Gear6 - Well Kept Secret (Not Yet)

LOL I wonder how much this

LOL I wonder how much this blog post cost Gear6. :)

Think: if what Gear6 offers is such a wonderful technological breakthrough, why it doesn't look like people are buying it?

Why Microsoft isn't buying it to boost software development productivity? Why Google isn't buying it in order to accelerate their search/YouTube server infrastructure? With their amount of traffic, they would surely benefit from the technology more than anyone else, right?

Short answer: smart folks don't buy crap.

They don't buy what can be built with commodity hw running linux. For a long time Google was running on a cluster of... Celerons and Pentium IIIs, no kidding. No big tin, no Itanium (see, e.g. http://query.nytimes.com/gst/fullpage.html?res=9D06E4DF1638F93AA1575AC0A9649C8B63&sec=&spon=&pagewanted=all), and certainly no CacheFX from Gear6 to help.

Did you hear anything about Engineered Intelligence? This is how Gear6 was called before (try www.engineeredintelligence.com). With a different team, but the same captain (Matt Oberdorfer) they tried selling CxC -- "a parallel programming language for scientists". Where is CxC now? Why did EI change the name? And why don't they even mention it on their pages?

In my opinion, CacheFX is CxC story all over again: trying to create an impression of breakthrough technology for idiots with fat wallets who will buy anything without checking what they are buying and whether money can be spent with a better return (think along the lines of hiring a knowledgeable consultant, who are abundant, and contracting him strictly on the cost-and-savings basis). And after even idiots are no longer buying, it's time to change company name and start again with another fad.

Gear6 can pay for blog posts and for articles praising CacheFX, no problem here, and some of above-mentioned idiots with fat wallets might even buy them.

The problem is winning customers like Google who would allegedly benefit from the technology, but who undoubtedly have well-educated people in charge of their server infrastructure, and to anyone with a brain that should serve as a good indicator as to what Gear6's technology is all about.

I hear you brother. But hear

I hear you brother. But hear me out - there is more than one way to skin a cat. Yes, you can go out and get a high performance NAS system and that is a road I encourage many to go down. But the majority of NAS systems deployed out in the field are none of the products you mentioned. I am a big fan of each of them but what about the common man? What about the bread and butter NAS environments? What about the gazillions of dollars already invested in existing infrastructure?

I agree - putting in a hunk of stupid memory or SSD isn't the way to do things. But putting in a cache coherent, scalable system that has intelligent caching algorithms and can support any number of NAS devices behind it is pretty cool. And we've found that 60% or more I/O hits are out of cache - so more cache helps the performance problem. I am not sure what you mean by saying - give you IOPS - IOPS are served out of cache as well as disk. And the Gear6 guys are saying they will give you a really big cache so you can keep more data in memory to immediately serve up I/O. They will not only provide you with intelligent network caching for cache friendly I/Os but also reduce the need to perform reads from disk. That is very cool as long as its cost effective. And remember - memory is 1,000 times faster than disk access.

But we can debate this forever. Let's see how well they do in the market. As Edison said - it is all about the utility of something. If it can't sell then it ain't worth sh*t. Well, I'm paraphrasing what Edison said.

Naw, these caching devices

Naw, these caching devices never fly, Tony. This is not rocket science here - many people have done it in various fashions or another. So it works with NFS only - so there are alot of high-perf NFS solutions that are massively scalable and yet still affordable. NetApp's GX (spinnaker) cluster, Isilon, Panasas, Ibrix, Polyserve (acquired by HP) all have super scalable, high performing NFS solutions. Most of these are highly scalable because of the nature of their deployments - high sequential workload environments. Mind you, those workloads are junk for heavy cache systems. Give me IOPS baby! Any other NFS environment is arguably low performance requirements. I know some folks needing a little more punch out of their NAS environments are deploying VIFs and 10Gb fiber these days.

respectfully,
anonymous