A brief comparison between Compellent and XIV

Can you cite any specific reason(s) for not considering EMC or HP in the search for a storage solution? Are you an IBM installed base customer?

David,
thank you for your comment,
I wrote this post to compare XIV and Compellent (two next generation storage systems) because I’m working on this competition analysis in these days.

BTW, you will find a lot of comparisons about Next and Old generation storage in this blog, one example here: http://www.cinetica.it/2010/02/09/your-new-array-isnt-better-than-my-old-array/

ciao,
Enrico

A ggod article. I know people tend to have their favourites and I am no exception having been delighted with our Compellent solution, but Compellent is clearly delivery innovation, operational advantages and a platform which I feel we will be able to work with for a long time to come.

I would mark this 5-0 for XIV and here is why:
Performance: XIV utilizes its cache very efficiently and will out perform FC arrays on the market today. If XIV did use your calculations for performance there would be no way XIV would sell into tier 1 space. XIV has many customer references stating that they replaced tier 1 DMX, Hitachi, 3Par… for performance reasons.

Scalability: XIV starts at a 6 module (27TB usable) unit and scales to 15 modules (161TB usable).

Management: I have personally worked with HP, EMC, IBM, and Compellant arrays and XIV is by FAR the easiest to manage and most intuitive. It takes around 30 minutes to get someone to be able to do >95% of the XIV management. The other <5% being data migration and replication.

TCO: Once you factor in space, heating and cooling efficiencies, personnel… XIV comes ahead. Once again, many customer testimonials.

TCA: This all comes down to how good of a negotiator you are.

One thing that I didn’t see on this list is resiliency. XIV can rebuild a 1TB in 30 minutes or less. What happens when hard drives becomes 4 or 8TB, are you willing to wait 1 or 2 weeks to rebuild a drive.

I see that you are very biased and you can probably tell that I am as well, but it looks like you don’t have all of your facts regarding XIV.

I would mark this 5-0 for XIV and here is why:
Performance: XIV utilizes its cache very efficiently and will out perform FC arrays on the market today. If XIV did use your calculations for performance there would be no way XIV would sell into tier 1 space. XIV has many customer references stating that they replaced tier 1 DMX, Hitachi, 3Par… for performance reasons.

Scalability: XIV starts at a 6 module (27TB usable) unit and scales to 15 modules (161TB usable).

Management: I have personally worked with HP, EMC, IBM, and Compellant arrays and XIV is by FAR the easiest to manage and most intuitive. It takes around 30 minutes to get someone to be able to do >95% of the XIV management. The other <5% being data migration and replication.

TCO: Once you factor in space, heating and cooling efficiencies, personnel… XIV comes ahead. Once again, many customer testimonials.

TCA: This all comes down to how good of a negotiator you are.

One thing that I didn't see on this list is resiliency. XIV can rebuild a 1TB in 30 minutes or less. What happens when hard drives becomes 4 or 8TB, are you willing to wait 1 or 2 weeks to rebuild a drive.

I see that you are very biased and you can probably tell that I am as well, but it looks like you don't have all of your facts regarding XIV.

Juan,
Performance? please show us numbers of some 100% random workloads with big data sets. never seen nothing published yet.

Scalability? 160TB isn’t scalability, Compellent can scale up to 1008 drives per system! and soon it will scale more with the next generation controllers. Compellent is also working on federation (see Live volume!)

Management: have you ever seen Compellent management tools? If you need some video examples go here: http://connect.compellent.com/?elqPURLPage=134
The XIV management tool is easy to use but far away from Compellent Enterpise Manager and Replay Manager tools!

TCO: 80TB/160TB of usable space per rack, with UPS, and controllers needs more power and more space than same usable space from compellent. (and I can say the same thing about IOPS). fixed expansions aren’t good for TCO (with XIV, if you need two TB for a DB you need to buy a whole tray or more!!!).
on TCO we can add that if you buy now 1TB disks system you cannot add 2TB but you need to expand with old drives.
Compellent supports every new feature in every old model, XIV doesn’t support nothing new on old models (every generation differs from previous ones, if you buy now 1TB disk you can add 2TB disks, and so on). i.e.: On compellent you can add FCoE connectivity or SSDs to old models can you do the same in XIV?

TCA: I seen 85% discounts from IBM about XIV… desperate or failing list price?

Resiliency: 1TB in 30 Mins? probably you mean 30Mins to rebuild an empty disk! please show us the rebuild time of an array full of data. Compellent has the same rebuild policy of XIV (only allocated space and per volume). But Compellent can also store RAID1, RAID1 DP and RAID 6 blocks in the same disk granting your data from a double fail!!!

so , adding resiliency, we are 6-0 now.
thank you for your help,

Juan,
Performance? please show us numbers of some 100% random workloads with big data sets. never seen nothing published yet.

Scalability? 160TB isn’t scalability, Compellent can scale up to 1008 drives per system! and soon it will scale more with the next generation controllers. Compellent is also working on federation (see Live volume!)

Management: have you ever seen Compellent management tools? If you need some video examples go here: http://connect.compellent.com/?elqPURLPage=134
The XIV management tool is easy to use but far away from Compellent Enterpise Manager and Replay Manager tools!

TCO: 80TB/160TB of usable space per rack, with UPS, and controllers needs more power and more space than same usable space from compellent. (and I can say the same thing about IOPS). fixed expansions aren’t good for TCO (with XIV, if you need two TB for a DB you need to buy a whole tray or more!!!).
on TCO we can add that if you buy now 1TB disks system you cannot add 2TB but you need to expand with old drives.
Compellent supports every new feature in every old model, XIV doesn’t support nothing new on old models (every generation differs from previous ones, if you buy now 1TB disk you can add 2TB disks, and so on). i.e.: On compellent you can add FCoE connectivity or SSDs to old models can you do the same in XIV?

TCA: I seen 85% discounts from IBM about XIV… desperate or failing list price?

Resiliency: 1TB in 30 Mins? probably you mean 30Mins to rebuild an empty disk! please show us the rebuild time of an array full of data. Compellent has the same rebuild policy of XIV (only allocated space and per volume). But Compellent can also store RAID1, RAID1 DP and RAID 6 blocks in the same disk granting your data from a double fail!!!

so , adding resiliency, we are 6-0 now.
thank you for your help,

It is interesting that you state that you will ‘only represent the facts’. I’m not sure I’ve seen many facts form the XIV side of the discussion.

As the most obvious example, regarding the rebuilding of a 1TB drive, maybe you should view a demo. This is real world and a clear indication that you should do a little more research prior to submitting facts such as these and many other ‘assumptions’ above.

It is interesting that you state that you will ‘only represent the facts’. I’m not sure I’ve seen many facts form the XIV side of the discussion.

As the most obvious example, regarding the rebuilding of a 1TB drive, maybe you should view a demo. This is real world and a clear indication that you should do a little more research prior to submitting facts such as these and many other ‘assumptions’ above.

Adam,
I asked informations many times to italian IBM representatives, customers and on the internet without answers! Faith is not enough for me.

So I’m very glad that you have something more to say. Please share the links with the videos (with detailed configurations, XIV load during the rebuild, used space, and so on). Articles with the internals are very appreciated too.

At the moment I’m still waiting informations about IOPS on 100% random workloads with huge data sets ( you can follow the complete discussion on “Storage Consultants – UK & Europe” http://www.linkedin.com/groupItem?view=&gid=152703&type=member&item=31277707&qid=b1d436a6-a69a-48a1-8730-cecd5320121d&goback=%2Egmp_152703

if you have something to add about all my XIV doubts I’m here and ready to change my opinion about XIV when I will see the evidences.

thanks a lot in advance,
Enrico

Adam,
I asked informations many times to italian IBM representatives, customers and on the internet without answers! Faith is not enough for me.

So I’m very glad that you have something more to say. Please share the links with the videos (with detailed configurations, XIV load during the rebuild, used space, and so on). Articles with the internals are very appreciated too.

At the moment I’m still waiting informations about IOPS on 100% random workloads with huge data sets ( you can follow the complete discussion on “Storage Consultants – UK & Europe” http://www.linkedin.com/groupItem?view=&gid=152703&type=member&item=31277707&qid=b1d436a6-a69a-48a1-8730-cecd5320121d&goback=%2Egmp_152703

if you have something to add about all my XIV doubts I’m here and ready to change my opinion about XIV when I will see the evidences.

thanks a lot in advance,
Enrico

I was with an IT decision-maker yesterday and asked him if he ever looked at the overall health of the companies he considers when making acquisitions. He indicated he most definitely does. This has not been addressed here and probably should be. Comparisons/contrasts can be made about specs, performance, etc. but if the supposed “better” solution comes from a company that is financially suspect, they don’t really mean much. Check out the financial health of Compellent. It’s very suspect and that’s a fact. I’m pretty sure they’ve only had one profitable quarter, maybe none. I don’t even have to cite IBM’s financial strength. Bottom line…it’s one thing to acquire what’s perceived to be a better product, but when it comes to the service and support in the years to come, I’d want to work with a company that I know is going to be around years down the road,

I was with an IT decision-maker yesterday and asked him if he ever looked at the overall health of the companies he considers when making acquisitions. He indicated he most definitely does. This has not been addressed here and probably should be. Comparisons/contrasts can be made about specs, performance, etc. but if the supposed “better” solution comes from a company that is financially suspect, they don’t really mean much. Check out the financial health of Compellent. It’s very suspect and that’s a fact. I’m pretty sure they’ve only had one profitable quarter, maybe none. I don’t even have to cite IBM’s financial strength. Bottom line…it’s one thing to acquire what’s perceived to be a better product, but when it comes to the service and support in the years to come, I’d want to work with a company that I know is going to be around years down the road,

Ed,
I agree completely with you about the risk of buying from financial unstable companies.

But this is not the case of Compellent, and you are totally misinformed about their current situation (or you are trying to do some FUD, ).

First of all they are posting something like 15 or 16 sequential Qs of positive growth, they have more than120M$ in cash, they are growing very fast in the storage market, they are making money. What else?

it’s not difficult to look for $CML there are a lot of sites that can help you in this: http://finance.yahoo.com/q?s=CML is the best for me (full of links)

And here you are the last Q report: http://finance.yahoo.com/news/Compellent-Reports-First-iw-921754671.html?x=0&.v=1

My point of view is that IBM can’t match Compellent in features and function so they try some FUD… but it doesn’t work because Compellent is traded on NYSE and all financials are well known.

Ed,
I agree completely with you about the risk of buying from financial unstable companies.

But this is not the case of Compellent, and you are totally misinformed about their current situation (or you are trying to do some FUD, ).

First of all they are posting something like 15 or 16 sequential Qs of positive growth, they have more than120M$ in cash, they are growing very fast in the storage market, they are making money. What else?

it’s not difficult to look for $CML there are a lot of sites that can help you in this: http://finance.yahoo.com/q?s=CML is the best for me (full of links)

And here you are the last Q report: http://finance.yahoo.com/news/Compellent-Reports-First-iw-921754671.html?x=0&.v=1

My point of view is that IBM can’t match Compellent in features and function so they try some FUD… but it doesn’t work because Compellent is traded on NYSE and all financials are well known.

Well I guess I’m somewhat a bit late to the party. But I do agree with Enrico that the XIV is outperformed by Compellent and I think it is one of the dumbest solutions ever, I really dont believe in XIV atleast not in its current version. I think I would rather go for a regular SAN than going with an XIV solution. And I guess that Ed’s doubts about Compellents economical issues is nothing to worry about anymore since Dell bought them. I was actually hoping for IBM to buy Compellent.

Regards
Niklas

It is true that there is a lot to consider about the companies that represent these products and Enrico makes some interesting points as far as the technology goes but one thing that caught my attention is the fact that Dell and HP fought very hard for the acquisition of 3Par after IBM bought XIV. It seemed like they were trying to play catch up in the Storage vitualization space after they saw IBM was having some success. It makes me a bit suspect that only after the bid war for 3Par, where Dell lost, they then looked at Compellent and settled on that acquisition. If I were buying right now I would probably be looking at 3Par vs. XIV as opposed to Compellent vs. XIV. IBM, HP and Dell all have agressive acquisition strategies, IBM probably being the most agressive, why would they fight over 3Par so hard if Compellent is better and for that matter why would HP and Dell go after 3Par unless they were trying to play catch up with IBM and XIV?

On the technical aspect of the discussion I’m not as convinced with Compellent or even 3Par since I would consider them somewhat legacy architecture due to the fact that they still rely on 2 controllers only. I would really compare them to NetApps where there is a traditional storage architecture with a lot of SW options that are very good. They are all still concentrated on RAID Arrays, Tiering and reliability/resiliency is solely relying on dual controllers. I’m not a big fan of tiering and where I see XIV being very different is in the fact that they have made storage dummy proof…easy to manage with massive amounts of space, no need to worry about tiering and the management headache that goes along with that complexity.

The truth is that only if you have very demanding and high IOPS intensive applications should you look at tiering and QoS, most companies would fly well under the radar with 15K IOPS.

I also question the fact that the comparison of scalability only mentions that Compellent can allow for 1008 disks as opposed to telling me what is the amount of TB’s I can have on a full box. I don’t think most people are worried about the amount of disks, they are probably more interested in the amount of usable storage.

I also question the power consumption analysis…seems to me that 1008 disks would consume more power than 180 disks.

Don’t mean to sound biased but these are some of the things that jumped out at me as not being to clear in the analysis. The truth is that I’m not very convinced especially after considering that it is a dual controller based system that will no doubt have some kind of bottleneck when capacity is added.

I’m late as well. @Eric…Tiered storage is based on pay as you grow.  Why would you want 80% of your storage whiz is barely used on expensive disk.  Allow it to migrate down to cheap disk.  Also the higher number of spindles (as Compellent stripes across all disks) will increase IOPS.  180 TB is very limited.  1008 disks is at the very least 1.4 PB.  Having the ability to share a tray with a combination of SSD, 15k, 10k, 7k all at the same time is very valuable, efficient and extremely cost effective.  Know throw in Fast Track which writes to the outermost portion of the platter and know you have increased efficiencies by up to 30%.  So running everything at 15k is a waste of capital.  I would much rather put the money in the hardware and software of the SAN than in disk and so do most companies demo the Compellent solution.

As for why IBM and Dell were fighting for 3Par and not Compellent?  Compellent was smart and waited.  Why get in the middle and reduce the value of your company…it’s about strategy.  They waited and received the amount that was acceptable and from a company that had their best interest and future at heart…Dell.

More than a year is gone since this was published.
Both (both means Dell Compellent and IBM XIV) lacks of public & reliable performance benchmark.
Has anyone any idea how to get a real life independent report ?

haha, what a bunch of lies and FUD.

Always fun to read these types of entries. I’m a biased IBM marketing guy – sorry.  Thought I’d relay some experiences.

What I’m told is Compellent is limited on the performance side due to cache limitations.   XIV has boatloads of cache – but more importantly its distributed caching (localized prefetching) is the real secret sauce.  If you understand this concept, it is the lightbulb that helps you understand why XIV is not a drive dependent architecture.  The cache that is local to 12 Sata/NL SAS drives on XIV, prefetches only for those drives.  Instead of a funneled/centralized cache in pretty much all other systems, where prefetching is done for the entire system,  all prefetching is local. Net – almost all IOs are served via cache instead of disk on XIV, so drive speed is less relevant.  If you get this, you will understand that for most workloads even the older Gen2 XIV would likely outperform compellent… obviously workload and config dependent.  Benchmarks are weak indicators – need to do side by side workload compares or bake offs.  Gen3 XIV with the SSDs for caching would crush compellent.  You mention 100% random workloads – obviously not a real world request.

You are quite biased regarding your management piece and likely do not yet understand XIV, and have probably not seen a demo – XIV is far simpler to manage than compellent – but not because of the interface (which I would argue is superior on xiv)… here is big concept number 2 – Compellent’s GUI is fantastic – as is XIV’s… but a GUI is not what makes XIV simple.  XIV is raid-less, tierless, and tuneless…  Gone are the spreadsheets and RAID layout planning sessions. If you have managed an array before at a shop of any decent size, you will realize the implications of this.  Cutting LUN carving times is one thing (which both arrays do v. emc or hds), eliminating several operational processes that admins spend big time on, is quite another.  When you look at all the tasks storage teams do – XIV doesn’t simplify management – it nearly eliminates it.  That is why XIV shops state their TB managed to FTE numbers in PB managed to FTE, instead of TB.  One admin can easily administer several PB of disk across several arrays.

This leads to Scale – why do fortune 5 organizations run XIV when a single XIV can handle only 243TBu?
because managing 12 of them for enterprise workloads is much easier than
managing one vmax or vsp.  Is one compellent vs 2 or 3 xivs easier?
possibly. My point is that scale is a factor but in the xiv world
its not as big a factor – because other value areas outweigh this point.  Please do not dismiss the mangement thing. This is the single most important reason people buy XIV – you will only fully understand it if you see a demo from an IBM rep and if you have background managing SANs.

cost – depends on discounts, I would have assumed compellent was cheaper as there are fewer CPUs and less cache and both are commodity based systems.

You don’t mention reliability.  I believe compellent can cluster controllers, which is nice (similar to IBM’s midrange box, the v7000), but raid rebuild times (raid 6 on 2TB drive = 48+ hours vs 50 min on xiv) and less redundancy make xiv a winner in my (biased) book.

end of the day – which provides the most business value to your client for the lowest cost?  This is the only question that matters.  For a shop managing 10TB or less, compellent could very well be a winner, but in a shop of that size, they should be looking at v7000 from IBM.  Most of the time, however,  XIV takes the cake.  It also depends on workloads, pricing, and preferences.

A couple great things both systems share: fantastic snap technology, fantastic replication software (best in class in both these areas), and simple pricing models.  The license based models that Netapp and EMC are pushing are crazy.

Hope this helps.

first a disclaimer…I’m an IBMer
I got to this article by a chance. Since then XIV Gen3 is out. with 2 and 3TB disks as standard, infiniband inter-connectivity optional SSD as cache layer and many more improvement. Performance in some scenarios are even 4 times those of XIV Gen2…I believe the comparison result would be completely different today comparing compellent and XIV Gen3