Sign in to follow this  
Kevin OBrien

NVELO Drives Storage Caching to the Masses with Dataplex

Recommended Posts

We've spent a lot of time talking about caching this week, from our overview of caching techniques to reviews of the OCZ Synapse and RevoDrive Hybrid. In both cases with the OCZ products, their caching magic is handled by NVELO's Dataplex software. Dataplex offers behavior-based algorithms to bridge the gap between a high-capacity hard drive or drive array and a high-performance SSD. As we've seen in the recent reviews, NVELO has done a brilliant job making the caching not only work in a near seamless way, but they've actually delivered on the promise of SSD speeds and hard drive capacity...without a big price tag.

Read Full Article

Share this post


Link to post
Share on other sites

Hi Kevin,

First of all, this is a much needed insight into a new burgeoning area. I am especially thrilled that you bothered to do a cost/benefit analysis - something most tech reviewers don't bother to include. Having said that, I found it hard to believe in the numbers that the synapse cached solution wasn't doing better after what I'd read. First, I could not replicate your "Price-Performance-GB" data. Secondly - what does that even *mean*?

I think I know what you're getting at though (accounting for all 3 variables to find best value) but I get very different results. I used 2 different methods using your data and both showed similar trends. I will use the Gaming Perf benchmark on the Synapse vs RevoDrive to demonstrate:

1) I calculate the same Price/GB as you have. Here, lower is better obviously (as it means cheaper/gb). Now if we take the performance figures (MB/s so higher is better), it would be best to divide it by the price/gb right? So with the resultant figure, a high performing but expensive solution is tempered and a low performing but cheap solution is bolstered. In Gaming, this gives me final numbers of (where higher is still better):

1654 - Synapse

1122 - Revo

234 - Vertex

194 - Momentus

These are actually in line with expectations when I examine the numbers - the Momentus is cheap but not that large and is punished by being *so* slow. The Vertex performs well but is punished by being most expensive *and* not that large. The Revo is expensive but is quite well compensated by being large and very fast. The Synapse has a great mix of everything.

2) Alternatively, you could say that the total 'firepower' of a drive is its performance (mb/s) *multiplied* by its capacity (gb). This accounts for the small and fast trade off of SSDs vs the slow and huge trade off of HDDs. Finally you would then divide this total 'firepower' by the price to give its 'value proposition' (cost per firepower). Here I get (again higher is better):

4660 - Synapse

2773 - Revo

1921 - Momentus

300 - Vertex

Sure the Momentus and the Vertex have swapped places but that's largely academic to me - just a different methodology giving different results.

Maybe I'm barking up the wrong tree but in that case I'd urge you to at least explain briefly how you got your final figures and what they mean.

Otherwise I urge you to carefully rethink what the last 3 lines of your table actually shows as to me this is the crux of the article and I'm not sure it shows what you may think it shows.

I'm genuinely not writing all this as criticism but because I really want to understand how valuable and cost effective this new tech is. You're welcome to PM me about it (if indeed you're able to on these forums). Otherwise, great write up.

Noli

Share this post


Link to post
Share on other sites

For our C/B analysis, we used the same metrics given by NVELO, but applied the equation to our trace results instead of the PCMark speed they used in some charts. I do agree that it is a bit odd to look at, but being an arbitrary metric I suppose you could show the results in many different ways.

Our results are calculated by taking the price and dividing it by the benchmark score, normalized to 1TB. Taking the Synapse price of $350 / 124.22 = 2.82 then dividing that by 2 to normalize it to 1TB.

The main purpose of the measurement was to compare the “storage value” of a product across all 3 dimensions of Price, Performance (MB/s), and Capacity (GB). In the past, SSDs and HDD were such different beasts that comparisons were near useless. SSDs could never compete against HDD in $/GB. And HDD could never compete against SSD in price-performance. So, no one ever needed to measure this. Enter caching... creating many interesting TCO equations to show its value ;)

Share this post


Link to post
Share on other sites

...but being an arbitrary metric I suppose you could show the results in many different ways.

Our results are calculated by taking the price and dividing it by the benchmark score, normalized to 1TB. Taking the Synapse price of $350 / 124.22 = 2.82 then dividing that by 2 to normalize it to 1TB.

Thanks for the time to reply. Ok I can now see what you did there. Obviously the lower the score the better. In which case I think you may have made a mistake in your calculations. Take a look again at the calculation for Gaming on the Synapse. $350 / 289.45 = 1.209 then divide that by 2 for the terabytes and you get 0.60 which is not the 1.21 that you publish (I think you forgot to divide by 2). Clearly 0.60 is much better and now makes the cached system surpass the RevoDrive for value in Gaming as well as the other two benchmark types.

[bTW technically I don't think it is an arbitrary metric - it specifically shows, and can be labelled: $ per MB/s per TB (lower is better) - but I don't really care about that as much as the value conclusion which I think is really important.]

Share this post


Link to post
Share on other sites

Yes, you are correct, we botched that last division on the gaming trace, thanks for catching that. Also made a small edit to chart to try to help it be more meaningful per your suggestion.

Share this post


Link to post
Share on other sites

Doh! You are completely right on that error. We just fixed that and will reword appropriately.

Thanks for the response and apologies if I sounded pushy. Whilst what they've created isn't completely revolutionary, it seems to me to be by far the best compromise of speed and size and cost for storage. As such, I'm really surprised it hasn't gotten more coverage in the tech review world / blogosphere, compared to all the yet-another-identical-sandforce2-based-SSD reviews! (not that they're pointless but I think we all know what to expect before reading those particular article).

Noli

Share this post


Link to post
Share on other sites

Not pushy at all, we needed to correct that.

As to the world at large, you're right, this solution isn't revolutionary in general, but at the client level it really is. You can't just walk into Best Buy or go to NE and get this kind of caching solution outside of the Synapse or RD Hybrid. If NVELO was public, I'd be a buyer.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this