Are Cloud Based Memory Architectures the Next Big Thing?

by randomwalker on 3/17/2009, 7:06 AM with 6 comments
  • by lsb on 3/17/2009, 1:48 PM

    Here's 10 minutes of your life:

      1) RAM is a lot faster than disk, keep things there for speed.
  2) "RAM is the new disk."  "Or is disk the new RAM?"
  3) 5 ads for Java companies.
    Nowhere does it say that Facebook is running the world's largest memcached installation. Or, for that matter, that HN is all in memory.

  • by oomkiller on 3/18/2009, 4:30 AM

    I think this article is close, but misses an important player, SSDs. SSDs bring super-fast data rates that can be amplified by traditional techniques (striping etc), but work enough like disk systems so that applications don't have to be rewritten. I believe that we will see SSD speeds continue to increase, along with widespread adoption and price reduction. On the software side, I believe that many commonly used applications will begin to detect SSDs and (un)optimize their code for them. In my opinion, memory will always be the fastest, but SSD gives the reliability and backwards computability that will cause mass-adoption.

    I think we'll see (and have already seen) the early-adopters and power users start adopting SSDs at a high rate. Business adoption usually follows after a few years. So my prediction is that we'll see 1TB or larger sized SSDs, that are increasingly faster, within a few years. I think we'll start seeing DB arrays with multiple SSDs in a Raid0+1 or similar, probably up to 10TB or larger.

    Does anyone else feel this way.

  • by rgrieselhuber on 3/17/2009, 1:13 PM

    Enjoyed this post. Most of the in-memory solutions seemed focused around Java / .NET. What are the major open source (if any) solutions for Python / Rails / PHP?