Hitachi VSP (Virtual Storage Platform) & Command Suite 7– Technology, Comparisons, Architecture
A Deepdive on VSP Technology, Command Suite 7 overview, Technology within the VSP, Comparison of VSP vs USPV, some architecture discussions and marketing message. Along with this discussion, also see architecture block diagrams and Videos from the event.
Hitachi and its US Subsidiary Hitachi Data Systems announced its next generation Storage platform on the 27th of Sept, 2010. It’s proven technology of storage virtualization that surfaced back in 2007/2008 is being offered in its latest platform code name “VICTORIA” now called VSP – Virtual Storage Platform.
Though I do not want to speculate too much on the naming, VSP – Virtual Storage Platform is a relevant name to the technology, But is the name VSP somehow influenced from the name VMAX (Virtual Matrix)??
The same day HP also announced its P9500 storage platform, which it rebrands from Hitachi with an HP logo and HP management software. The looks of the HP version of VSP, (P9500) is very attractive compared to the Hitachi looks.
Wonder the HP – 3Par acquisition will put some pressure on the OEM relationship between HP and Hitachi Ltd, Japan, since essentially the game would be to compete in the same market space now. Though to my understanding 3Par doesn’t offer Mainframe support with its storage as Hitachi does today with FICON.
Though do not get deceived by the name or the looks, the technology that VSP brings to datacenters (let me correct, virtual datacenters), is one that is revolutionary and will help customers build more resilient and efficient environments.
Hitachi VSP at Hitachi Information Forum in Santa Clara, CA
The color of the VSP cabinet is Green, indicating it’s a step forward towards a highly energy efficient system. As the datacenters are now being completely virtualized with computing environments and no geographical boundaries, the requirement for storage virtualization becomes key in being able to keep these environments resilient, scalable, reduced footprint and manageable.
Victoria was the code name for VSP, during our last visit to HDS at the Geek Day 0.9 in Santa Clara; we were given some hints about this project. But during our visit to Odawara (Tokyo), Japan, Hitachi along with its US subsidiary (HDS) gave us a preview of VSP, the underlying technology and architecture of VSP, a one for all platform Block, File, Object storage. Though we did not blog about the technology discussion that took place in Japan, they were highly focused around engineering & architecture aspects of the VSP technology.
VSP brings architectural enhancements, added flexibility, reduced footprint, higher response times, reduced management, concepts of storage economics, etc natively within the platform.
It is expected VSP will also be the core storage platform on HDS’s UCP (Universal Compute Platform) along with Hitachi Symphony Servers, a networking partner (****) and Microsoft Operations Manager as its orchestration software.
Nigel Poulton also has a Technical Deepdive Post on VSP, and it is very technical in nature.
The Marketing Message
The core messaging behind the VSP platform includes 3D Scaling, which are Scale Out, Scale Deep and Scale Up. In the past we have seen blog posts from Hu Yoshida and Michael Hay about 3D Cartesian scaling and affects of it on Storage Platforms.
Some additional pitches from HDS on VSP include being able to provide Virtualized, Automated, Cloud-Ready and Sustainable platform. Though I necessarily do not understand what Cloud-Ready means. The messaging around Cloud was particularly missing during the Hitachi Information Forum.
Scale Up refers to the tightly coupled storage environment that is easy to expand and manage.
Scale Out refers to the priority queues, dynamic allocation of resources and system that helps customers expand as business needs and workloads change.
Scale Deep refers to the Storage Virtualization piece that allows a single VSP system to grow using external storage through centralized management to more than 255 PB of data.
Storage Virtualization is a great technology and the benefits of it are being seen around the industry today. Manufacturers that did not have this technology a few years ago are all jumping in now. These days talking to customers about the managed services businesses and understanding the value storage virtualization brings to the table with technologies like Hitachi USPV, IBM SVC, HP SVSP, EMC VPLEX and now Hitachi VSP is pretty phenomenal.
On the VSP Hitachi also introduced the SAS II drives 2.5 inch form factor, reducing the footprint substantially. With a 2048 drive system, customers are typically looking at 6 standard cabinets vs an EMC VMAX that may utilize 10 cabinets for the same number of drives. The largest drive supported today on the VSP is 1 TB drive.
Along with the added number of drives to the VSP technology, the Storage Virtualization technology enables 255 PB’s of Storage behind a VSP or essentially 1TB x 255000 drives in a single federated storage system.
After the leap by EMC into the Intel Architecture with it enterprise Storage system VMAX earlier last year, Hitachi is the next storage manufacturer to take advantage of the great engineering work that is currently being done by Intel for Enterprise computing. Along with the Intel Xeon CPU’s on the Virtual Storage Processors, Hitachi also uses Hitachi ASIC’s on its controllers for specialized functions within the VSP.
The number of ports has also been now doubled with VSP for host connectivity, substantially less power consumption which numbers seem to be in the range of 40% to 50% energy efficient systems for power savings.
VSP also enables XTS-AES256 bit Encryption of data as its being written to its disk. This technology more than likely could be a third party plug-in that enables this feature. It will need to be purchased and enabled through software keys within the VSP.
One size fits all (Scale Up, Scale Down)
As you are aware, the USPV came in two flavors, the USPV and the USPVM. If the customer had invested into a USPVM and as the business demand (applications requirements, IOPS, workloads) increase, the only option the customer might have is purchase another system. There are similar offerings from EMC in this space with its VMAX and VMAX-SE frames. The VSP goes back to the basics of purchase a system and expand it based on your needs without the necessity to purchase a new system.
With the new generation of storage virtualization technology just brought to the market by Hitachi, there are differences between its predecessor, the USPV.
While the VMAX today offers 128 cores, the VSP starts at 32 Cores, but using Storage Virtualization, you can add thousands of Cores behind it.
Okay, just an example….
One of the large financial houses that were on the panel at Hitachi Information Forum virtualizes DMX-4′s behind USPV’s today. If a VSP supports 255000 drives, you can practically have 106 fully populated (2400 drive configured) VMAX systems behind one VSP.
Since manufacturers leverage technology and its inter-workings in different ways, a side-by-side comparison of VMAX and VSP may not be a fair comparison.
Though I want to point out differences between VSP technology and USPV technology relating to architecture and configurations.
Hitachi VSP vs USPV
|VSP Technology||USPV Technology|
|Name||VSP: Virtual Storage Platform||USPV: Universal Storage Platform – Virtualization|
|Cabinet||Min: 1 Cabinet
Max: 6 Cabinets (2 Systems)
|Drives (2.5 inch SAS)||Min: 0 Drives (External Storage)
Max: 2.5 inch x 2048 SAS II Drives
|3.5 inch x 1152 FC drives|
|Drives (3.5 inch SAS)||Min: 0 Drives (External Storage)
Max: 3.5 inch x 1280 SAS II Drives
|3.5 inch x 1152 FC drives|
|Federation||Min: Single System
Max: Two Systems tightly coupled using the Hitachi Star Fabric over PCIe
|External Storage (Federation – Virtualization)||Max: 255 PB||Max: 247 PB|
|External Storage (Federation Drives)||1TB x 255000 drives||1TB x 247000 drives|
|Storage||2 PB’s Internal||1 PB Internal|
|Processors||INTEL Quad Core Processors plus ASICs on FED / BED||ASICs|
|Single Controller (System)||Min: 1 Cabinet
Max: 3 Cabinets
|Virtual Storage Director Blades||4 Cores per Blade
Min: 2 Blades (8 Cores)
Max: 8 Blades (32 Cores)
|FED (Front End Directors)||Each FED has 8 ports
Min: 2 FED (16 ports)
Max: 24 FED (192 Cores)
|FED Port Speed||8 Gbps||4 Gbps|
|FED Port types||Min: 16 – FC Ports (8GB)
Max: 192 – FC Ports (8GB),
192 FICON Ports
|224 – FC (4GB ports),
112 – FC (8GB ports),
112 – FICON
|FCoE||It may be supported in a short duration but no support with Release 1||Not Supported|
|iSCSI||Not Supported||Not Supported|
|Infiniband||Not Supported||Not Supported|
|BED (Back End Directors)||Each BED has 4 SAS links
Min: 0 (incase of external Storage)
Max: 64 BED (64 SAS links – Two federated VSP’s)
|64 FC loops – Half duplex AL loops|
|BED Speed||6 Gbps SAS||4 Gbps FC-AL|
|Cache||32GB or 64GB Adapters
Min: 2 (64GB cache)
Max: 16 (1024 GB Cache)
|512 GB Cache|
|Cache Protection||Flash plus Battery||Big Batteries|
|Power Consumption||30KW (one phase) – 1024 drives||33.2 KW (three phase)|
|Automated Dynamic Tiering||LUN Level and Sub LUN level||LUN Level|
|VAAI – vStorageAPI support||Yes (expected late 2010)||No|
|Native within UCP (expected)||Yes||No|
|Support for 2.5 inch drives and/or 3.5 inch drives||2.5 inch and 3.5 inch drives||3.5 inch drives only|
|SATA drive support||Yes||Yes|
|SAS and SATA drive support – intermix||Yes||SAS not supported|
|FC and SATA drive support – intermix||FC not supported||Yes|
|Cache Mirroring||Write Cache Mirroring||Write Cache Mirroring|
|Command Suite 7||Supports||Supports|
|Operating System||BOS / BOS V||BOS / BOS V|
|LUN based Tiering||Yes||Yes|
|Sub LUN based Tiering||Yes||No|
|Drive Formats||??? bytes (expecting 520) with 8 bytes ECC||??? bytes (expecting 520) with 8 bytes ECC|
|Microcode||Runs on Virtual Storage Processors||Runs on FED, BED|
|Rack System||19 Inch Racks, 42U||19 inch Racks|
|Airflow||Hot – Cold Aisle||Hot – Cold Aisle|
|RAID||Mirroring, RAID 5, RAID 6||Mirroring, RAID 5, RAID 6|
|Cooling Fans||Noise Reduction||Non|
|Cooling Fans Speed||3 level speeds for cooling||1 level speed for cooling|
|Control Memory||On Virtual Storage Directors||On FED/BED|
|Color of the Cabinet||Green||Bluish / purple|
|Purchased as||VSP only with Controllers or with X number of drives scalable to 2048 drives||USPVM only sold with Controllers. USPV sold with controllers and drives. USPVM cannot be upgraded to USPV|
Cabinet Numbering and Structure
Below are how two VSP systems are coupled together using the Hitachi Star Switch (PCIe Connect), which enables the expansion of two VSP into a single system scalable to 2048 drives with 1024 GB of cache.
|Cab 12||Cab 11||Cab 1||Cab 0||Cab 1||Cab 2|
|Drives only||Drives only||Controller 1 + Drives||Controller 2 + Drives||Drives only||Drives only|
Each system (VSP Controller Unit) includes
4 x Virtual Storage Director,
8 x Data Cache Adapter,
8 x Front End Directors
4 x Back End Directors
4 x Grid Switch
2 x Drive Chasis in Controller Cabinet
3 x Drive Chasis in each – Drives only Cabinet
Totally 8 Drive Chasis
Each Drive chasis supports 128 drives (SAS)
FRONT of the UNIT includes
4 DataCache Adpaters
4 Virtual Storage Directors
4 Data Cache Adapters
Drive bays have FANs in the front of the unit
BACK of the UNIT includes
4 Front End Directors
2 Back End Directors
4 Data Grid Switches
2 Back End Directors
4 Front End Directors
Virtual Storage Director (The Brains behind the VSP)
There are 4 Virtual Storage Directors in each system
Each Virtual Storage Director has 4 Cores
You can have 16 Cores per system
These processors manage the internal workings of the VSP along with the LUNs, eLUNs, Addressing, data mapping, virtual partition manager, layered software interface, references, SAS drives if internal, operational control data memory.
You can expand this to 32 Cores using the PCIe Hitachi Data Switch Grid and 3 additional cabinets along with a controllers.
Serves at L2 Cache
On Virtual Storage Directors
Responsible for managing and maintaining Metadata, mappings, etc
Data Cache (Global Cache)
Primarily used as Cache for read/write
Caches data during read operations from BED, similarly caches data from FED for write operations
Only write data is mirrored in cache (not all data is mirrored like the VMAX)
1024GB of total cache for 2 VSP’s tightly coupled
Read Operations only require one copy of cached data
Cache backed up to onboard Flash drive, reducing the amounts of needed batteries
BED / FED and ASICs
Though the Virtual Storage Directors use the Intel Quad Core processors, BED’s and FED’s use special purpose ASICs for I/O operations, which enables a much better, and flexible data movement and associated performance
Unlike VMAX which uses Rapid IO for coupling its engines, Hitachi uses its custom designed Hitachi Star Fabric for tightly coupling its Internal Network to manage data which includes the Drives, Virtual Storage Directors, Data Cache, BED & FED. This switch also connects two VSP’s together to form a 6 cabinet, 2048 drive system, which is connected through PCIe at a CPU level.
Dynamic Automated Tiering (Sub LUN Tiering)
Once a disk is assigned to the VSP whether it is native within the VSP or external virtualized (eLUN), VSP will utilize it for Dynamic Automated Tiering. With the announcement of VSP, HDS is also including policy based Sub LUN Tiering to this platform, allowing automated data movements in page size of 42MBs.
Dynamic Automated Tiering is shipping day one with VSP platform. Look at Sub LUN Tiering as a technology that will move the data real time based on policies setup in the environment. As a certain page gets a higher heat index, the underlying technology will automatically upgrade the tier only for that page.
If over time, a certain page falls on the heat index chart, the data is moved to a lower tier. This technology helps bring more efficiency into environments and does not require having all your data stored on one tier and helps you save on the use of expensive SSD’s for all your data.
Again this offering as it stands today is very unique in the industry and other vendors are moving or have a road map focused towards this technology.
VMware, SMT, Cloud
Day 1, VSP will not support VAAI, the vStorageAPI for VMware offloading VMware related tasks locally within Storage controllers bringing much added flexibility to virtualized environments.
But within the next 45/60 days as first code release rolls out, VAAI support is on the roadmap to be included as supported on VSP.
Also the VSP claims to have SMT (Secure MultiTenacy) built into the architecture that allows the system to operate in virtual partitions including cache & host ports. Though still not sure how the VSP manages to offer audits, resource management etc within an SMT environment. In an SMT environment, encryption becomes a very viable offering where tenants could have its own individual key as the VSP natively supports 32 different keys.
Very noticeable product differentiation from previous generation Hitachi USPV. Improvements visible in terms of I/O, tight coupling of systems, SAS drives, drive speeds, port speeds, data routing. Expansion of external storage to 255PB total, internal drives to 2048 with reduction in floor space and hot and cold aisle friendly cabinet design shows the move towards the next generation thought process.
No Total cache mirroring
No VAAI support day 1
Cloud ready message needs to be more refined
The product seems to miss the flashiness
Having the front bezel green in color is a great message.
Color coded cables in the cabinet based on different loops possibly, cable colors are black, grey, white and are very visible.
HDS Command Suite 7 for VSP, AMS and USPV
Along with the announcement of the VSP, HDS also announced its next generation SRM tool Command Suite 7 that enables management of the VSP along with its predecessor USPV & USP, along lines its mid tier storage platform AMS.
The message driving the Command Suite like the VSP platform is the 3D Cartesian scaling, which is Manage Up, Manage Out and Manage Deep.
The Command Suite 7 is being compared to solutions from other vendors in this space. Along with managing the Hitachi systems at an element manager level, Command Suite 7 also offers heterogeneous storage support for other vendors.
It seems the Command Suite 7 or its elements might end up within the UCP announcement likely to happen early 2011, which will offer an integrated stack of solution from HDS.
Command Suite 7 is a move towards being able to manage & discover all storage, either virtualized behind a VSP, non virtualized but in the environment, VMware host, Virtual Machines, Switches, hypervisor and wants to impact as an infrastructure monitoring tool.
Manage Up: References the ability to manage the Hitachi environments that will enable automated dynamic tiering. Command Suite 7 also boasts the management of 255 PB of storage and 5Million objects through a single installed instance
Manage Out: Single Solution for management whether is File, Block or Object storage. Along with the ability to manage VM hosts, VM’s, applications, it also enables the management of heterogeneous storage. It is not expected to replace your existing native element managers within your storage environment by Command Suite. It is also expected that within Hitachi storage, Command Suite 7 will be able to manage the objects without agents being deployed in the environment.
Manage Deep: Through the use of reporting, single pane of glass view, capacity monitoring, performance monitoring, Command suite 7 enables a granular management of your storage environment adding automation reducing operations to accomplish required tasks.
Automated Dynamic Tiering
Hitachi introduced Sub LUN level tiering with its VSP and Command Suite 7 offering. The automated tiering will work either at File / Object or Block levels. Through a policy-based engine any of the above can be migrated to either a lower or a higher Tier based on SLA’s, performance, time of the day, application requirements or cost.
With the Sub LUN Tiering, Hitachi allows the movement of its data in 42MB page size, which is a standard within its storage environments and enables it to be promoted or demoted based on policy.
The sub-lun level tiering enables only the hot blocks of data or file to be moved rather than an entire LUN. Also by default all new incoming data gets written to the highest performance tier first and gradually gets demoted to a lower tier as the activity on it is reduced, again all this can happens at a LUN level or a sub-LUN level.
The Command Suite 7 ships as a BOS (Basic Operating System) or BOS V (Basic Operating System V), which includes the following Software modules as part of the offering.
Hitachi Device Manager (BOS)
Hitachi Universal Volume Manager (BOS V)
Hitachi Dynamic Link Manager Advanced (BOS / BOS V)
Hitachi Dynamic Provisioning (BOS / BOS V)
Hitachi Dynamic Tiering (BOS / BOS V)
Hitachi Command Director (BOS / BOS V)
Hitachi Storage Capacity Reporter (BOS / BOS V)
Hitachi Tiered Storage Manager (BOS / BOS V)
Hitachi Tuning Manager (BOS / BOS V)
Hitachi Virtual Server Reporter (BOS / BOS V)
** Have no idea how Command Suite is licensed today, but I would think the pricing for both the BOS and BOS V are different and possibly would have a Base price plus a per TB of licensing cost.
Security, SMT, Cloud
Along with the discussion above around VSP in SMT and Cloud environment, Command Suite 7 offers additional benefits towards applications being hosted in a cloud environment.
Through the use of industry standard implementations of Active Directory, Radius etc, the storage managers, administrators, backup admins, etc get authenticated in the environment.
Through the use of Virtual partitioning, resources can be allocated in partitions and only managed through those with the correct permissions, giving the management of the environment the much-needed granularity. There is also support for Host port segregation, reporting and management includes support for provisioning, migration and replication.
For SLA and Reporting, Command Suite 7 has support around Automated Policy based TIering. There is also support for Hypervisor discovery and reporting along with Charge backs that should be included in the near future.
Okay that was long!!!
Some More….a discussion with RIck Vanover, Chris Evans and Claus Mikkelsen (Chief Scientist, HDS) about the VSP Technology and release.
Quick Discussion with Rick Vanover and the introduction with a Japanese Band
More discussions on storage virtualization coming up in the next blog posts..
Disclaimer: I do not work for HDS. Access to this information was given by HDS over the past few months helping understand the architecture of the VSP platform. I have attended HDS Geek Day 0.9, Hitachi Japan Trip and Hitachi Information Forum in Santa Clara and learned about the technology at these events. All airfares, lodging and boarding was paid by HDS. I have not received any monetary compensation during these visits nor any gadgets.
This is just an attempt to put some light on Hitachi VSP technology and what Storage Virtualization may enable in virtualized environments.