This morning i saw the post from Marc Farley (3Par) i feel free to add something to his thoughts about 3Par and Compellent…
Well, we have a common “enemy”: EMC, that has poor features and now they are beginning to speak about risk of buying new and awesome technologies because who produce them isn’t a big company like EMC ;-).. sterile position, hope in better arguments in the next future!
Now EMC can’t ignore any longer Compellent since its installed base is growing very fast, the brand’s popularity grows, customers are happy (very happy), the TCO is one of the lowest of the market, upgrades are less expensive than competition, and so on!
The risk is not to buy Compellent, the risk is to spend more money in EMC!!!
But the real topic of this post is to answer Marc!
I repaet that 3Par and Compellent have several things in common: they have good and innovative products, the former targeted to big enterprises while the latter is targeted to the SME market but I find very difficult to agree with Marc upon his latest post for some reasons.
Very different architectures means different approches and results!
Compellent choosed to write a lot of software and use standard hardware, on the other hand 3Par choosed to design chips and to implement in hardware some critical features.
Standard hardware means the ability to invest a lot in R&D for software features and, indeed, Compellent has a lot of features and you can get new awesome features every year… and more, Compellent is very fast to adopt any new technology when avalilable because they don’t need to develop specific hardware but just buy it from someone! This ‘someone’ is a market leader with best in class products (Qlogic, Intel, Seagate, Stec, Xyratex, ecc., ecc.).
3Par needs to spend a lot in R&D to develop its own and unique architecture, it’s fast but closed and risky… if they get a buggy chip the entire production will be compromised and, with actual sales volumes, it can impact a lot to customers and revenues…. do you remember the intel xeon bug some years ago? or the first AMD opterons quad core bugs? and they produce a lot of chips!!!
The big difference in Compellent, compared to the rest of the world, is the Dynamic Block Architecture (DBA). Every page of block has some metadata with some important informations about the block itself (creation time, raid level, frequency of access, and others), this kind of informations are valuable to perform high level features (snapshots, data movements, raid reorganization, restriping, ecc, ecc)… probably a lot of more features will be released in the future to help customer to better manage their data.
Hard to say that 100 disks is a limit!
Compellent scale from 6 to more than 1000 disks… if you buy 3Par you can scale further in the same system but in a different way! But it’s insane in 2009 to compare systems from the number of disks and not for real performance/features/TCO capabilities!
if you need performance you don’t need 100s of disks you need a bounch of SSDs (oops! 3Par can’t use SSDs) so you get iops and low latency! I have seen some Compellent systems installed and i truly don’t think the limit is 100 disks… i personally installed systems with 50 disks and we reached, without any problem, the limit of the spindles while controller’s CPUs were taxed no more than 5%!!! i don’t think doubling the number of disks changes a lot this equation, Until now every disk added to a Compellent systems added iops and space as espected!
Data progression (Automated tiered storage), (i agree with you: the feature not invented but well implemented by Compellent) has some important benefits that help to gain more perfomance without adding 100s of disks!
First of all you can write data on SSD or in the external tracks of FCs disks in raid 10 (you can chose to write everything in raid 10 and get maximum performance from the system!) and than the less used data will be migrated in slow parts of the disk and, if necessary, the raid level will be changed!!!!!
Result? You always write in the fastest disks, in the best location available with the best raid level available for performance, then the system will move blocks to the more convenient places accordingly the usage of the data!!!! it’s not a matter of having 100s of disks it just depends on how smart is your system!!!
Less disks means: less floor space, less power consumption, less TCO!!!
Compellent has no dynamic optimization, Compellent has Automagically optimization!
its a question of TCO! When you add disks (SAT/SAS/FC/SATA) they will be logically placed to the appropriate tier and they will be used immediately! a restripe will be done and Data Progression let the new extra space and performance available to the hosts!!!
Simple as drinking a cold beer in a warm day!
To recap: yes, we are different and most of the times we are not in competiton, but the bar is not 100 disks! The only common factor is that we both are best of EMC.
I could keep on chattering but i’m just a little partner for Compellent, i’m not Compellent. My field experience with Compellent comes from winning against obsolete traditional storage vendors, and not from establishing corporate strategies.
I don’t think my non-vendor blog has the same importance of the vendor one… 😉 I am waiting for a prompt answer from Compellent, the same way you are!
ciao,
Enrico