watts.

Using 10TB of active data as an example, let’s do our math.  Though we all wish we had the latest disk technology, this 10TB is probably housed on many sizes of drives.  If we assume an average disk size of 250GB, which is probably way too high but I like easy math, this equates to 40 disk drives.  40 Drives * 30 Watts/Drive = 1200 Watts.  About the same power consumption and heat generation as an electric hair dryer on low power, 24 hours a day, 7 days a week, 365 days a year.  And to add insult to injury you have to get rid of that heat from your computer room.  We’re not even looking at AC costs here!  Hmmm.  I won’t speculate as to what we might say to our teenage daughter about this blatant waste of electricity.

If reference data is six times our active data add six more hair dryers to the computer room.  No wonder it’s so hot in there.  And how many computer rooms have only 10TB of active data?  Alex, there are no rhetorical questions in Final Jeopardy.

There isn’t much we can do about reducing our energy consumption for our active data except to purchase newer, more efficient disk drives and enclosures.  So, let’s focus on our reference data.  Do we really need sub second access times to all of this data?  We certainly may require fast access for some of it, but definitely not for all of it.  In fact, the vast majority of it will never be accessed again, so why have it spinning at all?  Usually, if we access reference data, we access data that is less than a couple of days old.  The older the data, the less likely we are to access it.  So our strategy should be to keep relatively new reference data more accessible and older reference data less accessible.  How much new reference data is there?  Generally new reference data is a fairly small percentage of the active data.  If the reference data consists almost entirely of backup data (as oppossed to archive data) then we can use a typical 10% daily change rate factor to get started.  10% of 10TB is 1TB or about four 250GB disk drives and 120 watts.  You can double or triple this number to add a couple of days of reference data stored on disk for fast access.
What should we do with the other 50 to 57TB of backup data?  We want access to it.  Since we don’t expect to access this data very often, we should be willing to extend the access times.  How much though?  That depends, but certainly if we can gain access to this data in less than an hour that should meet many if not all of our most stringent requirements.

Final Jeopardy Question: Alex, What is an LTO4 tape cartridge?  That is correct!  Let’s see your wager.
If we assume an LTO4 cartridge capacity of 1TB, all of our reference data will fit on 60 cartridges.  Let’s do our power math on a tape library that holds 60 cartridges.  The library consumes on average 250 watts.  Each tape drive, when reading or writing data, consumes about 50 watts.  When the tape drive is sitting idle it consumes about 10 watts.  Writing 1TB of data to tape requires approximately four hours of tape drive time at an average of 250GB/hour/tape drive.  To write our 1TB of incremental backup data in a two tape drive library we’ll consume 350 watts of power for two hours a day and then 270 watts the rest of the day.  We’ll use an average power consumption of 300 watts.

60TB of reference data on disk consumes 1200 watts.  That same 60TB of reference data with 1TB on disk and the other 59TB in a 60 slot, two tape drive library consumes about 420 watts (300 watts in the library and 120 watts on four disks) or almost three times less!  That’s three times less power you need to purchase to store your reference data.  Oh, and a 120 slot library with two drives consumes the same amount of power so you can double your reference data size for no additional power dollars.  If you double that same reference data on disk you’ll double your power dollars.  And that does not include increasing your AC requirements to remove all the heat from the computer room.

Will you triple your power consumption to provide nearly instant access to all of your reference data?  Not Green.