Withdrawals and my addiction to speed

It’s 2:37am as I’m writing this blog. No, I’m not under the influence of some illicit drug that’s keeping me up all night, just suffering from jet lag as I’m adjusting to the local time here in Shanghai where Intel is holding IDF this week.

In my last blog, I shared my frustrations with the experience of my IT laptop and I mentioned that I have been particularly frustrated lately. So why am I particularly frustrated any more than usual, and why lately?

Flash back to a couple weeks ago. In order to collect some real-world usage information and gain additional experience with the technology in an IT setting, I played the part of Guinea Pig and had one of our pre-production solid state drives (SSD) installed in my IT laptop (my IT guys will flip when they read this).

Although I was quite familiar with its capabilities from all the performance characterization data, I was unprepared for the powerful instant high it gave my system. It was such a dramatic difference in how my system responded that I found myself uninhibited in doing things that I previously would have shied away from.

I no longer aborted our backup client software whenever it launched itself (which is still at the most inopportune time possible, like when you’re in the middle of a presentation, since the ability of the IT software to detect the most inconvenient times to do things still appears to work perfectly), nor did I need to go for coffee while it ran. I actually continued working while it was running. I even ran it intentionally a couple times for fun to prove to myself that I wasn’t just imagining the fact that it had no noticeable impact on my system responsiveness. I no longer launched IE to do some surfing while Outlook loaded up in the morning (which of course only makes Outlook load even more slowly). This threatened to mess up my entire routine, not to mention my relationships with colleagues who I had been going to coffee with each morning while our systems struggle to make themselves useful.

I quickly got used to this new way of doing things and after a short while I started to think that this was normal. Then the day came that my SSD was retrieved for data mining (that was the whole idea behind the trial) and my original hard-disk was put back into my laptop. There’s no way to feel the pain quite as intensely as having to go back – and that is why I’m particularly frustrated more than usual and why I’m so frustrated lately.

So a word of warning to those that might be considering dabbling with the use of our new SSD technology. It can give you quite a rush, and once you have enjoyed its effects, it can be quite difficult to ever go back.

16 Responses to Withdrawals and my addiction to speed

  1. James says:

    Just wanted to say that I am hoping that Intel SSD’s are released soon, and that initial claims of performance (greater than 100MBps read/write) are correct.
    From what I understand, You guys are using MLC, so it will be interesting to pair off an Intel offering with something like an MTRON 7000 series drive.
    Good luck to you guys! And here’s wishing that competition drives down the price so that a 64GB SSD is feasible for the masses as their primary OS/app drive.

  2. PowerPaul says:

    Hi there,
    how fast was your test ssd?
    More than 60 MByte/s or even 100 MByte/s?
    Or was it just the faster access time in contrast to normal hdds?
    Thanks in advance!

  3. Corey says:

    I think the real news here is that Intel is formally launching a real SSD brand. After hearing about what is happening with PCM technology and Numonyx, I think its pretty clear that Intel will be forging an SSD market to eventually lead PCM SSDs into it in the years to come.
    Overall, very exciting technology, and I cannot wait for the day I replace my last system drive with an SSD, because noise and limited performance going from servers to workstations and especially down to laptops is becoming ridiculous. All that needs to happen now is for prices to drop. If that happens, every man, woman, child and their dog will have SSDs!
    High-capacity storage requirements is still another story, though…

  4. Kaushal says:

    Absolutely Knut. I too had a chance to be on the pilot. The speed gain was noticeable because I have a habit of keeping lots of windows and too many outlook folders open at any given time. Initially I had attributed this performance gain to my new PC, but when I had to go back to the hard drive, I could not believe that my brand new T61 was actually this slow without SSD. I agree, it will be extremely hard for anybody to go back to hard drives after having a taste of these awesome SSDs.

  5. Chas says:

    Knut,
    The speed issues you write about are encouraging. What about the loss of capacity that you experienced in going from HDD (say, 200 GB) to SSD (perhaps 64GB) ? If your laptop is anything like most of ours, the HDD is constantly running at over 90% capacity.
    Second, what can you say about SSDs in enterprise applications?

  6. Knut Grimsrud says:

    Thanks to everyone for your interest and comments. I’m back from IDF and trying to catch up with my backlog.
    A few of you asked some questions in your comments, so I’ll try to address them:
    – How fast is the SSD?
    In my IDF presentation I shared some relative performance figures, but we have not yet published absolute performance figures. I suspect that will be covered at the formal launch. In general, to get a complete performance picture, both the bandwidth and the latency (or access time) of the device must be considered. Although some streaming workloads benefit from high bandwidths, often responsiveness improvement comes from improved latencies.
    – What about the lower capacity?
    The Intel SSD will be available in a range of capacities from 32GB to 160GB in the 2.5” and 1.8” formfactors. This range of capacities covers a reasonable spectrum of usage, and in cases requiring very high capacities it might make sense to use an SSD for the high-speed primary storage and a mechanical hard drive to store bulk data that might be less performance sensitive. As the technology matures further, I expect that the capacities will scale to strike a balance with demand.
    For my corporate IT laptop, my usage is actually surprisingly modest as I don’t store personal media (pictures and music) on my corporate laptop. I just took a look at my drive and I’m only using 26.4GB of my 80GB drive. I’m running XP which is a little more compact than the latest operating environment, but my usage is probably not too far out of mainstream for a corporate laptop. For a personal laptop the usage model is obviously quite different and I expect the desired capacities to be higher than for corporate laptops.
    In my particular case, I have an 80GB hard disk in my brand new IT laptop, and swapping that with an 80GB SSD is not a capacity compromise for me (and I’m happy to make that swap).
    – What can you say about SSDs in enterprise applications?
    I think SSDs hold a lot of promise in enterprise applications. For enterprise configurations that balance the I/O performance with the rest of the system, it is not unusual to have an excessive number of disk drives in order to realize the necessary I/Os per second (IOPS) performance. With SSDs that have substantially higher IOPS capability, it will be feasible to construct balanced systems that are more space and power efficient using a smaller number of higher-performance devices.
    There was a presentation at the recent IDF made by Sun that highlighted the use of SSDs in their application. I think you can find the IDF presentations posted online somewhere.
    Knut

  7. Prasanto K Roy says:

    Doesn’t flash have a rather limited lifetime in terms of write cycles? How does the flash in the SSD handle the constant read-write activity of a hard disk? Does the expected lifetime (of the SSD) match a hard disk’s? Or is it limited by the number of write cycles, or something?

  8. Knut Grimsrud says:

    The question about the limited number of program cycles for NAND is perhaps one of the most commonly raised issues with NAND-based SSD, and I’m glad you asked about it.
    At the recent Intel Developer Forum in Shanghai I devoted a bit of time to this topic and it’s addressed in slides 8 through 16 in my presentation. My IDF presentation can be downloaded by going to http://www.intel.com/idf and clicking on the “Content Catalog” link under the “Conference Materials” heading. My presentation ID is “MASS001″ and it’s in the “Memory Applications and Solutions” track.
    In practice, for client usage models the Intel technology has attributes that does not stress the cycling capability of the underlying NAND. Two characteristics combine to provide ample margin for client applications: 1) Research into the write intensity that client usage models expose the drive to indicates that the write rates for client usage is quite modest (we designed for write rates substantially higher than the highest we measured in a fairly large-scale internal study), and 2) The efficiency of the Intel SSD algorithms result in low write amplification so the available NAND cycles translate well to overall SSD write endurance (see my presentation for details and how some traditional technologies do not exhibit this).
    For higher-stress usage models like server applications this is a bit more interesting since the applied sustained write intensity may be much higher. Our SSD technology targeted at that market segment is different than for the client application in order to account for the sustained high-intensity 24×7 server usage model.
    Knut

  9. Matt says:

    I was recently awarded a patent Patent #7,239,509 for a stackable modular computer case system. I would like to form a strategic partnership with Intel to help get it into production. Perhaps the biggest issue is motherboard design to suit a stand alone case.
    Would Intel be interested in adapting the motherboard to fit into the system?
    I would readily consider an exclusive deal if it would help get Intel on board.
    As there is now a greater emphasis on going green. This product could greatly reduce landfill waste and use of energy to manufacture parts that can be reused. It would also build value into Intels motherboards by offering the versatility of this system to its customers.

  10. battery says:

    […] 11.9 MB/s read speed, and a slower 6.7 MB/s write speed. This read speed puts it ahead of Seagate’s Pocket Drive, but behind Lexar’s Lightning Drive. It is also considerably faster than […]

  11. tant04 says:

    A disruptive technology starts out being too small, too slow, or too expensive for the mainstream user. It appeals to people with special needs in niche markets. It then improves more rapidly than the conventional technology, invading the mainstream market “from below.” Finally it displaces the mainstream product.

  12. o ring says:

    I suspect that will be covered at the formal launch. In general, to get a complete performance picture, both the bandwidth and the latency (or access time) of the device must be considered.