Lifespan: Up to 6 petabytes of Writes

Seagate is calling the Pulsar a 5 year drive. If you are absolutely dying to know how many writes you can do to the drive, there is a way of finding out.

Remember the sustained write figures from earlier? Seagate guarantees that you’ll be able to run at those sustained write rates for a period of 5 years before the drive dies.

I’ve done the math below:

Drive Sustained 4KB Random Write IOPS KB Written per Second # of Seconds in 5 Years (1825 days) Total Writes in 5 Years
200GB 10,500 42000 157,680,000 seconds 6167 TB
100GB 5300 21200 3113 TB
50GB 2600 10400 1527 TB

 

At 1527TB (or 1.49 petabytes) of random 4KB writes in a 5 year period, the 50GB Pulsar has a worse lifespan than Intel’s X25-E (the 64GB drive is specced at 2 petabytes of lifetime random writes).

The 200GB drive however can allegedly withstand over 6100TB (6 petabytes) of random writes before it goes to SSD heaven.

With no standardized method of testing or verifying any of these claims, we can’t do much more than take Seagate’s word for it. Given Seagate’s reputation in the industry, I’d expect that its word is worth it.

Final Words

There’s little more to say about Pulsar than what I’ve already mentioned. Seagate is finally getting in the game and, at least on paper, Pulsar appears to be in the class of Intel’s X25-E.

If Seagate can deliver on its promises, this should be good. Currently the enterprise SSD market is occupied by Intel, Samsung and a bunch of other contenders with unproven track records. Using Indilinx drives in a server environment is very tempting, but also potentially dangerous. There isn’t sufficient data out there today to say how long their drives will last under constant random writes. I tend to believe Seagate when they commit to Pulsar being a 5 year drive capable of over 6PB of random writes.

Pricing is another major unknown, but I’d expect the drives to be no cheaper than Intel’s X25-E: at least $11 per GB.

Even more exciting is the promise of consumer level SSDs from Seagate, possibly as early as 2010. Western Digital is next, whom I expect to join the race in 2010 as well. Once all of the major HDD players are in the race, we’ll see some real competition in both the high end and mainstream client SSD markets.

Seagate Goes Light on the Details
POST A COMMENT

52 Comments

View All Comments

  • araczynski - Tuesday, December 8, 2009 - link

    SSD revolution?? someone forgot to wake me up...

    good to see seagate putting in some rebranded hardware into the mix, should help the prices drop a little faster, to the levels of mass market acceptance....

    was just throwing out some old purchase order records....couldn't believe i (well, the company) paid $178 for a dvd+-rw drive back in the day... so maybe in 5 years the HD's will be the floppy disks.
    Reply
  • Kougar - Tuesday, December 8, 2009 - link

    Bright Side of News is indicating the Pulsar drives will utilize "up to" 16 channels per drive, more than Intel's 10.

    http://www.brightsideofnews.com/news/2009/12/8/sea...">http://www.brightsideofnews.com/news/20...aunches-...
    Reply
  • Casper42 - Tuesday, December 8, 2009 - link

    Being that the 200GB (256GB) model here uses 8 channels, that would imply the 16 channel model would be 512GB (or 400GB according to Seagate). Reply
  • PandaBear - Tuesday, December 8, 2009 - link

    that Seagate and Micron has worked together in a secret JV (joint venture) with a small team of engineers since about 1 to 1.5 years ago.

    The focus of the JV is to put a programmable analog amplifier (the one called read channel in the hard drive industry) inside the flash to improve its voltage sensing performance. HD has these sort of analog components like pre-amp (the amplifier on the actuator, because head signal is too weak to reach the controller by itself) and read channel (programmable A2D converter and decoder) for quite some time, and flash is starting to see problems that can only be fixed by analog mean.

    So it seems like Seagate's flash may have a special way of adjusting the memory's parameter via vendor unique commands that aren't available to others (except Micron), and they use that to their advantage in their own designed controller (with 3rd party IP, but everyone has 3rd party IP for their SATA interface, ARM processor, embedded memory controller, etc, so it is not saying much.
    Reply
  • Roland00 - Tuesday, December 8, 2009 - link

    It appears there is a new controller on the block with the Phison Controller found in the Patriot PS-100.

    Specs on the box make the drive look like it is Indilinx comparable and I have seen this drive priced at 2 dollars per gigabyte after rebate.
    Reply
  • SleepyGreg - Tuesday, December 8, 2009 - link

    $11 per GB! Is that a typo? That would make the 200GB model over $2000! Reply
  • crimson117 - Tuesday, December 8, 2009 - link

    Remember, assumption makes an ass out of u and mption. Reply
  • arnavvdesai - Tuesday, December 8, 2009 - link

    At a cost of 11$/GB & a reduction to 3W from 7W, are these drives cost effective. Yes I understand you are speed restricted but is the difference in cost recovered by the difference in speed & power consumption. Also, I am curious whether the higher speeds help reduce the number of servers we might need especially larger data centers considering that we can match capacities. Would having higher drive speeds cause a unit of 'work' be executed fast enough that we would need less machines than using standard mangetic drives. Reply
  • Mr Perfect - Tuesday, December 8, 2009 - link

    It would depend on everyone's individual case, I'd think.

    First you'd need to know how many drives you have in each of your servers and then multiply that by 5 watts(assuming your drives are 7 watt, andthe SSDs you buy are 3 watt). Then you'd have to take that per-server power savings and multiply it by how many servers you had. At that point you'd know how many fewer watts you'd be pulling off of the PSUs, but you'd still have to figure how many fewer watts those PSUs will pull form the wall(are they 75% efficient? 80%? 85%?). When you get the drop in power from the wall, you could multiply it out by whatever your local power utility is charging you per KWH. Of course you'd also want to figure out how much money you'd save by not having to cool all of that extra power dissipation too.

    Quick, to the beancounterarium! :D

    Reply
  • Kibbles - Tuesday, December 8, 2009 - link

    "Such low write speeds are either artificial or the result of the same sort of issue that continues to plague Intel’s X25-M."

    Is that supposed to be Seagate? Since it says the 50GB, and Intel only sells one size.
    Reply

Log in

Don't have an account? Sign up now