In the second part of Pat’s keynote, I did a quick demonstration of future technologies from EMC and VMware to support IaaS use cases:
- “An Infrastructure Portal” = The next version of EMC Ionix Unified Infrastructure Manager 2.0 (aka UIM) – target Q3
- “An End-user self-service portal” = future capabilities of VMware (“project Redwood”)
The point is to show where we’re going (after all – half the fun of EMC world are big “unveil” announcements, the other half is seeing what we’re thinking – and then arguing about it :-)
The idea that’s important here is that long term, we think infrastructure HAS to become much simpler, and will be abstracted across the stack of server, network, and storage. These abstractions of elements into “service offerings” needs to be simple. Then, ultimately the infrastructure needs to be consumed in new self-service models by end users along with IT.
If you’re interested to get a preview into the future read on….
This demontration and discussion portends a new IT role – I don’t know what to call it, so I’m calling it an “Infrastructure Administrator”. Its an amalgam of people with Server/Network/Storage expertise/skills (kind of like the vSpecialist qualification list).
It’s likely that this will be a battleground between the server/network/storage folks, but a likely future outcome if you ask me. For people looking at this the right way – it’s a great opportunity. Remember when you think about the external cloud offerings out there at large scale, most of them don’t have the “cylinder of excellence” silo model that is prevalent in IT today.
As Sanjay Mirchandani, EMC’s CIO said at our Exec QBR last week “I have a EMC IT job opening for a Cloud Architect. That’s a job that just didn’t exist last year”.
So – what is shown in the demo?
- First – we show SIMPLIFICATION. We show how UIM 2.0 provides a “infrastructure dashboard” – consumption across the infrastructure landscape – but below the VMware administrator.
- We show how UIM 2.0 can make adding another whole Vblock a very, very simple task. Remember, stop thinking about it as a “set of equipment” and start thinking about it as a product – a “VM hosting black box” and you are on the same wavelength the VCE Coalition is.
- EMC Ionix UIM can subdivide the multiple Vblock stacks into multiple tenant/administrative domains.
- FYI - Longer term, we’re also pushing hard with VMware (via vCenter integration, but also the vStorage/vNetwork initiatives) where perhaps we can get to the point where even this isn’t needed – but file that under “advanced R&D”, not “product development”.
- Second – we show ABSTRACTION. We show how UIM 2.0 can then take those Vblock resources can be quicky and easily used in the definitions of service offerings (CPU, Memory, storage, network combinations) based not on specific elements, but rather SLA definitions.
- Third – we show how DYNAMIC this can be. A better word may be “elasticity”, but those are related ideas. The service offerings (end to end) can be provisioned (and deprovisioned) in a single click across the stack, including everything needed to build the ESX clusters. The service offering also has built-in elasticity itself (since you provide “grades” and “scaling ranges” when creating the service offering.
I’m going to pause here for second and make some observations. My vSpecialist team and I (along with the VCE SST folks and the VMware vCloud teams) are in CONSTANT engagement with many customers building and many who are running Cloud IaaS services (both enterprises and service providers). Some use the VMware/Cisco/EMC technology stacks (but not Vblocks), some have embraced standardized Vblocks to simplify, some have rejected 1, 2, or all 3 of the VCE Coalition companies :-)
Every single one of them have constructed some sort of hardware orchestration engine underneath their virtualization layer to integrate the provisioning process. Some have used other enterprise frameworks (BMC and others). Some have built incredible home grown tools. Some of those homegrown tools eventually got spun out as their own companies (look at Credit Suisse and DynamicOps as an example).
What Ionix UIM tries to do is be laser focused. Since it is focused on Vblock exclusively, there’s no real hardware variation – which is a huge, HUGE problem with orchestration of the hardware layer.
Some people will call this “lock in”, but I don’t get that. VMware, Cisco and EMC continue to offer up the “hey, we think our individual elements all rock, and we all support open standards, and you can mix and match how you want”. The flip side is that you will build your own orchestration, and we all integrate with the big folks in this domain.
You have the choice: a la carte IT, like we’ve always done or the new prix fixe integrated model.
ok… back to the super-cool second part of the demo…
The next section takes us OUT of the “infrastructure administrator” domain, and into the domain of the “VMware administrator and end user”.
- Fourth – we show how SELF-SERVICE is now simple and easy for the end user via Project Redwood.
- logging in as the administrator, we can see that project Redwood can aggregate many datacenters (including external vClouds, many vSphere clusters, many clusters). We also see that Redwood itself is multi-tenant, and abstracts all the underlying infrastructure.
- Note that this is the linkage between Project Redwood and Vblock and UIM. Redwood makes all the resources provided by vSphere elastic pools that can be consumed by the end users, but assumes that the pools themselves are static. UIM and Vblock make the underlying physical infrastructure supporting vSphere elastic.
Instead of me saying all that, you could just watch the demo :-)
UPDATED: Since VMware “Project Redwood” is pre-release, VMware asked me to take down the video, so what I did is edited out the Redwood part (note that I didn’t name it, or suggest a release date :-) … It remains, “…that which shall not be named” :-) Likewise, I’ve edited the high-rez demos. If you want to see a preview, contact a vSpecialist, who will engage a VMware resource to show you (no dates, no names)
Download the high-rez versions here: WMV and MOV
Some important caveats and important notes on UIM (I can’t say much about project Redwood, don’t bother asking :-)
Q: How is UIM going?
A: We got the first release of UIM 1.0 (which had no dashboard, and only did servers/network) out at the end of last year and sold to the first customer on last day of the year! we have seen about a 70% attach rate to Vblock deals so far in 2010.
Q: Is UIM a required element in a Vblock?
A: This is a very important comment, so pay close attention. NO. UIM is not a required Vblock element. UIM is what is called “recommended”. This means it’s still ABSOLUTELY a Vblock without UIM. All the elements of the stack support open, public APIs, and most have good integration with enterprise management frameworks. UIM is of course, in my opinion, ahead in most places (behind in others), but also accelerating super-fast due to it’s laser-focus.
Long term, (warning, personal opinion) I think the value of mass heterogenous hardware management will decline (will alway be there, just decline relative to doing ONE bit of hardware right). Likewise, the advent of vApp containerization of policy and state portends that the “single massively centralized truth of a CMDB” is also likely less and less valuable in the future.
Q: Can I use UIM with ______ server and ______ network?
A: No. UIM is for VCE Vblock management at this time. Please don’t distract the UIM team :-) This is how the UIM team is able to move as fast as they are. They have a very focused target.
Q: Will UIM 2.0 support Type 2 Vblocks and the upcoming Unified Type 1 update, or Type 0?
A: UIM 2.0 will support server and network configuration across the board, but recognize that this is the first release that integrates the storage elements of the stack. UIMv2 will only support provisioning of block storage models in a Type 1 Vblock. There will be a UIMv3 – we’re working on it, and it will support both block and NAS models, and across the Vblock product family.
Q: Will UIM 2.0 support an API?
A: This is a smart question IMO. It highlights that you’re thinking about the Vblock as a product, and UIM as an “element manager”, since all “element managers” need to have an API to integrate into other tools. The answer is YES. Fred Crable will be giving a developer sessions at EMC World on how to control UIM 2.0 with the public API.
Hope you dig this – I think it’s super cool! Please tell me what you think? Are we on the right track? Wrong track?
Yeah UIM is BOTH super-powerful and super-cool.
Good stuff.
Posted by: CC | May 11, 2010 at 09:00 AM
Chad,
UIM 2.0 looks really impressive. The only other automation tool i've seen was from HP (Matrix) which converges Compute, storage and network but the challenge with this type of integration is that it's dependent on all the layers. If one of these layers gets i.e upgraded the entire stack can easily break. In fact, from talking with an ex HP SW developer who was part of that project this was by far their biggest issue...and they controlled all 3 layers. How's EMC going to deal this given that although it integrates with does not control the Compute, nor the Networking layer?
The other question I have is with regards to end-to-end visibility from a fault isolation perspective. Clearly providing an abstraction layer and hiding the various complexities from the administrator is a good thing, but the tradeoff here is how does one deal with fault isolation? Even the slightest anomaly, will have an administrator "in the dark" given he has no idea not only as to how the system is provisioned not even where the storage is coming from? Same goes for compute and Network. So how is this addressed?
Thanks
Nick
Posted by: Nick Triantos | May 15, 2010 at 07:51 PM