HDS VSP: my point of view

What is the cost of an outage? There are business that cannot afford to be down and the “good enough” do not really work for them. This is one of the reason many large accounts have NTAP in fronmt and other vendors (many HDS) as their hardware raid system….VSP cheap some away for the front end too. By the way, do you know what is the entry cost of the VSP? It may surprise you.

What is the cost of an outage? There are business that cannot afford to be down and the “good enough” do not really work for them. This is one of the reason many large accounts have NTAP in fronmt and other vendors (many HDS) as their hardware raid system….VSP cheap some away for the front end too. By the way, do you know what is the entry cost of the VSP? It may surprise you.

[...] Yesterday, HDS announced the availability of HDT (Hitachi dynamic tiering) on their shiny new VSP. [...]

Roberto,
thank you very much for your comment.

NetApp storages are certified for a 99,999% of uptime, I don’t think we can say that NetApp is only “good enough”. I know Italian very large (happy) accounts migrated from old monolithic arrays to NetApp storages as their primary storage running their core Apps, VMs and so on.

About the cost of a VSP I can say very much, but I know very well the costs of USP-V/VM. I don’t think the costs will differ so much and the problem isn’t only the TCA but the support, software expansion (per TBs in the USP licensing model!!!) and so on.

NetApp has solution starting under 10K$ capable to deliver all software features of their biggest iron, can you say the same?

Roberto,
thank you very much for your comment.

NetApp storages are certified for a 99,999% of uptime, I don’t think we can say that NetApp is only “good enough”. I know Italian very large (happy) accounts migrated from old monolithic arrays to NetApp storages as their primary storage running their core Apps, VMs and so on.

About the cost of a VSP I can say very much, but I know very well the costs of USP-V/VM. I don’t think the costs will differ so much and the problem isn’t only the TCA but the support, software expansion (per TBs in the USP licensing model!!!) and so on.

NetApp has solution starting under 10K$ capable to deliver all software features of their biggest iron, can you say the same?

Hi Enrico – your recommendation of going with NetApp or Compellent leads me to believe that you really don’t understand the value proposition of the HP StorageWorks P9500 or the HDS version, the VSP. The P9500 is mission-critical storage – for customers that will turn it on once and never turn it off again (or have any downtime). We know that NetApp markets “mission-critical” for solutions like MetroCluster that don’t come close to mission critical (You can see my blog about it here: http://bit.ly/9rW8K1. Compellent is a midrange array – it also doesn’t have the availability and reliability features built into the P9500.

I agree with you – it isn’t for everybody. But customers that are driving their infrastructure as insanely close to zero downtime love this platform. I have several posts on my blog talking about the P9500: http://h30507.www3.hp.com/t5/Around-the-Storage-Block-Blog/bg-p/139/label-name/p9500.

Ciao,
Calvin Zito, @HPStorageGuy

Hi Enrico – your recommendation of going with NetApp or Compellent leads me to believe that you really don’t understand the value proposition of the HP StorageWorks P9500 or the HDS version, the VSP. The P9500 is mission-critical storage – for customers that will turn it on once and never turn it off again (or have any downtime). We know that NetApp markets “mission-critical” for solutions like MetroCluster that don’t come close to mission critical (You can see my blog about it here: http://bit.ly/9rW8K1. Compellent is a midrange array – it also doesn’t have the availability and reliability features built into the P9500.

I agree with you – it isn’t for everybody. But customers that are driving their infrastructure as insanely close to zero downtime love this platform. I have several posts on my blog talking about the P9500: http://h30507.www3.hp.com/t5/Around-the-Storage-Block-Blog/bg-p/139/label-name/p9500.

Ciao,
Calvin Zito, @HPStorageGuy

Calvin,

I know your point very well but NetApp says 99,999% of uptime ( http://www.netapp.com/us/products/protocols/fc-san/ ) and I don’t think that this number is quite different from the numbers shown by Hitachi.
Metrocluster customers are very happy (I manage a bunch of them) and they will never migrate back to old monolithic arrays!

Yes, Compellent is considered a Mid-range array but they are winning a lot of deals (also in italy, ) against monolithic offerings because theory and practice don’t match in the real world.

Customers are looking for next generation automated storages, integrated with VMware/Hyper-v, easy to use, scalable, cheap, with a low overall TCO: HP bought 3Par for these reasons!

The world is changing very fast: VMware rules, commodity HW rules, cheap-automated-easy-reliable storage rules.

Calvin,

I know your point very well but NetApp says 99,999% of uptime ( http://www.netapp.com/us/products/protocols/fc-san/ ) and I don’t think that this number is quite different from the numbers shown by Hitachi.
Metrocluster customers are very happy (I manage a bunch of them) and they will never migrate back to old monolithic arrays!

Yes, Compellent is considered a Mid-range array but they are winning a lot of deals (also in italy, ) against monolithic offerings because theory and practice don’t match in the real world.

Customers are looking for next generation automated storages, integrated with VMware/Hyper-v, easy to use, scalable, cheap, with a low overall TCO: HP bought 3Par for these reasons!

The world is changing very fast: VMware rules, commodity HW rules, cheap-automated-easy-reliable storage rules.

I think we agree, mostly but I’ll limit my reply to the point where we don’t agree. I really don’t care what NetApp says – there’s no way a two node system, geographically separated (leaving one node at each side of the cluster) is 99.999% available. Two failures and you’re down and there are many single points of failure on each side of the cluster. That just isn’t mission-critical by anyone’s definition except NetApp. Metrocluster is a poor man’s DR solution that leaves customers exposed to a second failure. It’s kind of like putting your kids on a bike, giving them a helmet and turning them loose on the streets of Rome. Personally, I’d rather they were in a Volvo or a Mercedes. Sure, it costs more but I value my kid’s life. Customers that value their data use platforms like the P9500 for exactly that reason.

I think we agree, mostly but I’ll limit my reply to the point where we don’t agree. I really don’t care what NetApp says – there’s no way a two node system, geographically separated (leaving one node at each side of the cluster) is 99.999% available. Two failures and you’re down and there are many single points of failure on each side of the cluster. That just isn’t mission-critical by anyone’s definition except NetApp. Metrocluster is a poor man’s DR solution that leaves customers exposed to a second failure. It’s kind of like putting your kids on a bike, giving them a helmet and turning them loose on the streets of Rome. Personally, I’d rather they were in a Volvo or a Mercedes. Sure, it costs more but I value my kid’s life. Customers that value their data use platforms like the P9500 for exactly that reason.

Calvin,
I like your example but I disagree.
Most of my NetApp metrocluster customers haven’t their storage in WAN but are in one or two near datacenters (not for DR). It’s a great solution if used in the right way.

Calvin,
I like your example but I disagree.
Most of my NetApp metrocluster customers haven’t their storage in WAN but are in one or two near datacenters (not for DR). It’s a great solution if used in the right way.

I like Calvin’s arguments for rock solid arrays and Enrico makes a good point too – put the less than rock solid arrays behind the VSP (or IBM’s SVC) and get rock solid availability from T2 arrays. With SVC you can create a logical mirror so your LUN is supported by two separate arrays eliminating that as a single point of failure. I’m not sure if VSP can do that or not…

I like Calvin’s arguments for rock solid arrays and Enrico makes a good point too – put the less than rock solid arrays behind the VSP (or IBM’s SVC) and get rock solid availability from T2 arrays. With SVC you can create a logical mirror so your LUN is supported by two separate arrays eliminating that as a single point of failure. I’m not sure if VSP can do that or not…

Enrico,
you cannot judge the costs based on something that came before and take in consideration what you can actually do with one system long term..Everyone is happy with the first NTAP, no so much with their 10th. Stating a 99.999% is meaningless; I’ve to arms thus I’m redundant….no, I will be crippled if I lose one and that’s what those systems are when a part fail and that is why so many enterprise arrays are actually installed behind their gateways.

About the “even more for their crappy software”, you must be offering to some old version or some very personal experience. All the Hitachi management software came a long way since it is first appeared (and infancy deficiency is in every product, not just Hitachi ones);today it is as good or better than what the major competitors offer and it is proved by the vase installed based that the Hitachi Software Products have……

Enrico,
you cannot judge the costs based on something that came before and take in consideration what you can actually do with one system long term..Everyone is happy with the first NTAP, no so much with their 10th. Stating a 99.999% is meaningless; I’ve to arms thus I’m redundant….no, I will be crippled if I lose one and that’s what those systems are when a part fail and that is why so many enterprise arrays are actually installed behind their gateways.

About the “even more for their crappy software”, you must be offering to some old version or some very personal experience. All the Hitachi management software came a long way since it is first appeared (and infancy deficiency is in every product, not just Hitachi ones);today it is as good or better than what the major competitors offer and it is proved by the vase installed based that the Hitachi Software Products have……

The VSP, like the USP V actually, can do that. If the LUN is externally virtualized, it can be in two separate DC….the VSP, likewise the P9500, can be two physical system linked together to appear as a single large system but, to all the effect, still separate harwdare modulaes with total control of teh same virtualized farm

The VSP, like the USP V actually, can do that. If the LUN is externally virtualized, it can be in two separate DC….the VSP, likewise the P9500, can be two physical system linked together to appear as a single large system but, to all the effect, still separate harwdare modulaes with total control of teh same virtualized farm

Roberto,

about TCA:
VSP is a costly system: starting price is 234K $ (http://www.channelregister.co.uk/2010/09/27/hitachi_vsp/) and it is not clear what you obtain with that price: disk space, # of controllers, cache, software options, etc.

Compellent and NetApp are cheap, scalable, with all-in-one software options. think to replica on Compellent and NetApp you have only a software option that does sync, async, snapshot copies, 1:N, 1:1, N:N, QOS, sync of LUNs on different protocols (Local lun in FC replicated via iSCSI without appliance on IP networks), and much more… in two words: cheap and flexible

about TCO:
cheap and flexible is a good start but I can add that CML/NetApp are virtualized by design: the disk pool/aggregates are the only (right) way of thinking. Thin provisioning, wide striping, unlimited snapshot per volume, ease of use, automated tiering are the standard! Live Volumes and data motion add a lot of flexibility…
They have new way to store data (not very old RAID levels and raid groups to manage)!!!!

integration with operating systems, hypervisor and even application software is the only way to go to save time and money! an example? with NetApp you can clone a SAP environment in minutes, or recover a single mailbox from exchange in seconds, (only to do 2 examples).

with Compellent and NetAPP it is not a problem to manage one or many boxes it is simple!

about crappy management software your company is very famous, I hope to see the new management software as soon as possible to change my ideas but I’m not sure it will match Enterprise manager from Compellent.

I know customers with PBs environments that are very happy with their Compellents and NetApps coming from old monolithic storage (most of them are published case studies), can you show me customers migrated from Compellent to USP?

We have a lot of success cases about customers migrated from old generation storages to next generation ones (here you are a case: http://www.cinetica.it/2010/02/04/the-best-space-guarantee-program/. ).
this is not from VSP (he is from AMS) but when he asked a new storage HDS tried to sell him a USP-VM. He bought Compellent for the ease of management, better features, costs, reliability, automated tiering (two year ago!).
Now they have 3 systems (one in USA, two in italy) and they can replicate the US one here in italy with deduplicated snapshots replicas with a 15mbit MPLS!!! can you do this with VSP?

Roberto,

about TCA:
VSP is a costly system: starting price is 234K $ (http://www.channelregister.co.uk/2010/09/27/hitachi_vsp/) and it is not clear what you obtain with that price: disk space, # of controllers, cache, software options, etc.

Compellent and NetApp are cheap, scalable, with all-in-one software options. think to replica on Compellent and NetApp you have only a software option that does sync, async, snapshot copies, 1:N, 1:1, N:N, QOS, sync of LUNs on different protocols (Local lun in FC replicated via iSCSI without appliance on IP networks), and much more… in two words: cheap and flexible

about TCO:
cheap and flexible is a good start but I can add that CML/NetApp are virtualized by design: the disk pool/aggregates are the only (right) way of thinking. Thin provisioning, wide striping, unlimited snapshot per volume, ease of use, automated tiering are the standard! Live Volumes and data motion add a lot of flexibility…
They have new way to store data (not very old RAID levels and raid groups to manage)!!!!

integration with operating systems, hypervisor and even application software is the only way to go to save time and money! an example? with NetApp you can clone a SAP environment in minutes, or recover a single mailbox from exchange in seconds, (only to do 2 examples).

with Compellent and NetAPP it is not a problem to manage one or many boxes it is simple!

about crappy management software your company is very famous, I hope to see the new management software as soon as possible to change my ideas but I’m not sure it will match Enterprise manager from Compellent.

I know customers with PBs environments that are very happy with their Compellents and NetApps coming from old monolithic storage (most of them are published case studies), can you show me customers migrated from Compellent to USP?

We have a lot of success cases about customers migrated from old generation storages to next generation ones (here you are a case: http://www.cinetica.it/2010/02/04/the-best-space-guarantee-program/. ).
this is not from VSP (he is from AMS) but when he asked a new storage HDS tried to sell him a USP-VM. He bought Compellent for the ease of management, better features, costs, reliability, automated tiering (two year ago!).
Now they have 3 systems (one in USA, two in italy) and they can replicate the US one here in italy with deduplicated snapshots replicas with a 15mbit MPLS!!! can you do this with VSP?

Steve-
Comepellent already has this feature on his arrays it is called Live Volume. http://www.compellent.com/Community/Blog/Posts/2010/5/CDrive-Live-Volume.aspx

and you don’t need SVC or other hardware on top of your arrays to manage.

Steve-
Comepellent already has this feature on his arrays it is called Live Volume. http://www.compellent.com/Community/Blog/Posts/2010/5/CDrive-Live-Volume.aspx

and you don’t need SVC or other hardware on top of your arrays to manage.

$234K are list prices, you should know the value of a list price. We can provide the configuration that goes along with that.

TCO: while I recognize that the licensing is somewhat simpler on the mentioned products (why don’t you include even easier to license ones is a mystery), the functionality offered is rather better on USP/VSP. What is the RTO/RPO on those devices? How long it takes to rebuild from snaps? Do you think that a large organization can afford a day to rebuild a corrupted DB? Those are the costs that you have to consider. A easy and cheap backup is meaningless w/o a safe restore. Why, 8mm were cheap tapes….except that you could never read them back….

Again, the organization that really care do not set with the good enough, and use serious enterprise class units behind the NTAP’s and similar

$234K are list prices, you should know the value of a list price. We can provide the configuration that goes along with that.

TCO: while I recognize that the licensing is somewhat simpler on the mentioned products (why don’t you include even easier to license ones is a mystery), the functionality offered is rather better on USP/VSP. What is the RTO/RPO on those devices? How long it takes to rebuild from snaps? Do you think that a large organization can afford a day to rebuild a corrupted DB? Those are the costs that you have to consider. A easy and cheap backup is meaningless w/o a safe restore. Why, 8mm were cheap tapes….except that you could never read them back….

Again, the organization that really care do not set with the good enough, and use serious enterprise class units behind the NTAP’s and similar

Roberto-

About 234K, I know that this is a list price but I’m pretty sure that it doesn’t take count about support. I’m very curious to know the list of hardware and software your company give away for 234K (HDP, HDT, Snapshots, clones and how many disks, controllers, cache, ports? ).

About functionality your are wrong, only to do an example CML and NTAP snapshots are pointers based, HDS snapshots are copy-on-write! Pointer based snapshots are more flexible, easy to use, and a better RPORTO for sure. Large companies prefer to have point and click recovery tools (like snap managers -ntap- or replay manager -Compellent-) instead of old style snapshots/cones.

E

Roberto-

About 234K, I know that this is a list price but I’m pretty sure that it doesn’t take count about support. I’m very curious to know the list of hardware and software your company give away for 234K (HDP, HDT, Snapshots, clones and how many disks, controllers, cache, ports? ).

About functionality your are wrong, only to do an example CML and NTAP snapshots are pointers based, HDS snapshots are copy-on-write! Pointer based snapshots are more flexible, easy to use, and a better RPORTO for sure. Large companies prefer to have point and click recovery tools (like snap managers -ntap- or replay manager -Compellent-) instead of old style snapshots/cones.

E

When I was 10 years old I remember watching Rocky IV where Rocky had a robot butler in

his mansion. This robot butler was amazing! — he knew when you wanted a Coke and went

and got it for you. This robot butler knew what you wanted, when you wanted it, and

made sure it was right there in front of you pretty much before you knew you even wanted

it!

Sadly, I grew up and realized that robot butler technology was a myth and was not going

to happen during my lifetime — at least not until after flying cars.

“Automatic Tiering” is that robot butler, and people that buy into it are fools. Have

any questions? Would be more than happy to talk you off the ledge.

Ryan,
thank for your post here.
And I replay with the same comment of the Linkedin disussion here: http://www.linkedin.com/groupItem?view=&gid=2523196&type=member&item=30815398&commentID=23899427&report%2Esuccess=8ULbKyXO6NDvmoK7o030UNOYGZKrvdhBhypZ_w8EpQrrQI-BBjkmxwkEOwBjLE28YyDIxcyEO7_TA_giuRN#commentID_23899427

Customers want AUTOMATED storage because their data are exploding. Automated tiered storage is one of the most important features for a modern storage like thin provisioning, wide striping, snapshots, hypervisors integration and ease of management.

IT departments can’t hire new people but they need to manage more and more storage every year. the only way to go is automation and ease of use.

Customers are looking for these features because they want to save time and money!

Ryan,
thank for your post here.
And I replay with the same comment of the Linkedin disussion here: http://www.linkedin.com/groupItem?view=&gid=2523196&type=member&item=30815398&commentID=23899427&report%2Esuccess=8ULbKyXO6NDvmoK7o030UNOYGZKrvdhBhypZ_w8EpQrrQI-BBjkmxwkEOwBjLE28YyDIxcyEO7_TA_giuRN#commentID_23899427

Customers want AUTOMATED storage because their data are exploding. Automated tiered storage is one of the most important features for a modern storage like thin provisioning, wide striping, snapshots, hypervisors integration and ease of management.

IT departments can’t hire new people but they need to manage more and more storage every year. the only way to go is automation and ease of use.

Customers are looking for these features because they want to save time and money!

Hi how can i find a standard LOM for a VSP hitachi ?