Emc logo

Vblock Review
1st Year in Review - pre-sales vs. reality


Well we have just passed a year of Vblock ownership and the last year has passed rather painlessly.

Our Vblock was one of the first out there, delivered in November 2011. I wanted to provide some pros and cons of Vblock ownership. Some of the themes are not Vblock specific, but worth bearing in mind because there will always be a gap between what you hear from pre-sales and what the reality is.

Pros:

VCE – The company has been constantly improving which is good to see. Not content to rest on their laurels, they really have grabbed the bull by the horns and they are innovating in a lot of areas.

Vblock – The concept of the Vblock itself deserves a mention. VCE are definitely on the right path… it’s like the first generation Model T Ford. I’m sure old Henry had hundred’s of suppliers that provided the components for his Model T and he came along with the assembly line production and he put it all together. This is like what is happening over at VCE. Over time I’m hoping that the integration between components will become more and more seamless as the demand for pre-configured virtualisation platforms grows and grows and the designers behind each of the components are forced to work closer together.

Management and Support - If you have a bloated IT support team in large sprawling organisation, a Vblock can help reduce your head count by simplifying your environment. One thing converged infrastructure platforms are good for, is breaking down the traditional support silos with regards to storage, network, compute, virtualisation. When all the components are so tightly integrated, your silo’d operations team morphs into one.

Compatibility Matrix – This has to be the biggest selling point in my book. Taking away the pain of ensuring compatibility between so many different components. The VCE matrix is far more stringent than individual vendor product testing and therefore far more trust worthy. Try getting a complete infrastructure upgrade over a single weekend across storage, network, compute and virtualisation components through your change management team. It’s not going to happen unless it’s been pre-tested.

Single line of support – Being able to call a single number when there is any issue, immensely simplifies fault finding and problem resolution. Worth it alone just for this and the matrix.

Single pain of glass – This is where UIMp is starting to come into its own. It’s been a long road, but the future looks good. VCE’s goal is to replace each of the individual management consoles so that VCE customers can use UIMp for all their automated provisioning. When it works, it really does simplify provisioning.

Customer Advocate – In my experience the customer advocate offers great value. Extremely useful when managing high severity incidents and ensuring your environment remains up to date and in support, with regular services reviews and providing an easy path into VCE to organise training sessions, bodies to fill gaps in support, provide direct line of contact to escalation engineers and just deal with any queries and questions you may have about your environment.

Cons:

The AMP – the major design flaw in the AMP for me is the 1GB network. Data transfers between VMs in our 10GB service cluster can achieve 300 Mbps; as soon as the AMP is involved it drops to 30Mbps. Really annoying and what is in the AMP' vCenter, which is used to import virtual machines. Let’s say you are doing a migration of 1000 VMs for example… that 30Mbps is going to get really annoying and it has.

Cost – The Vblock hardware isn’t so bad, but what really surprised me is the amount of and cost of the licenses. Want to add a UCS Blade' No problem, that will be £5k for the blade and about £3k for the licenses – UCS, UIMp, VNX, vSphere,  etc. It all adds up pretty quickly. Ensuring you adequately size your UCS blades up front, i.e. plenty of memory and CPU is really important.

Management & Support – Converged Infrastructure Platforms require a lot of ongoing support and management. This is an issue not limited to VCE. It’s just the nature of the beast. If you have  an immature IT organisation and have had a fairly piecemeal IT infrastructure and support team up until now, you will be in for a shock when you purchase a converged infrastructure platform. There’s no doubt a Vblock is an excellent product, but it’s excellent because it uses the latest and greatest, which can be complex. It also comprises multiple products  from 3 different vendors – EMC, Cisco and VMware, so you need the right skillset to manage it, which can be expensive to find and train. It takes at least a year for someone to become familiar with all components of the Vblock  You’re always going to have employees with core skills like virtualisation, storage, network, compute, etc, but you do want people to broaden their skills and be comfortable with the entire stack.

Integration between products – See above, multiple products from 3 different vendors. At the moment the VCE wrapper is just that, little more than a well designed wrapper, lots of testing and a single line of support. Ok, so EMC own VMware, but it seems to make little difference. EMC can’t even align products within their own company, how on earth can they expect to align products with a subsidiary'  If the Vblock is going to be a single vendor product, then all 3x vendors need to invest in closer co-operation to align product lifecycles and integration. VMware release vCenter 5.1 and Powerpath have to release an emergency patch to support it' Going back to my Model T analogy, the Vblock is never going to become a real Model T until Cisco buys EMC or EMC drop Cisco and start making the compute\network components. Not so far fetched.

Complexity – The VCE wrapper hasn’t changed the complexity. (This is the same with HP or Flexpod.) This is another myth. “We’ve made it simple!”. Er, no, you haven’t. You’ve just done all the design work and testing for us. Until the integration above takes places, which will allow for simplification of the overall package its going to remain just a wrapper and it’s still going to remain an extremely complex piece of kit. VCE have focused efforts on improving UIMp to simplify Vblock provisioning and to simplify Vblock management through a single interface but really these are just band aids if the individual components are made by separate companies.

Patching – Even though there is a compatibility matrix, which does the integration and regression testing for you, it still doesn’t take away the pain\effort of actually deploying the patches. Having a Vblock doesn’t mean there is no patching required. This is a common pre-sales myth, ‘Don’t worry about it, we’ll do all the patching for you.’ Sure, but at what cost' Security patches, bug fixes and feature enhancements come out more or less monthly and this has to be factored in to your budget and over time costs.

Monitoring and Reporting – This is a pain and I know there are plans afoot at VCE to simplify this, but currently there is no single management point you can query to monitor the vitals of a Vblock  If you want to know the status of UCS: UCS manager, VNX: Unisphere, ESXi: vCenter, etc. For example, you buy VCOps but that only plugs into vCenter, so you are only aware of what resources vCenter has been assigned. To get a helicopter view of the entire Vblock from a single console is impossible. UIMp gives you a bit of a storage overview: available vs provisioned, but does not give you much more than that. So you end up buying these tactical solutions for each of the individual components, like VNX Monitoring and Reporting. Hopefully soon we will be able to query a single device and get up to date health checks and alerting for all Vblock components.

Niggles – There have been a few small niggles, mainly issues between vCenter/Cisco 1000V and vCenter/VNX 7500 but overall for the amount of kit we purchased it has not been bad. I think a lot of these issues had to do with vCenter 5\ESXi 5. As soon as Upgrade 1 came out, everything settled down. Note to self don’t be quick up upgrade to vCenter 6/ESXi 6!

Disclosure: I am a real user, and this review is based on my own experience and opinions.
3 visitors found this review helpful
B6bd6286 973d 4ac6 a9b1 ee1fe77f0838 avatarAndrew sventek li?1416341496Anonymous avatar x30

6 Comments

B6bd6286 973d 4ac6 a9b1 ee1fe77f0838 avatar
PatrickITOPsReal User

Gareth, what an amazing and thorough review! You bring up some very good points to consider for these "beasts" but the overall review seems positive. Thanks!

Like (1)02 March 13
Picture eric dirst
Eric DirstReal UserTOP 5LEADERBOARD

Thanks for all the details Gareth. This will help greatly with our evaluation of the product.

Like (1)06 March 13
Picture trevor roberts jr

Hello Gareth. Thank you for the great and thorough Vblock Review. There were a few points that I disagreed with, and I tried to summarize them as best I could below...

DISCLAIMER: I am a VCE Employee, and this comment is published with permission from IT Central CEO: Russell Rothstein

The AMP as a design flaw?
It should be stressed that the AMP is not meant to be in the data path at all. While it's OK to use the vCenter to import the occasional one-off VM OVF\OVA, for large-scale migrations (the 1000 VMs mentioned), using the standard vCenter import functionality is not advisable.

Tools such as RecoverPoint (possibly in conjunction with SRM) should be utilized. Even if vCenter had a 10 Gb pipe, I don't think you would want to migrate that many VMs with a vCenter import. For a free alternative, vSphere Replication may be appropriate. I haven't used it before, but I'm pretty sure it does not have vCenter in the data path either.

VCE as just a wrapper?
The VMware 5.1 and PowerPath incompatibility problem was avoided by our customers, provided that they were following our Release Compatibility Matrix (RCM) :-). If customers upgraded to vSphere 5.1 before VCE published the RCM that includes it, that is out of our control.

That being said, VCE makes significant investments in R&D to test all the components of the Vblock in an integrated fashion so that we can eliminate the majority of interoperability issues that customers face when they follow individual matrices from VCE's Investor Companies.

True, upgrades are non-trivial (it's the same for any stack), but we eliminate some of that pain by:
a. publishing which combination of firmware\software levels you should go to next
b. publishing an upgrade guide that lists the steps in which components should be updated
c. the Customer Advocates being available to advise the customer on how to proceed.

You mentioned these advantages, but I cannot underscore enough how critical these capabilities are in time and OpEx savings for our customers.

Monitoring and Reporting Concerns?
VCE understands that a single view of the entire platform, regardless of whether it is running virtual workloads or bare metal, is necessary. That is why we developed VCE Vision Intelligent Operations to provide a single view of the entire infrastructure regardless of the management platform.

So, for example, you can see your Vblock components correlated with your virtual infrastructure within the vSphere Web Client with our new plugin. UIM and other management tools will soon be leveraging VCE Vision to improve their Vblock reporting capabilities.

If you'd like more info, I wrote a brief article about it here: http://vmtrooper.com/vce-vision-intelligent-operations-i-see-a-vblock/

...and we have even more functionality on the horizon. Stay tuned and keep giving us this kind of feedback!

Like (1)06 March 13
Picture 1508 1362053212
kleegeekReal UserTOP REVIEWERPOPULAR

Thank you Gareth for this fantastic review. I really appreciate the depth of thought that you put into this review, and will be referencing this directly in presales meetings that I am in.

Like (1)18 April 13
Picture 1549 1362044785
Gareth LloydReal UserPOPULAR

Glad you enjoyed the write up guys. I'd like to add a few bits, as on reflection it may sound like I was VCE bashing a bit.

Just to clear things up: I love the vBlock

Why?

What really matters in any organisation is the product and the support - with regards to VCE I cannot recommend them enough.

What I think any organisation wants is a relationship similar to what Microsoft have with most organisations. If you're a Microsoft shop with Windows XP\7 clients, Active Directory\DHCP\DNS, Exchange, SQL, File Services, Office, etc your main support point for software is going to be Microsoft Premier Support. Likewise you would want a similar relationship for your hardware stack. You don't want to have to handle multiple different contracts and act as the middle man between the different vendors during problem resolution.

VCE single line of support takes care of this for you by offering a single point of entry for your hardware stack. There will always be products outside VCE and there are obviously still EMC products that are VCE exceptions but the list is getting shorter and shorter as the VCE product range grows. Single line of support was a stroke of genius.

I have a great relationship with our Customer Advocate and the Support guys in Cork. Thanks to Trevor for reminding us all about the Pros of vBlock ownership, it is much easier to criticise than to praise.

So one of the main reasons I wrote the article was to make people aware of some of the pitfalls to the big promises around converged infrastructure you might hear. I'm sure this is the same in all organisations but I don't enjoy the sales side, I feel they oversell the product when they could just as easily make sales on the merits. We actually had a running joke during the pre-sales period, "Don't worry, its in the vBlock", as any concern we raised or any question we had, garned the same response. I've seen organisations NOT pick vBlock for this very reason. Maybe it works for most organisations, but for more experienced, cynical, IT staff, big promises on how a vBlock is going to fix every problem you have is unrealastic and a turn off.

Its easy to get blown away by the pitch and get so excited, you stop to ask basic questions around the correct fit for your organisation and the ongoing costs. Anyway I think I have made my point - size your vBlock realistically up front, expect to patch it every 3 months like regular kit (but expect less pain because its been pre-tested) and you will get a much better idea of the capex and opex costs.

Re: Trevor's Comments

My post was written before the public release of Vision Intelligent Operations so I didn't want to say anything, but I am really looking forward to this. It plans to be a monitoring game changer and I cannot wait to integrate it with VCOps to give us a total vBlock view for our operations team. This has been the missing link in the Operations department. Its a huge leap forward being able to monitor and capacity plan at all our data centres and will hopefully avoid us having to buy tactical solutions like VNX Monitoring and Reporting.

I could have used better examples, but I think my points are valid. Whilst VCE adds tremendous value and mitigates many of the problems facing customers in such a fragmented industry, I would really like to see more consolidation in the hardware arena, so that one day your entire stack comes from a single vendor with a solution that has been designed from the bottom up to integrate together. Every component of the hardware stack is separately going through its own development cycle, whether by separate companies, or separate teams in the same companies. It would be great if they were all aligned and there are some components which I would like to see removed all together. The 1000V is a prime example. It is a support DMZ. The network team don't want to touch it because its virtual, the Virtualisation team don't want to touch it because its a switch. Its an endless source of pain. Some times I just want to go back to VMware distributed switches.

re: The AMP. vCenter is your access point into the vBlock. In our recent migration, we restored a couple of hundred VMs through various methods - manual OVF import, VEEAM Backup and VMware Converter and that is why I raised the point. Not everybody has SRM & RecoverPoint\vSphere Replication and even though we do, it was not suitable to import Virtual Machines. This problem isn't going away.

Further more, the AMP is your management cluster, so by default (as in installed by VCE) its going to contain: vCenter, VUM, 1000V Management VMs, SQL, UIMp, a jump box and an EMC support server.

What else could be housed in the AMP? Well the first thing we added to our AMP was a domain controller. Why? Well the main premise of the AMP is to allow you to manage your service cluster if it goes offline and if all your domain controllers are in your service cluster, your AMP (and vCenter) is going to be pretty useless when all your domain controllers go offline. Then you'll probably, like us, deploy VCOPs, EMC SRS, SRM, vCloud Director and vShield Manager amongst others.

The first thing we had to do was upgrade our AMP rackmount server RAM from 48GB to 96GB and double the storage because we ran out of space. So another point, review what you will put in the AMP and how many resources you need. VCE like to think the AMP is just for the software they install at build but that's a little naive. Any customer is going to use it for all their VMware and EMC management software whether it is delivered by VCE, EMC or VMware professional services.

So there is going to be quite a lot of kit in there and we really noticed the connectivity difference between the service cluster and the AMP in our deployment. It would be beneficial to us if the AMP network was 10GB too and I'm sure it will be the same for other customers.

Like (0)29 April 13
Picture trevor roberts jr

Gareth's review is honest, open, and the kind of feedback that our Sales Teams need to bring back to Engineering. I didn't take the article as VCE Bashing, and I hope my response wasn't interpreted as "defending" VCE. Rather, I wanted to explain to some of the thought behind our design decisions.

Regarding the AMP, VCE's intent is to give a low-cost management infrastructure, separate from production per VMware's recommendations, to run the critical management applications only. Some larger organizations may want to pay for the added options of the Cisco VIC or 10 Gb CNA\NIC and a pair of 10 Gb switches to accommodate bandwidth requirements for additional management components. Customers may also want flexible virtual networking options. In all such cases, we need to hear that from you guys to influence future AMP designs.

Regarding the Single vendor stack, VCE doesn't design all its parts, but our great Engineering organization gets us pretty close to that point. Bear with me as I shed some light on the great work that those ladies and gentlemen do.

The Design Team takes the best components from industry. Instead of relying on manufacturer best practices, the team works with their counterparts at the Investor companies to truly understand how the components work. Then, they design the Vblock so that the components work together in the most optimal way. In addition, for each Release Certification Matrix (the big table of which component firmware versions match up), additional time is spent making sure that interoperability issues are not experienced by the customers. All these decisions are then vetted and put into exhaustive testing by our QA Teams.

The aim is to prevent customers from having to think about the individual components and instead focus on the Vblock as a stack of optimized resources for critical applications. I'll break out the car analogy: When you purchase your car from , customers tend to care more about transportation rather than the components underneath (other than they are good quality).

The manufacturer may source the brakes from Brembo, the clutch and flywheel from ACT, the stereo from Bose, etc. However, what the customer gets is an optimized product after the source components are fully examined and engineered into a single combined product (typically, within ~45 days of ordering already racked and stacked in the case of Vblocks)

Now, there are customers who do care about individual components, as Gareth pointed out, and these concerns are based on real-world pain points. The Engineering Teams are always looking for feedback to make the Vblock better with each release. Keep the feedback coming, and keep your VCE Sales Reps in the loop too.

Like (1)30 April 13
Anonymous avatar x30
Guest
Why do you like it?

Sign Up with Email