AIMing for success

 Cloud, Vendors  No Responses »
May 212011
 

As I have mentioned, I recently joined Dell. I hope I am one of those who Storagebod talks about as having done this for smart reasons not the money. I had a choice of where (and indeed whether) to move, and the main reason I chose Dell was their approach to managing infrastructure.

These days it’s all about owning the stack. It’s why Dell bought Compellent, why EMC teamed up with Cisco, and why Oracle have done whatever weird shit they did this week (OK, with Oracle it’s about world domination first and owning the stack second).

Different vendors do this in different ways.

Cisco’s UCS is undoubtedly clever. I doubt any other vendor could have got away with a solution that combined new-to-market products, eye-watering expense and complete lock-in; Cisco’s presence in CIO mindshare and their golden Gartner halo has allowed them to be bolder than I’d have believed possible and get away with it.

HP are making a competent job of serving their core market with tools that make it reasonably easy for an HP shop to make the transition to “cloud-readiness”, for whatever version of the term cloud is in vogue this week. I don’t particularly like it but mainly for religious reasons (it smells too much of rip and replace and single-vendor; it does not work too smoothly even with their older product).

In the stack-to-cloud space Dell, as so often, is playing a game which at first glance looks suspiciously like following the crest of the wave and sweeping up the leavings, but actually I think it is a lot smarter than that (and I have views on the question of Dell and innovation). Yes, of course I do, that’s why I am working for Dell now. Please check my working:

The Dell Approach (and why I think it’s right)

There are two main parts of the Dell integration, they are called Advanced Infrastructure Manager (AIM) and Virtual Integrated System (VIS, definitely not pronounced Viz). The cricial thing about both of these is that they will work across multiple platforms; HP, Dell, VMware, Microsoft, Red Hat – all fair game. In a world where people increasingly want to get away from purchasing decisions dominated by vendor labels and towards a focus on the service provided, this seems to me to be the right approach.

If desktop virtualisation is driven by “BYO3” (of which more anon, hat-tip Barb Goldworm) then surely the next big move in server virtualisation must be “bring your own stack”, a term for which I claim copyright, trademark and quite likely T-shirt rights. Both the Dell products will work with what you have and probably with what you buy next even if it’s not Dell. That’s rather a powerful message.

AIM is the former Scalent product line, with some further development. Scalent was widely recognised as an innovative and intelligently designed product which performs for servers pretty much the function of user state virtualisation (the AppSense way, not the currently-laughable Microsoft version). The function of the server – the OS and software load – is abstracted from the hardware and delivered as a “persona” which can be targeted to another device or to a virtual machine. Today I have been moving SAN-booted Windows and Red Hat instances from bare metal to ESX to Hyper-V with remarkable ease. The UI takes some getting used to but there’s no doubt that the technology is seriously good. Importantly Scalent was an HP-first product. Some of Scalent’s biggest customers are HP shops and not only are there no plans to drop support for HP hardware, the test matrices actively embrace new kit from HP, IBM and others.

VIS is currently a mixture of Dell intellectual property and technologies licensed from other firms. I don’t think it’s any secret that the Creator component is based on Dynamic Ops Cloud Automation Center, which is a product I love, and the Director piece will be – ahem – hauntingly familiar to those who have compared such industry legends as vFoglight and Netuitive SI. Frustratingly, the precise content I would like to put here is covered by NDA and has been for easily six months, I am working on finding out when the embargo lifts, but suffice it to say that VIS is built on products generally considered best-of-breed. In this case “generally considered” means that not only do I think it myself but lots of other people have agreed.

The crucial points about both of these elements are:

  • Vendor agnostic (YMMV)
  • Built on well-known and widely respected products
  • Significant Dell investment in both new development and maintaining cross-vendor compatibility

The last is, for me, the most significant.

Now, I don’t pretend to have the judgment of Solomon but this is a fast-moving field and right now I think Dell is positioned in a very exciting place. And hopefully I’m along for the ride.

Bring what you have

 Virtualisation  No Responses »
Apr 152011
 

Some people have asked me why I joined Dell. I have many good friends in EMC who feel this was a bad call. They may be right, but here’s my reasoning.

As a virtualisation specialist I looked at a lot of presentations and RfP responses from companies active in the virtualisation and integration space. It’s been apparent for some time that the margin is currently in storage, integration and management tools – basically cloud-readiness – not in selling tin, so all the big vendors have been driving towards owning the stack. Let’s be clear here,though, I was very happy at SunGard and having a lot of fun – I did not go looking, I was recruited. Several firms came calling.

Some vendors I would not have been interested in whether they called or not. HP bought 3Par, which was a good call but way too expensive. I looked at HP’s integrated stack offering and although there was talk of being extensible to other vendors it was pretty clear that the bulk of functionality would not work other than on latest generation HP hardware. That’s lockin and rip and replace, to my eyes. HP has a lot of market share in servers and a good record in innovation but the approach seems to me to be wrong for the customer, even if it’s right for HP’s long-term lockin – I would love to work with @HPsisyphus but in my bit of the space, the operational life of the virtual estate, HP is going a way I am uneasy with. IBM has tin, has storage, clever software and the biggest, deepest and most bomb-proof silos on the planet. They have some incredibly clever people but I could never work there. Oracle… just no.

But there are two firms with whom I have a long-standing relationship as a customer which has been generally good, occasionally bad and sometimes spectacular. They are EMC and Dell, and when I started out in virtualisation they were hand in hand. The decree nisi is not yet signed but I guess we could say they are legally separated. It’s a shame, but I hope no regrets as the journey was good for both: EMC got penetration into a wider market and made good headway in midrange virtualisation, Dell got storage that was enterprise-credible at a time when its reputation had some residual tarnish form early generations of PowerEdge. We still bear the scars, sometimes literally – the sharp edges in Gen 3 PowerEdge caused many a bloodied finger.

Now the world is different. Dell and EMC have headed for converged stack, and with different partners. The reason EMC went with Cisco is probably that Cisco were hungrier for the deal, while Dell has a relationship with Microsoft that was always threatened by the ties between its storage partner and Microsoft’s current arch-rival VMware. EMC / VCE has some really cool stuff though there is a big perception that UCS is a stalking-horse to get Nexus through the door. Cisco’s reputation in production support is up to the standard of EMC’s and VMware’s, they are very firmly pitched in the datacenter and will continue to play well there but I see a lot of innovation outside the datacenter, what happens in the DC is maturing rapidly, it’s the interfaces that are in flux and the routes in for hybrid cloud.

I chose Dell because in my view they have it right: they are going with best-of-breed components to create an open(-ish) management platform that works with multiple hypervisors, uses provisioning tools which were HP-first for a long time and have broad platform support, and they are rapidly expanding services so it’s a good place to be if you’re a techie with ambition. Having joined, and staring to get some idea of the direction of key technologies, the scope of acquisition planning and so on, I think I made a very good choice. I can see why others would choose differently and I am certainly going to miss my EMC friends because we had some good times together. We can be frienemies.

EMC VNX

 Vendors  No Responses »
Jan 182011
 

Today amid much trumpeting EMC launched a new storage device, the VNX series. It effectively makes the AX redundant, in my view, which is really no loss, and is pitched at a market that is price-sensitive, small-scale but reliable IP-connected storage.

The leader in this space, for my money, is Dell EqualLogic. Dell are EMC’s biggest reseller but not for much longer. The relationship has looked rocky since Dell’s (ultimately unsuccessful) bid for 3Par last year and isn’t helped by Dell’s recent acquisition of Compellent. It’s telling that the launch presentation for the VNX prominently featured its competitive price advantage against EqualLogic.

I am expecting any day now an announcement from Dell that they are adding NAS functionality to EqualLogic, because it makes no sense not to. That puts both on the same field as NetApp’s biggest volume sales I think, and both will be cheaper.

The significance is, I think, mainly that storage vendors are chipping away at the per-gigabyte costs which have thus far made desktop virtualisation a tough sell.

Backup means vendor lock-in. We hate that.

 Vendors, Virtualisation  1 Response »
Nov 302010
 

As I said before: backup sucks, move on. Here’s why I think the manufacturers of backup software are not keen to fix this.

We are looking at a rip and replace of our backup software. The new vendor is prepared to buy out the existing licenses giving us a cost-neutral solution; they will make their money form the professional services and the support contract that wraps around it. Fair enough.

We’re lucky: most of our data is not covered by [W:Sarbanes-Oxley] or any other form of compliance or regulatory regime. For most people the data that gets backed up most carefully is going to be the data covered by compliance. Which means you have a bunch of data in a proprietary format, in order to restore which you need to retain copies of the then-current software.

That’s not exactly optimal in the tape world, and you are horribly exposed if you have to recover the data – if you stop paying maintenance because you changed vendors then typically the software companies will want to charge you all the back maintenance right back to when you stopped paying before they will give you any support at all should you need it, so that’s a substantial exposure. It could even be more than the value of the software, with renewals running typically at 20% of box price at the moment.

But that’s manageable and it might never happen. What about if you backup to disk?

Let’s say you use a DataDomain appliance as your backup target (or a VTL or any other reasonably sophisticated backup target). In general these use block level deduplication or some other single-instancing methodology to keep down the amount of storage you use. Guess what: the new backups could be different bitwise even for the same data.

You could end up doubling your storage usage.

Or you have to stream off to tape, which brings the whole world of retensioning, media lifecycle and tape handling.

What does this add up to? Vendor lock-in.

We hate vendor lock-in

We. Hate. Vendor. Lock-in.

Like almost nothing else, we hate vendor lock-in. Why? Because it magnifies the consequences of a mistake. If I buy BackupWorker today but next year find that they’re not supporting some new platform or VMware version or something, I have massive pain to go through if I want to switch to NetVault.

And actually this is all caused by something that we should not be doing in the first place: writing proprietary stream data instead of image data. A VMDK file should be backed up as a VMDK file. There should be no restore.

Symantec, the Great Yellow Peril, actually have a product that more or less does this: BackupExec System Recovery. It’s quite clever and it will even do a fair job of offline P-V conversion for weirdly configured systems. It writes a mountable, bootable image. Symantec market this in the SME space and it would be silly money to deploy widescale, but it shows it can be done. Veeam do it too. There is no excuse any more for pretending that disk targets are all virtual tape, virtual tape sucks as much as physical tape. Block level is the way to go. “Virtual tape sucks too, move on!” as those nice people at [W:EMC Data Domain] put it.

Why will vendors not give us what we want? There has to be a suspicion that at least some of it is down to vendor lock-in.

If all backup software provided nothing other than a means for turning VMs and physical machines to a consistent disk image, be it [W:VHD (file format)] or [W:VMDK], or something else, then switching would be easy. They would have to retain our business with things like excellent support and rapid adoption of new platforms, software and versions. They’d have to become as agile as we want to be, or lose our business.

I can see why they might prefer to keep us locked in.

Addendum: hat-tip @EwanToo who rightly points out that some file container is a good idea for error detection, and proposes the POSIX TAR standard. This does make sense especially for archival but is consistent with the idea of retaining a vendor-neutral file format. For recovery you really would rather not have to untar anything, but it’s not so hard.

EMC: Solid progress but a way to go yet

 Vendors  3 Responses »
Nov 182010
 

@EMC has an image in the industry for being a rather arrogant company; I have always disputed that but there’s no doubt that it did need to take some time to show its customers that it had changed in this respect. Of course my opinion is coloured by being part of the Customer Council community, since it is here that EMC’s willingness to listen is at its most obvious. I have no doubt that some of my friends will think I’ve drunk the Kool-Aid as a result, I will disagree. I like to work with partners not vendors, any company where I have a significant engagement as customer or as supplier, I will try to open up communication as much as possible, share the details of what we are trying to achieve and let them know how we are doing. I definitely do not have a monopoly on good ideas or understanding of my or any other business. It makes good sense to me to play an open hand wherever possible. I can absolutely see that this would not apply to some other engagements, companies and markets, but it works for me.

So, a few things that people have asked [W:EMC Corporation] to do and they have done:

  • Remember you are a storage company and do not try to take over the world. Selling [W:Ionix] to [W:VMware] was absolutely the right thing to do. By all means mitigate your exposure in storage by part-holding software and services companies, but EMC does storage.
  • When you develop new stuff, focus on time to value and ease of use (which is why [W:NetApp] are eating your lunch in the unified space). Unisphere is so much better than Navisphere and Celerra manager.
  • Remember that a big chunk of your business now is with people who are not storage nerds. Again, Unisphere is a giant leap forward, so is FAST v2.
  • Focus on best of breed and stop competing internally. Announcements about convergence of [W:Data Domain] and [W:Avamar] are the obvious sign of progress here, but there are others. Clear messages and direction are good.

Here are some messages I think still need to be reinforced.

  • Prices need to come down, the drive is to more and cheaper storage. For example, the price of storage on midrange arrays is currently at least 4x the price of the same drives presented as DAS in servers, a gap which badly needs to narrow. There are innovative companies out there which are delivering shared storage using JBOD or near-equivalently priced means. We will pay for value received (e.g. centralised management, security, planning tools, storage on demand, fully automated tiering) but we won’t pay over the odds for storage just on the basis that it embodies the Clever we already paid for when we bought the array. Storage cost is crippling the TCO case for virtual desktop, so this is going to get even more important next year.
  • [W:TCO] and [W:ROI] are not just fictions we use to get money out of the business.  Your cost and benefit modelling tools are seriously in need of a dose of reality, and must include ways of measuring that the value is delivered. Some companies are already auditing ROI, this is going to get much more common as service oriented IT delivers greater and more granular cost transparency.
  • AD integration is not optional. True, some people will want and need a completely separate mechanism for access control on their storage, but the small and medium sector is not going there. We want to be able to sign in to the management interface on every single product using the same AD credentials, and assign roles and rights through AD groups. vSphere 4i does this incredibly well, buy their code for Data Domain and Unisphere.
  • Work with competitors to provide common interfaces. Heterogeneous environments are a fact of life, we will always change vendors from time to time, we will always use great point solutions in some areas. DPA is cool, open the APIs so Veeam can write to them, but it’s much bigger than that. SLAs are the KPI for most things now so let’s just get it done. There has been work done on XML / Web Services schemas for SLA compliance, this is a great way forward. Old-school frameworks like [W:Tivoli], [W:OpenView] and [W:Unicenter] have horrific running costs and permanently lag the market, open standards can allow proper comparison between disparate systems.
  • And, of course, PowerPath needs to be free. Especially PowerPath VE. We bought EMC storage for our virtual infrastructure, we paid a lot of money for it. We could have bought NetApp and not worried about storage path load balancing. PowerPath fixes a problem that is yours, not mine, the [W:CLARiiON] active-passive model. Asking me to pay, at last quote, £600 per socket for a patch to cover your internal kludges? Please.

So there we go. Praise plus brickbats, as Mr. Foskett recommends.

3Par, Dell, HP

 Vendors  No Responses »
Aug 312010
 

I could be completely wrong here, but I seriously wonder if Dell’s main purpose in bidding for 3Par is to hurt HP financially. There is little doubt that, good though 3Part is, the current bids overvalue it.

HP cannot, I think, afford to lose this. XP and EVA are tired and need massive investment to become anything like competitive. Hitachi won’t sell HDS (the more logical choice for HP) because that goose lays nice golden eggs, so the options are to acquire one of a small number of lesser players or spend a lot of money on a long overdue tech refresh.

Actually, though, the refresh is not an option. HP needs storage because that’s where the margin is; tin is commodity and there is little differentiation. Will that storage margin last? And for how long? I guess not. New storage delivery and access models, and the inevitable war on state prompted by cloud computing, will, I think, undermine the achievable margins in storage. So a long-term development program is more of a risk than a quick acquisition.

Dell has a competent and well-liked entry level product in EqualLogic and also for the enterprise OEMs EMC. Obviously reselling is less lucrative, but Dell customers like EMC, a brand which is among the strongest in this conservative market. For Dell to acquire 3Par would be massively disruptive. Maybe in a good way, but maybe also they would drive some long-term enterprise customers to other vendors.

So for me 3Par is a better fit in HP than Dell, as part of the product portfolio. Culturally? Not so good. But mainly, Dell can afford to walk away in a way that HP really can’t.