..

VPLEX and USPV

EMC’s latest addition to the concept of storage federation is the VPLEX announcement that happened at EMC World 2010. VPLEX comes in two forms today, VPLEX Local and VPLEX Metro. Important features of VPLEX include cache coherency, distributed data and active-active clusters at metro distances. In the works are VPLEX Geo and VPLEX Global enabling inter-continental distances.

VPLEX contains no drives today, it is based on a version of Linux Open Source and runs on the same hardware as a VMAX engine. But said that, what prevents EMC from including VPLEX as part of every VMAX and Clariion sold today or may be just run it as a Virtual Appliance within the VMAX (Enginuity) or Clariion (Flare).

While HDS has a slightly different approach yielding almost the same result using Storage Virtualization, HDS approaches storage federation in its USPV platform. The USPV scales upto synchronous distances, I believe 100 kms max distance today.

USPV natively uses a combination of Universal Volume Manager (UVM), High Availability Manager (HAM), Dynamic Replicator, Hitachi TrueCopy Remote Replication and Hitachi Replication Manager to do synchronous distance (100 kms) replication with distributed data in an active-active clustered environment.

VPLEX local and VPLEX metro has been recent announcements, while the USPV has been offering similar features since the past few years now.

..

Use-Case

Service providers will be largest customers while the VPLEX is still being developed in the Geo and Global modes.

I would think, government customers like DISA, DoD and other cloud providers in the federal space may find VPLEX and USPV very interesting as well.

Migrations using both the VPLEX Local and USPV are a piece of cake, because of its underlying features.

And many more…

..

Future

Will the future of all storage subsystems have federation in it as a core component? It is most likely with virtualization technologies being designed and pushed today, that we will natively see some of these features into backend storage that can typically hold data in containers and these containers move based on requirements. Look at a VM as a container of information or application.

With a Front-end storage controller, call it a VPLEX or a USPV which doesn’t care what sort of disks are sitting behind it, natively add all storage features to it like snaps, clones, RAID, replication, high availability, virtualization and it doesn’t matter if you use the cheapest storage disk behind it.

Typically with a single storage subsystem, you are looking at scaling your storage to 600 drives or 2400 drives or 4096 drives or 8192 drives or 16384 drives max or does it even matter at this point.

Storage federation will allow a system to scale upto 100’s of PB of storage, for example a EMC VPLEX scales upto 8192 Volumes today, while a USPV scales upto 247PB’s of storage, in essence that is 1 TB x 592,000 disk drives in a single system (federated).

When you connect two of these VPLEX’s or two of the USPV’s at synchronous distances, you now start taking advantage of active-active clusters (datacenters) with distributed data. (Again I will be the first to say, I am not sure how much of cache coherency is built within the USPV today).

But that brings us to some important questions…

..

Questions

Is storage federation that important?

Is storage federation the future of all storage?

Do you care about active-active datacenters?

What is the use-case for federation outside of service providers?

Will this technology shape the future of how we do computing today by leveraging and pooling storage assets together for a higher availability and efficiency?

How large, a single namespace would you like to have? I believe HP IBRIX brings a similar concept of scaling storage to 16PB’s total in a single name space..

Does federation add latency, which limits its usage to only certain applications?

Is VPLEX the future of all EMC storage controller technology, and will that eliminate the Flare or Enginuity code?

If you add a few disk drives to the VPLEX locally, can it serve high demand IOPS applications?

How large will cache get on these storage controllers to minimize the impact of latency and cache coherency on devices at synchronous distances? Is PAM or Flash Cache that answer?

At that point, does it matter if you can do coupling on your systems to extend it like we initially thought the VMAX would have 16 or 32 engines or may be you can couple Clariion SP’s or AMS or USPV’s?

..

More Questions

Will the future VPLEX look like a USPV with local disk drives attached?

Though the big vision of VPLEX is Global replication creating active – active datacenters, does the next generation VBlocks meant for Service Provider market include a VPLEX natively within it?

Is EMC Virtual Storage just catching upto HDS technology? Or is VPLEX vision a big and unique one that will change the direction of EMC Storage in the future…

Is Storage federation game changing and is EMC ahead of HDS or HDS ahead of EMC…

..

This is an attempt to live blog at the Hitachi Data Systems (HDS) Bloggers Day: Session 4

Day 2:

Pete Gerr, Director, Strategic and Solutions Marketing

8:30: Kickoff…

8:31: HDS going through parking lot questions…some will be answered today, some later on one on one discussions.

Michael Hay, Senior Director, Product Strategy

8:40: John Deere is owned by Hitachi..

8:40 tons of tweets out there…..Nigel wins the award for the MVP and Devang wins the best dressed…..

8:41: Hitachi Research can be found at http://www.hitachi.com/rd

8:42: Even though HDS is moving into this arena, we are using HDS proxy servers to connect to the internet from here, the HDS Geek website is blocked internally. Here is the site that HDS might want to consider to open http://hds.com/go/geekday/index.html

8:45: High level Hitachi (not HDS) Vision, Strategy.

8:45: Gazopa.com another visual search engine

8:50: RAID discussion kicked off by Nigel Poulton and its getting quite interesting…

Claus Mikkelsen, Chief Strategist

Mark Adams, Senior Product Marketing Manager

9:12: Storage Tiering discussion

9:15: AMS discussion

9:16: HDS midtier storage (AMS) has explosive growth over the last 2 years….

9:20 AMS was released in 2008 and has been doing great…

9:30 ……

9:55: LUN Ownership and Failover question from yesterday

9:56: THere is LUN ownership on either Controller 0 and controller 1. Active – Active cluster, I/O and architectural diagram.

10:00 Break……but a big discussion on the AMS architecture (current) and potentials in the future….

XXXXX,xxxxxxxxxx xxxxxxxxxxx

XXXXX,xxxxxxxxxx xxxxxxxxxxx

10:20: Embargo session

11:00 Round table discussion….for another hour.

Disclaimer: Though we are not required, it is up to the attendees to blog and tweet about this event if they wish to. Our travel, boarding and lodging expenses for two days will be paid by Hitachi Data Systems (HDS).

This is an attempt to live blog at the Hitachi Data Systems (HDS) Bloggers Day: Session 3

John Harker: Product Marketing, HDP, Virtualization products and stragegy

John Yarborough: Technical Ops

Larry Korbus: Product Manager, Array Based Virtualization

11:32: USP and USP V storage arrays, external storage

11:33: Tiering within Arrays (using Tiering models like Tier 1, Tier 2, TIer 3, Tier 4) and external storage.

11:34: All the above is done through using Storage Virtualization using the storage virtualization layer that will help extend the life of legacy arrays.

11:44: Dynamic Provisioning is applicable both on the USP and the AMS platforms

11: 45: Simple Provisioning, capacity leveling, automated performance optimization, performance, load balancing, reclamation, savings related to replication, IT agility, reduce OPEX and CAPEX

11:48: HDP is included in AMS but is chargeable on the USP

11:55: Zero Page reclaim discussion.

11:57: Hitachi Dynamic Provisioning also uses Symantec Storage Foundation, VxFS. Cleaning filesystems

11:47: All questions related to Zero Page reclaim, HDP, external storage

12:11: White board session related HDP, External storage, storage virtualization

12:14: HDS denotes Cache in an array as $ (Cash)

12:15: These sessions are quite technical, very interesting, white boards

12:16: USPV does add latency to IO, though measurable very little

12:17: Hitachi Device Manager demo

12:30: I am loving this…demo

—-

1:00 to 2:15: Did an infosmack podcast with Greg Kneireimen, Hu Yoshida, Nigel Poulton

Miki Sandorfi,  Chief Strategist, File, Content and Cloud

Linda Xu, Senior Director, File, Content and Cloud

Missed most of this session, this was a great session on UCP (Universal Compute Platform), I wish someone recorded this

2:25: Talking about Orchestration layer on the UCP

2:30: Networking partner has not been disclosed yet……..BUT THINK OF IT AS…..

2:45: Lots of questions on UCP…..lots of answers on UCP, good discussion

2:46: Orchestration is provided using Microsoft Systems Manager and Virtual machine manager

2:47: A bit of comparison to VBlocks, some questions asked.

259: Integration platform to enable cloud like services for private cloud, public clouds, storage as a service, infrastructure as a service, orchestration, tiering using existing Hitachi Storage technologies.

3:06: Real life experience unified IT and how customers are driving more efficiency into organization and how they are managing to use automation.

3:15: Demo of the Unified Compute Platform Orchestration software

3:30 Lab tour now….and no pictures…

Disclaimer: Though we are not required, it is up to the attendees to blog and tweet about this event if they wish to. Our travel, boarding and lodging expenses for two days will be paid by Hitachi Data Systems (HDS).

This is an attempt to live blog at the Hitachi Data Systems (HDS) Bloggers Day. Session 2

Hu Yoshida, CTO Hitachi Data Systems

The technical discussion kicks off….video coming soon….

9:56: Common platform for all data, Storage Virtualization

9:57: Storage Virtualization requirements including applications, multitenancy, scalability, transparency, etc.

9:59: Storage Virtualization in-band, out-of-band vs USPV

10:01:Dynamically provision new services in seconds

10:02: Convert fat volumes into thin volumes by moving them into the pool

10:03: Optimize storage performance by spreading the I/O across more arms

10:04: Safe Multitenancy architecture

10:10: Application Transparency, Hitachi Command Portal. SLA’s, Allocation, etc, can be customized here, for could providers.

10:11: SCALE UP and SCALE OUT are important on storage platforms

Michael Hay, Chief Strategist, Strategic Product Operations and Planning

10:15: Waiting for the virtualization circle

10:18: Deepdive into HDS’s thought process and startegy….(it is hard to capture all the thoughts)

Michael Heffernan: Solutions Product Manager, Server Virtualization

10:28: Topics around Hypervisor, impact to customers, hypervisors and storage, options around vmware, hyper-v, hds

10: 29: Hitachi has very strong set of hardware storage products

10:34: Hitachi changing messaging around hypervisor and going back to the basics to understand the HCP and Vmware IO stacks. Hitachi is closing at it very closely.

10:46: Vmware strategy: to reduce impact on the host we do Active Active Symmetric clusters on modular arrays (AMS platform), so we dont need powerpath like software’s.

10:47: Chris Evans USP ness moment…

10:48: HDP: Hitachi Dynamic Provisioning, wide stripe, vm extends, vmfs volumes. Sub Lun level locking of USP LUNs with HAM coming soon…

10:55: Scalability discussion

11:11: AMS disscussion for small to medium businesses. HDS also has a hypervisor modeling and sizing tool that customers can use.

11:16: HDS is completely agnostics on hypervisors.

Disclaimer: Though we are not required, it is up to the attendees to blog and tweet about this event if they wish to. Our travel, boarding and lodging expenses for two days will be paid by Hitachi Data Systems (HDS).