HP vs HDS

On Thu, 08 Apr 2004 14:05:15 -0600, Rob Young wrote:

In article , "R Klute"
writes:

Perhaps, but that is the difference. The EVA only has a total of 4 2Gb
fibre connections, 2GB of cache in each controller, and a max of 35TB
Raid-0 capacity. The HDS / HP XP can have up to 32 2Gb fibre or 64 1Gb
fibre, up to 128GB of cache, and 129TB of usable raid protected
capacity.


Yeah. Scale-up versus scale-out.

Price out that 9980 with 128 GBytes of RAM and 1024 drives. I'll match
the capacity and spend a lot less. As a benefit, the EVA config will
have multiple frames. A single frame isn't an advantage. It sits in one
spot in one datacenter.



To match the storage capacity you'd need 3 EVA's and you'd still be 20
short of matching the port capacity of the 9980 and 122GB's short on
cache. So to completely match the capacity you would need 61
more controller pairs just to match the caching abilities of the 9980,
and at least 10 to match the port capactiy. I think that alone should
break the wallet in pricing the two. HDS's caching is also far superior
than the EVA.

To even consider the EVA in the same market as the 9980 is ludicris. At
most I'd place the EVA in a entry level enterprise. The advantage of the
HDS 9980 in a single frame is clear and thats scalability, and performance
with out having to upgrade the crontrollers/frame. Also, the HDS has
enterprise level services with it, including Continuous Access and BCV's
which are tested and very reliable on the HDS frame. However, the EVA is
still lacking in that area by just releasing a version of Continuous
Access, and BCV on the EVA has yet to match the capabilities of the HDS.

However, the usuability and easy of the EVA far surpasses the HDS 9980.
The time spent on allocating LUN's is a fraction with the virtualization
provided by the EVA.

--
Jake Roersma

On Fri, 09 Apr 2004 10:40:49 -0600, Rob Young wrote:

In article m, Jake
Roersma writes:
On Thu, 08 Apr 2004 14:05:15 -0600, Rob Young wrote:

I think that alone should
break the wallet in pricing the two. HDS's caching is also far superior
than the EVA.

I've read a fully loaded 9980 lists for $8 million. Maybe that is $4
million too high (50% discount). At $4 million that is still more than 6
EVAs. Quite a bit more. What pricing are you assuming?

Regarding caching - spouting about superiorty - what features make it
superior? Why so much cache? Profit margins?


Pricing is different for everyone considering the massive markups that
the retail prices have. Most vendors are willing to slash prices 50% or
more as you suggested. Which at $4m US for 129TB there isn't that much
of a price difference for 3 fully loaded EVA's. You aslo have to
remember for some companies price is not an issue as long as the array
meets the SLA.

HDS uses a crossbar switched backplane which is what seperated HDS from it
competitors which are still using bus based backplanes. This also means
that there is no single point of failure on the backplane. Since this is
a patented technology I'm pretty sure no one else is using it. The link
below gives a break down of HP's XP512 architecture, which is the older
generation HDS 9900; the same but faster achitecture is used in the 9980.

http://www.hp.com/products1/storage/...ance/paper.pdf
http://news.zdnet.co.uk/hardware/0,3...2109893,00.htm


To even consider the EVA in the same market as the 9980
isludicris. At
most I'd place the EVA in a entry level enterprise. The advantage of
the HDS 9980 in a single frame is clear and thats scalability, and
performance with out having to upgrade the crontrollers/frame.

Depends. If you had a small red-hot database that does a ton of Random
IO (let's say 18 GBytes) it is trivial to carve a LUN out of an EVA
that would serve up several thousand Random IO. How would you do this
in HDS? And yes I have databases that small and hot.

I'm not trying to displace the EVA completely, everything has its place in
your situation the EVA may be the better of the two. If I _had_ to place
this on the HDS I'd most likely be using Open-E-CVS(1000MB) LDEV's spread
across 19 Raid Groups. which would give me the access to 76 disks. I may
even consider to lower the LDEV size and use 190 Open-E-CVS-CM(100MB)
spread across all of the Raid Groups giving me access to 760 disks. Which
should yeild better performance than the 165 disks in the EVA. Assuming
you are using some form of software striping (LVM, etc) you can stripe
accross all of these LDEV's to gain better performance. Of course if I
had an 18GB database with the requirements you do I'd load the whole thing
into memory and place it on the cheaper EVA disk.


But
speaking of performance, I've looked but can't find HDS he

http://www.storageperformance.org/results.html

Correct me if I'm wrong but isn't all the testing still done by the
vendor here? How does that make these numbers any more valid than any of
the other numbers published by vendors?

Also, that HDS box sits in one place. If water poured in on it, it
wouldn't work well at all.

Last time I check the EVA was one piece too. Unless you are refering to
the fact that you'd have 3 fully loaded EVA's in different locations,
which is something to keep in mind when purchasing. If your requirements
have three locations then yes maybe the EVA is a better choice. If your
worried about water getting on a $8m array than you should reconsider your
data center design and policies. Are you using Continuous Access for the
EVA? I'd be interested in hearing how well it works for you.

Also, the HDS has
enterprise level services with it, including Continuous Access and
BCV's which are tested and very reliable on the HDS frame. However,
the EVA is still lacking in that area by just releasing a version of
Continuous Access, and BCV on the EVA has yet to match the capabilities
of the HDS.

But EVA offers VSnap, SnapClone. Maybe not as full featured as HDS,
but not broken either.

Up until recently the Snap operations on the EVA were not atomic, but are
now which is a big step forward. But yes, overtime I think that HP will
incorporate more and more of the enterprise functionality into the EVA.
I'm still waiting for a working implemenation of a call home feature.
Its nice to have the disks shipped to you before they actually go bad.
PRS seems to be a tossed together half assed implementation of call home
hopefully they make it better.

--
Jake Roersma

In article m, Jake Roersma writes:
On Thu, 08 Apr 2004 14:05:15 -0600, Rob Young wrote:

In article , "R Klute"
writes:

Perhaps, but that is the difference. The EVA only has a total of 4 2Gb
fibre connections, 2GB of cache in each controller, and a max of 35TB
Raid-0 capacity. The HDS / HP XP can have up to 32 2Gb fibre or 64 1Gb
fibre, up to 128GB of cache, and 129TB of usable raid protected
capacity.


Yeah. Scale-up versus scale-out.

Price out that 9980 with 128 GBytes of RAM and 1024 drives. I'll match
the capacity and spend a lot less. As a benefit, the EVA config will
have multiple frames. A single frame isn't an advantage. It sits in one
spot in one datacenter.



To match the storage capacity you'd need 3 EVA's and you'd still be 20
short of matching the port capacity of the 9980 and 122GB's short on
cache.

Actually - I wouldn't use extension cabinets just single frames.
Each frame holds 168 drives. To match the capacity of the 9980
would require 6 EVAs.

So to completely match the capacity you would need 61
more controller pairs just to match the caching abilities of the 9980,
and at least 10 to match the port capactiy.

Yeah - ports won't match up. But at 24 ports for the 6 EVAs -
not bad.

I think that alone should
break the wallet in pricing the two. HDS's caching is also far superior
than the EVA.

I've read a fully loaded 9980 lists for $8 million. Maybe that
is $4 million too high (50% discount). At $4 million that is still
more than 6 EVAs. Quite a bit more. What pricing are you assuming?

Regarding caching - spouting about superiorty - what features
make it superior? Why so much cache? Profit margins?

To even consider the EVA in the same market as the 9980 is ludicris. At
most I'd place the EVA in a entry level enterprise. The advantage of the
HDS 9980 in a single frame is clear and thats scalability, and performance
with out having to upgrade the crontrollers/frame.

Depends. If you had a small red-hot database that does a ton
of Random IO (let's say 18 GBytes) it is trivial to carve a LUN
out of an EVA that would serve up several thousand Random IO.
How would you do this in HDS? And yes I have databases that small
and hot.

But speaking of performance, I've looked but can't find HDS he

http://www.storageperformance.org/results.html

Also, that HDS box sits in one place. If water poured in on it,
it wouldn't work well at all.

Also, the HDS has
enterprise level services with it, including Continuous Access and BCV's
which are tested and very reliable on the HDS frame. However, the EVA is
still lacking in that area by just releasing a version of Continuous
Access, and BCV on the EVA has yet to match the capabilities of the HDS.

But EVA offers VSnap, SnapClone. Maybe not as full featured as
HDS, but not broken either.

Rob

On Fri, 09 Apr 2004 13:09:47 -0600, Rob Young wrote:


Nice. But what does any of that have to do with cache? Tell me about
cache superiority and why so much is required.



Well the first document has a lot to do with cache. If you read pages
10-12 is describes the architecture and the caching alogrithms used by the
XP. By no means are you _required_ to have that much cache. For
starters the mere fact that the 9980 can handle more than 6 times the
amount of drives that the EVA can, it makes sense that it has ability to
handle that amount of cache. Of course it sure doesn't hurt to have more
depending on the application's placed on the array.

But
speaking of performance, I've looked but can't find HDS he

http://www.storageperformance.org/results.html

Correct me if I'm wrong but isn't all the testing still done by the
vendor here? How does that make these numbers any more valid than any
of the other numbers published by vendors?


Why attempt to diss the benchamrk? The benchmark isn't
flawed - evidence of that is wide industry participation and a total
lack of anything bad said about it.

I wasn't dissing I was asking. I've seen the site referenced and have
used the numbers a few times. Is it too much ask how valid it truely is? I
couldn't find on the site where the auditing process was described; I'll
have to take another look. I does seem odd that both HDS and HP have
failed to participate with the 9900 series arrays, considering there are a
number of people that seem to swear by them. Even more odd is the lack of
HDS participation at all.

Jake

In article m, Jake Roersma writes:
On Fri, 09 Apr 2004 10:40:49 -0600, Rob Young wrote:

In article m, Jake
Roersma writes:
On Thu, 08 Apr 2004 14:05:15 -0600, Rob Young wrote:

I think that alone should
break the wallet in pricing the two. HDS's caching is also far superior
than the EVA.

I've read a fully loaded 9980 lists for $8 million. Maybe that is $4
million too high (50% discount). At $4 million that is still more than 6
EVAs. Quite a bit more. What pricing are you assuming?

Regarding caching - spouting about superiorty - what features make it
superior? Why so much cache? Profit margins?


Pricing is different for everyone considering the massive markups that
the retail prices have. Most vendors are willing to slash prices 50% or
more as you suggested. Which at $4m US for 129TB there isn't that much
of a price difference for 3 fully loaded EVA's.

3 fully loaded EVAs would go for about $1.5 million toss in a
few hundred thousand for software.

You aslo have to
remember for some companies price is not an issue as long as the array
meets the SLA.


Right. Operative word is some. In most cases, price is a big
point in the Decision Matrix.


HDS uses a crossbar switched backplane which is what seperated HDS from it
competitors which are still using bus based backplanes. This also means
that there is no single point of failure on the backplane. Since this is
a patented technology I'm pretty sure no one else is using it. The link
below gives a break down of HP's XP512 architecture, which is the older
generation HDS 9900; the same but faster achitecture is used in the 9980.

http://www.hp.com/products1/storage/...ance/paper.pdf
http://news.zdnet.co.uk/hardware/0,3...2109893,00.htm


Nice. But what does any of that have to do with cache? Tell
me about cache superiority and why so much is required.


But
speaking of performance, I've looked but can't find HDS he

http://www.storageperformance.org/results.html

Correct me if I'm wrong but isn't all the testing still done by the
vendor here? How does that make these numbers any more valid than any of
the other numbers published by vendors?


Auditing.

Auditor

Storage Performance Council www.storageperformance.org
Walter E. Baker
1060 El Camino Real, Suite F
Redwood City, CA 94063
Phone: (650) 556-9384
FAX: (650) 556-9385

Why attempt to diss the benchamrk? The benchmark isn't
flawed - evidence of that is wide industry participation and
a total lack of anything bad said about it.


in different locations,
which is something to keep in mind when purchasing. If your requirements
have three locations then yes maybe the EVA is a better choice. If your
worried about water getting on a $8m array than you should reconsider your
data center design and policies.

That was an example. Data Centers worldwide take hits every day.

Rob

"Jake Roersma" wrote in message
s.com...
On Fri, 09 Apr 2004 13:09:47 -0600, Rob Young wrote:


Nice. But what does any of that have to do with cache? Tell me about
cache superiority and why so much is required.



Well the first document has a lot to do with cache. If you read pages
10-12 is describes the architecture and the caching alogrithms used by the
XP.

There's nothing there that isn't dirt-standard these days: I'd be extremely
surprised if what's described held any advantage whatsoever over most
current mid-range-to-high-end storage boxes (Symmetrix didn't replicate
cached dirty data at least at one time, but that was known to be a
deficiency a decade or more ago, and other vendors which did it right
knocked Symmetrix for it).

- bill

Faeandar wrote in message

And the performance is insane, about the only other array that
comapres is the new EMC DMX. Of course I hate EMC so....

~F

Actually, there is at least one array that exceeds the HDS/EMC/HP-EVA
systems substantially.

You should take a look at 3par (http://www.3par.com) instead/as well.

Performance (over 100,000 on SPC-1 for a single 3par system)
vs EVA5000 is around 20,000. HDS has no official result for 9580.
The 9980 (higher end than 9580) is significantly
less than 67,000 on SPC-1 (Sun had to withdraw an SPC-1 result
of around 67,000 on the 9980 because of a bug that caused them to
run the test over a much smaller volume size than they claimed;
if they used a reasonable volume size their results are guaranteed
to be significanlty lower). EMC has not participated either. I
have no reason to give EMC DMX the benefit of the doubt especially
since I've heard that they barely equal the HDS
9980 in benchmarks some customers have run).

3par excells in other characteristics as well:

scalabilty (16 to 2560 disks for a single 3par system),
capacity (over 300 TB), port count (up to 192 ports),
density (40 disks per 4U).
Redundant everything.

Pricing per disk seems comparable to EVA (see SPC-1 summary)

And 3par's software offering is pretty good as well.
Virtualization, provisioning and ease of use are excellent.
They can stripe even small volumes across a large number of disks.
They have a whole suite of advanced software including
- advanced snapshot copies
- 100s of snapshot copies per volume
- read-only as well as read-write snapshots
- can take snapshots of snapshots
- full copy with fast re-sync
- remote copy over GigE
- thin provisioning (ability to provision more virtual space than
physical space with automatic provisioning)
and a bunch more.

-nw

In article m, Jake Roersma writes:
On Fri, 09 Apr 2004 13:09:47 -0600, Rob Young wrote:

Why attempt to diss the benchamrk? The benchmark isn't
flawed - evidence of that is wide industry participation and a total
lack of anything bad said about it.

I wasn't dissing I was asking. I've seen the site referenced and have
used the numbers a few times. Is it too much ask how valid it truely is? I
couldn't find on the site where the auditing process was described; I'll
have to take another look. I does seem odd that both HDS and HP have
failed to participate with the 9900 series arrays, considering there are a
number of people that seem to swear by them.


The oddness is quite expected. After all, HDS
and EMC work within a boutique pricing model. Ever find published
list prices down to the component level? You'll notice the SPC
has a $/SPC metric. In order for that to be calculated, you actually
have to post the prices - shocker!

Even more odd is the lack of
HDS participation at all.

Nope. They'd actually have to tell the unwashed masses what
their list prices are. If that information got around, the unwashed
masses would suddenly be empowered. Can't let that happen - now
can we?

The folks at Storage Magazine have pleaded off and on about DMX being
benchmarked. You need a subscription to read this in full:

http://storagemagazine.techtarget.co...884731,00.html

Editorial: Here we go again
by: Mark Schlack
Issue: Mar 2003

Last month and this month, we've been guilty of something I hope you never have
to do: evaluating vendors' performance claims in the abstract. In our February
cover story ("Inside the new Symmetrix") and this month's lead Trends story,
("Symmetrix DMX: Is it hot or not?") we've written a lot about the putative
performance advantages of EMC's new Symmetrix DMX architecture, and how it
compares to Hitachi Data Systems' Lightning and IBM's Shark.

We did that without ever reading or writing a single bit to any of those
devices. Nor did we report any data from anyone who did, outside of the vendors
themselves. It's the sad truth, but a critical review of architectures on paper
is the best you can do right now.

Although there's a vendor-neutral performance benchmark (the Storage
Performance Council's SPC-1), only IBM--of those three vendors--has seen fit,
as of this writing, to report results. Perhaps HDS will. EMC hasn't even joined
the Storage Performance Council.

---

They'll probably be a run on ice skates in Hades before you see
a DMX or a 9980 benched for SPC numbers.

Rob

On Fri, 09 Apr 2004 14:01:56 GMT, Jake Roersma
wrote:



To match the storage capacity you'd need 3 EVA's and you'd still be 20
short of matching the port capacity of the 9980 and 122GB's short on
cache. So to completely match the capacity you would need 61
more controller pairs just to match the caching abilities of the 9980,
and at least 10 to match the port capactiy. I think that alone should
break the wallet in pricing the two. HDS's caching is also far superior
than the EVA.

Why are you worrying about cache as if it's the measure of performance?
It's certainly not how MY customers measure performance.

The virtual configuration of the underlying hardware on the EVA reduces the
amount of ache actually needed to achieve performance.


To even consider the EVA in the same market as the 9980 is ludicris. At
most I'd place the EVA in a entry level enterprise. The advantage of the
HDS 9980 in a single frame is clear and thats scalability, and performance
with out having to upgrade the crontrollers/frame.

Er... scale up vs scale out - scaling up is not the be-all and end-all in
most customer sites.

Also, the HDS has
enterprise level services with it, including Continuous Access and BCV's
which are tested and very reliable on the HDS frame. However, the EVA is
still lacking in that area by just releasing a version of Continuous
Access, and BCV on the EVA has yet to match the capabilities of the HDS.

What are you talking about? EVA also has enterprise level services, and CA
and BC. They work great. Exactly which capabilities of HDS are not
matched by the EVA?

--- jls
The preceding message was personal opinion only.
I do not speak in any authorized capacity for anyone,
and certainly not my employer.
(get rid of the xxxz in my address to e-mail)

On Mon, 12 Apr 2004 14:41:12 +0000, jlsue wrote:

On Fri, 09 Apr 2004 14:01:56 GMT, Jake Roersma
wrote:


What are you talking about? EVA also has enterprise level services, and
CA and BC. They work great. Exactly which capabilities of HDS are not
matched by the EVA?


I never said they didn't exsist. BC for the EVA lacks the ability to
resync snapshots, your only option is to delete and recreate the snapshot
with the same WWLUN and for most OS's the same LUN ID (it would be nice
to have this built in). I find using the interface for Business Copy on
the EVA clunky and doesn't work well with backup software (Netbackup,
etc.). I'd like to see some of the capabilities for BC on the EVA pushed
to the operating system so your not stuck creating and maintaining jobs
through the java interface. As I've stated before up until recently the
snap shots where _not_ atomic. However, I do like the zero capacity
snapshots.

Continuous Access was just introduced a few months ago so I haven't had a
chance to test it to a T. I'd be interested in hearing your thoughts on
it. Does it support asyncronous mode (last I heard it doesn't)? Can you
seperate your ports to have them dedicated to CA? Although I'd like
to be able to, considering you can only have four ports its most
likely not an option. With the four port limtation of the EVA I'd be
interested to see how well you could replicate a fully loaded EVA while
still being able to use it with respectable throughput.

The call home feature of the XP and 9980 is very usful, which may only be
due to pricing gap between the two. PRS for the EVA is junk. This is a
small but key component to an enterprise storage array.

--
Jake Roersma