Ahead IT

This is Part II in a Four Part Series discussing VCE Vblock and NetApp FlexPod technology and marketing approach.   Links to the other articles can be found at the bottom of this document.

In the introduction to this series, I outlined a number of considerations that typically play into conversations around whether a customer should look at building a “stack” of infrastructure compute. The most heavily marketed and most mature “stacks” today are the VCE Vblock and NetApp’s FlexPod. Prior to digging into the questions around flexibility and investment, I wanted to take a couple of minutes to talk about each infrastructure stack. Specific focus areas include: what they are, how they are similar, how they are different, and what are the advantages to either approach.

At the highest level, it’s important to understand a key differentiator in the difference between NetApp’s approach with FlexPod and VCE’s approach to Vblock:

• NetApp’s Flexpod is a Reference Architecture – a set of designs and pre-tested / validated architecture design documentation from which customers can build their own stack (think blueprint)
• VCE’s Vblock is a Product – Vblock’s are pre-designed, pre-tested, pre-architected, pre-built, and shipped as a customer consumable product (think pre-fabrication)

As you will see, while many of the architecture components are similar (with the exception of the storage of course), how the various technology components are deployed and how flexible they are largely depends on the reference architecture vs. product support. Each company has chosen a specific path targeting specific use cases as outlined in the previous post – much of this will be uncovered as we continue to discuss.

Let’s start with defining a “stack” which is typically a unit of datacenter infrastructure that includes networking, compute (CPU/Memory), storage, and likely an operating system or hypervisor. Stacks have been getting a lot of press lately as companies market their visions of private cloud. Stacks are sold as the great enabler to private cloud, otherwise known as datacenter as a service, or as an internal “infrastructure as a service” (IaaS) offering. To understand the differences and similarities between the FlexPod and Vblock approach to stack, let’s take a look at each of the layers of the stack from the top down.

OS / Hypervisor

Both VCE and NetApp have “gone to market” with VMware vSphere as the core operating system of their respective stacks. VMware vSphere and its corresponding products (vCloud Director, vShield, etc) provide the virtualization layer that has become the trademark of IaaS allowing IT organizations to massively consolidate workload on fewer and fewer servers while being able to implement multi-tenant security, virtual machine management, and ultimately an evolving self-provisioning portal with vCloud Director. vSphere, as a hypervisor, works equally well on either platform as both use the same compute infrastructure.

The primary difference between the VCE and FlexPod approach to OS / Hypervisor is that the FlexPod approach seems to be more open.

• NetApp has recently released documentation and reference architecture material for Microsoft Hyper-V for virtualization on the FlexPod in addition to providing reference architecture for bare metal operating systems.
• The VCE approach originally was targeted for pure virtualization with VMware as the only supported Hypervisor / OS. Recently however, VCE has announced support for limited deployment bare metal workload based on real or perceived limitations (technical, licensing, or political) around virtualizing specific workloads.

While Hyper-V is trying to make a move into the core enterprise datacenter, VMware arguably continues to hold and maintain the lion share of the enterprise server virtualization market.

For customers who are bought into the VMware virtualization story – this category is a technical dead heat between FlexPod and VCE. For customers with Hyper-V or other requirements, FlexPod carries a slight advantage due to its flexible architecture approach.

Compute Infrastructure

Both VCE and NetApp have “gone to market” with Cisco UCS as the “compute” or “server layer” for their respective stacks. Cisco UCS has been on the market for 2+ years now and is quickly becoming a $1B annual revenue business for Cisco. Cisco’s innovative UCS infrastructure brings significant advantages to a virtualized infrastructure including:

• a “wire once” virtualized infrastructure scaling to up to 160 physical hosts
• tremendously scalable, low cost server memory architecture
• converged networking, VN-link, and top-to-bottom Nexus support
• virtualized interface adapters (Palo) which provide up to 55+ NICS/HBAs to each server,
• a common infrastructure management component (UCS Manager) to manage the infrastructure from a single pane of glass – up to and including implementation of a stateless compute platform using server profiles.

There are many articles available on the merits, advantages and differences that Cisco UCS brings to the table. They key for this article is that both VCE and NetApp leverage Cisco UCS as the core compute layer of their infrastructure and UCS in and of itself does not differentiate either platform. The clear winner in this category is neither FlexPod or VCE – it’s Cisco – as they continue to grow marketshare against HP, IBM and Dell regardless of stack decision.

Network / SAN Infrastructure

Similar to the compute layer, VCE and NetApp have effective parity when comparing network and SAN infrastructure when considered as part of their respective stacks. Specifically, the networking and SAN components of both architectures are remarkably similar. Both start with the embedded FCOE enabled by UCS which connects each UCS chassis (and server) over 10GB FCOE to the Fabric Interconnect Modules (FIM) (network/SAN switches) from which “external” connectivity to northbound network and storage infrastructure is established. When you compare the reference architecture of the FlexPod to the VCE Bill of Materials, you will see a couple of slight differences.

• The NetApp Flexpod approach currently leverages a pair of Cisco 5548 switches to provide northbound IP connectivity from the UCS FIMs. The 5548s also provide FC SAN connectivity if required. Effectively the NetApp storage infrastructure connects to this aggregation switch in which both traditional Ethernet and Fiber Channel are used.
• The VCE Vblock approach currently leverages a pair of Cisco 5548 switches to provide northbound and storage IP connectivity. The Vblock also includes a pair of Cisco 9148 SAN switches for connectivity from the FIM to the FC storage front end, again using traditional FC connectivity.

In both cases, larger northbound switch infrastructure as well as larger FC switch infrastructure is available as an option or enhancement to the reference architecture. The requirement for more ports (FC and Ethernet) will be largely driven by ultimate scalability requirements of the infrastructure.

As technology continues to mature and multi-hop FCOE becomes adopted as more of a datacenter standard, I fully expect to see both the FlexPod reference architecture as well as the VCE architecture to converge (eliminate separate FC SAN requirement) and ultimately offer FCOE connectivity throughout their solutions from server to storage.

The other difference specific to networking within the two architectures is focused on the management network layer. In the VCE product design, there is a separate management network deployed as part of the “Management Pod”. This dual-switch network provides an IP aggregation point for all of the management ports for all of the Vblock components, from which it can be connected to the 5548 and subsequently northbound to the customer IP network. In the FlexPod approach, each management port for various components could connect directly to to the 5548s (using expensive 1/10GB ports) or directly to the customer network (core or management.)

From a pure technology perspective, there is no clear winner – as both stacks leverage nearly identical technology for networking and SAN. The VCE approach to the management aggregation layer is a nice touch because it allows the Vblock to be a fully contained unit leverage lowest cost port technology. Overall, FlexPod may ostensibly carry a (temporary) slight price advantage because it does not require the separate FC SAN switches or management aggregation layer – but as previously mentioned, pricing is subjective based on company and customer.

Management Infrastructure

I include Management infrastructure as a component of the stack because VCE brings a management infrastructure component to the Vblock that it considers to be a differentiator. VCE provides a EMC branded product – UIM – which is part of the EMC Ionix management suite. UIM is ultimately intended to provide a common pane of glass for both reporting and general provisioning across all Vblock components with hooks into higher level management systems and lower level element managers.

The idea behind UIM ties to the basic VCE precept that the Vblock is an appliance and as such it should provide a common interface from which reporting and general work can be performed, relegating the element managers to specific engineering and configuration roles. The UIM and other management tools run on the aforementioned “Management Pod” which consists of dedicated hardware, storage and networking gear located inside the Vblock separate from the customer compute and storage infrastructure.

As of the time of this writing, I have not seen that the NetApp FlexPod approach brings any specific management infrastructure (software or hardware) to the equation as part of their solution, instead relying on native element managers or enterprise systems for configuration, provisioning and reporting.

In this category, Vblock wins over FlexPod because it carries the advantage of added functionality (specifically software functionality that usually resonates with customer executives.)

So, for a quick review between NetApp FlexPod and VCE Vblock so far:

So – what’s left?

Storage Infrastructure

What drives the ultimate technology difference between VCE’s Vblock and NetApp’s Flexpod? Quite simply – it’s the storage infrastructure. It is no surprise that there are two main stakeholders and sales forces who are driving the FlexPod vs Vblock message. Cisco and VMware both win regardless of the stack ultimately sold and deployed – so it comes down to EMC storage vs. NetApp storage.

There are two different approaches taken by VCE and NetApp specific to fighting the storage battle. NetApp will focus almost exclusively on array functionality and integration between NetApp and VMware. Focusing on features like deduplication, overall storage efficiency, unified storage (FC, NFS, iSCSI options), high order read cache (FlashCache), system performance, and overall virtualization integration as a normal storage campaign seems to be the norm.

VCE in general takes a different approach to selling the Vblock, with more of a focus on delivering an integrated technology solution that meets specific use case, infrastructure requirements and application / performance SLAs.The VCE (and EMC) sales teams will absolutely roll out and discuss the technology differentiators included in the Vblock storage options – but at an executive level, the discussion is less likely to compare speeds and feeds. VCE takes a more general high level approach that the Vblock is an appliance that includes purpose-built storage for virtualization that is part of an infrastructure designed to deliver a performance SLA. When asked – or as part of a Vblock deep-dive, you can certainly expect to hear about unified storage (FC, NFS, iSCSI options), FASTCache, FAST (fully automated storage tiering, storage efficiency, IOPs performance (especially given the power of their leapfrog VNX platform) and overall virtualization integration (maybe not in that order!)

Notice the similarities? The purpose of this article is not to rip apart the NetApp FAS approach nor the EMC storage approach with VNX or VMax or provide a side-by-side storage function/feature comparison. Some virtualization workloads will absolutely function extraordinarily well on a NetApp storage array using NFS with full deduplication and FlashCache utilization. Some workloads will absolutely run better on EMC storage technology by taking advantage of significant write cache (or FAST cache) capabilities with FAST tiered storage. For a pure “stateless” infrastructure perspective – boot from SAN is a clear requirement (iSCSI or FC) – most industry watchers would recommend that an EMC solution for FC boot from SAN will be better supported than a NetApp FC implementation.

Storage differentiators do matter. The storage differentiators that each vendor bring to market play to general and specific work loads and use cases. Key to the selection process is to evaluate your application requirements and then determine which storage platform best meets those requirements. That being said, if either vendor’s solution turns out to be basically “fit for purpose” for providing storage to a highly virtualization infrastructure – the final decision should likely come down to analysis of the whole stack (and it’s approach) vs. just the storage component.

Ironically it would seem that for most customers, the storage decision is in many cases already made at either a technology or a strategy level prior to engaging in a stack conversation. Long-time NetApp customers are going to be more open to a FlexPod message. Long-time EMC customers are going to be more open to Vblock. The question is – will either company’s approach help land their footprint as competitive beachheads at their competitor’s accounts?

Which approach (reference architecture or product) will win over the hearts and minds of executives and technologists and by their choice help the customer meet their business and application requirements in the most cost effective, timely, and technologically effective model?

The question is – why? And more critically – what are the differentiators that customers must compare and contrast when deciding if:

1. Is a stack right for me? And…
2. Do I stay with what I know … or go with something new?
3. What does going with something new mean from a flexibility and support perspective?

Links:

• Let’s Build a House: Prelude to Vblock or FlexPod
• Vblock or FlexPod Part I
• Vblock or FlexPod Part II
• Vblock or FlexPod Part III
• Vblock or FlexPod Part IV