I was sent a link to a chalk-talk by a customer the other day that I thought I would address point by point. Tegile recently announced our new T-Series all flash hybrid arrays, and it seems like everyone else is getting in on the game as well. I thought it would make sense to show just how impressive these new solutions are and leverage this video as a guide to some talking points.
A few points regarding the HP 3PAR All Flash Deduplication ChalkTalk
- Let Pure and EMC continue to battle over silly legal and who said what battles. It detracts from the fact that customers need a viable flash driven solution and why Tegile has enjoyed a quiet place out of the spotlight doing just that. Solving customer challenges with badass flash driven technology.
- ONLY then can deliver enterprise all flash? I’d argue that there are a lot of companies that are proving enterprise all flash.
- All flash at less than $2/gb? Tegile has an all flash array that is priced at $1/gb using a 5x deduplication factor. What factor is HP touting to drive to $2/gb? 4:1 or 10:1? Also note that Tegile uses eMLC instead of cMLC and a software patch. These eMLC drives provide 10x the write endurance of a cMLC drive.
- We can agree that pricing is the same as a legacy 15K spinning disk array. I like it when we can agree.
- ONLY they can deliver hardware accelerated inline deduplication? I don’t know if that’s similar to Lowes and Home Depot using a SKU that is off by one number to guarantee they have the lowest price because they are the ONLY one’s carrying that particular SKU even though the product in the box is identical, but Tegile provides inline deduplication leveraging flash hardware to deliver that technology without any performance penalty. Ever. We also don’t need a million dollar marketing gimmick to prove it. Just try one of our systems out for yourself.
- Adaptive sparing sounds like a way to work around the fact that they are using cMLC drives.
- The 20% more capacity they reference is likely the reserve cells leveraged in an SSD drive to accommodate cells that have failed. One of the reasons cMLC drives are more cost effective is because they have fewer of these reserve cells. Fewer reserves mean reaching the beginning performance degradation curve faster. Sure, you can create intelligent software to write to the drives in an efficient manner (Like Tegile and everyone else does) but you can’t overcome simple physics.
- Guaranteed 6 nines? That’s 31.5 seconds of downtime per year. That’s 2 and half minutes over the 5 year life of the array. And also note that you have to get 4 of their systems to qualify. Frankly, if you need that kind of uptime, then Tegile is not the right fit for that workload in your environment. Sure, I don’t go down. Sure, I have no single point of failure. Sure, I ship with a spares kit that includes an entire controller and both SSD and HDD drives so you can be back online in minutes. Sure, I have active/active controllers that will failover in millisecond. Sure, I go to extreme lengths to ensure that all of my data is committed to disk in way that preserves its integrity. Sure my engineering principles are integrity first, and performance second. But even with all of that, I think a guarantee that you’re not down more than 30 seconds every year is suspect. Better read the fine print. I can’t get a copy without talking to their sales or channel partner, so I’m not calling BS, just saying…better check and better understand what you end users are REALLY expecting from you. (http://www8.hp.com/us/en/products/data-storage/data-storage-products.html?compURI=1649739#.U5h1ExbVp4M) Is your network and compute infrastructure built to the same standard? Are all of your client endpoints built to that standard? Do you have similar guarantees from your ISP, the internet as a whole? National governments? The Sun? Mother Nature? There are so many external factors that may impact your availability beside just your array.
- Quick blurb on scale. I can build in blocks of 168tb of USABLE capacity. Now, this merits a definition. Usable to me means all data that can be addressed after accounting for all system overhead (RAID, Metadata, OS, etc). EFFECTIVE is a formula. I use USABLE * REDUCTION RATIO = EFFECTIVE. Marketing likes to use 5x because many of the other vendors do as well. So I will also. That said, I can provide 1.6PB of Effective Capacity in 10U which I would argue is Tier 1 enterprise scale.
- You got me. I don’t do synchronous replication today. I can do asynchronous however and use a 15 minute interval. Pretty dang close to real time. Also, if you have have a Tier 1 application that needs to be in an identical state at more than one site, do you really want to have your array do all of that work for you? Or do you want to use the applications own builtin tools (Like DAG for SQL/Exchange, Oracle RAC, etc)? Just asking.
- Warranty and Support is just maintenance. I do that along with everyone else, and I am typically a fraction of the cost of the legacy vendors and comparable Niche (using your terms) players as well. In fact, my latest system can support a 7 year maintenance term. Impressive by any comparison.
- Comparison time! Yay!
- Yep…predictable sub millisecond response. Got it.
- Equivalent pricing to all 15K arrays? Yep…in fact better.
- Enterprise scale? Yep. 1.6pb equals Enterprise according to them and in fact is bigger than their 1.3pb.
- Beat me on Zero RPO. But with any one of the multiple available 3rd party hardware/software products I can. You can use a small portion of all the money you save with Tegile to include that in your solution.
Compromise is a theme in this chalk-talk and we at Tegile agree that you should compromise nothing. Tegile all flash, and flash hybrid arrays provide more choices, more flexibility and usually at a lower cost. It’s worth taking a look to see what Tegile can do for you.