The Register® — Biting the hand that feeds IT

Feeds

3PAR thins storage arrays

Revs its ASIC to slim down fat volumes

SaaS data loss: The problem you didn’t know you had

Today 3PAR is doing its bit to solve the storage obesity problem with new T-class InServ storage servers featuring a third-generation ASIC and hardware-assisted fat-to-thin volume transformation.

Currently 3PAR offers two S-class storage servers, the S400 scaling to 300TB with Fibre Channel SATA drives and S800 scaling to 600TB with the same drives. A storage server is composed of two or more controller nodes with drive arrays. The controller nodes are clustered - interconnected - by a full mesh passive backplane which operates at 1GB/sec and connects to an ASIC in each node. The S400 can have 2 to 4 controller nodes, the S800 2 to 8.

There is also an entry-level E200 system scaling to 96TB with just 2 controller nodes. All products have the same operating system and firmware. The S400 and S800 were launched in 2004 and have had in-model upgrades since then with larger capacity FC and SATA drives and a second generation ASIC.

The new machines are effectively the old ones with a storage ASIC brain implant plus a coming operating system revision. There is no change in the backplane design, port counts or hard drive types and numbers. Like the existing S400 the T400 has 64 Fibre Channel (FC) ports plus GigE and iSCSI support, and scales up to 640 drives. The larger T800 has 128 FC ports and can have up to 1,280 hard drives just like the S800.

An SPC-1 benchmark, the fastest single system one, for the T800 has been submitted and it shows 245,989,65 SPC-1 IOPS at $9.30 per IOPS. No short-stroking of drives was needed for this and the array was 83 percent capacity utilised. For comparison a high-end HP XP24000 recorded 200,245.73 SPC-1 IOPS at a cost of $17.96/IOPS. The S800 had an May, 2004 SPC-1 rating less than half that of the T800.

The new ASIC has new 'zero-detection capabilities' which help a storage volume become, and stay, thinly provisioned.

Thin provisioning means an application 'thinks' it has, say 20TB of capacity available, into which it has written, say, 5TB of data. In fact only 5TB plus a margin is allocated to it with the array controller allocating more storage as it is needed for writing data. This avoids having a lot of pre-allocated but unused disk space in an array.

Zero-detection means that an existing storage volume, perhaps migrated to the new systems, can be checked by the hardware and allocated but unwritten capacity - full of zeroes - detected and reclaimed in a fat-to-thin volume conversion. This zero-detection silicon is separate from the controller CPU and RAM resources so that ongoing storage operations are not affected by the fat-to-thin conversion process. 3PAR says a software-based fat-to-thin process would have used up controller CPU and RAM resources diminishing service levels for users. However this special hardware will only be enabled by a coming release of the InForm O/S for the new systems.

The new ASICS also run data movement in parallel with the metadata processing performed on the Intel CPU-driven controller nodes. They support an accelerated RAID 5 implementation with a built-in RAID 5 XOR engine. Apparently it delivers performance levels comparable to RAID 1 without the overhead.

3PAR says the new products will enable it to strengthen its message of massive scalability within one system making it more attractive to cloud storage and storage-as-a-service applications, which are becoming a core customer area for its machines.

An entry-level T400 with 2 controllers, and 16 147GB FC drives will cost about $130,000. In general there is a small premium over the effectively replaced S-class products.

It is known that the IBM XIV marketing people think that 3PAR is their main competitive focus in cloud SaaS applications. These new T400 and T800 models will increase the pressure on IBM's XIV developers to come up with the product delivery goods and quickly release an upgraded system.

®

Steps to Take Before Choosing a Business Continuity Partner

Latest Comments

Minis and Ferraris

M'lud, may I respectfully beg to differ? The 3PAR Mini went faster than the HP XP2400 Ferrari. In the SPC-1 summary the 3PAR box is listed as having 146GB 15K FC drives. So, at face value, one Fibre channel drive array went faster than another Fibre Channel drive array. 3PAR sells against DS6000/DS8000 from IBM, Symmetrix and high-end Clariion from EMC and XP/high-end EVA from HP. It is an enterprise drive array supplier in my view. I've seen an IBM XIV in-house training doc and IBM reckons that 3PAR is the XIV's main competitive focus with the XIV being positioned as a sell into cloud storage/storage as a service customers where the DS8000 has been rejected. This InServ array of 3PAR's is, I'd suggest, a bit more appealing to enterprises than you might give it credit for.

I am, respectfully y'r honour, y'r obedient servant, etc. etc.

0
0

only 6 nodes

The SPC-1 test only has 6 nodes, imagine what the box can do with all 8! Seems to be roughly 74k IOPS per pair of nodes, so another 2 nodes would put it up in the 300k IOPS range, whew.

Can't wait to get my T400 eval!

0
0

My array is NOT better than yours...

How can you seriously compare a 3Par T800 with an HP XP24000? They are not in the same league. The 3Par is a Tier 2 device running only SATA drives and the XP is a top end Tier 1 solution running FC drives and with significantly more scalability without even going into the other factors that make a real Tier 1 array. To compare $/IOPS on these is like comparing a Mini with a Ferrari or Accrington Stanley with Real Madrid.

SPC benchmarks are notourisouly dubious anyway as the configurations used to gain these numbers in no way compare to real-life scenarios or configurations. They are just marketing hype that no one in the real world takes seriously.

0
0

More from The Register

SCO vs. IBM battle resumes over ownership of Unix
Zombie lawsuit back and wants to suck the brains out of Linux
 breaking news
You don't need phone lines or cable for ANYTHING, says Dish
The satellite-dish man can sort you out with phone and broadband over the air too
 breaking news
What's HP got under wraps? Looks awfully flash and tape shaped
What happens in Vegas won't stay there - we've got the details
Microsoft borks botnet takedown in Citadel snafu
Stupid Redmond kicked over our honeypots, wail white hats
IBM's $1bn layoffs latest: Now axe swings in US, Canada - reports
Union claims 121 storage bods canned after dismal sales
NetApp musters muscular cluster bluster for ONTAP busters
Storage array OS overhauled to juggle more nodes, go down on you, er, less
HP adds 'Haswell' Xeon E3s to entry ProLiant servers
Gussies up MicroServer for SMBs, adds baby switches