A perspective on the Cisco / Whiptail Acquisition

Cisco has acquired (or a least, announced the “intent to acquire” at the time of this writing) a “storage company”, Whiptail. There has been speculation that Cisco would enter the storage market through acquisition for years, just as there has been speculation on numerous other M&A activities that haven’t come to fruition. The question everyone seems to be asking is, what now? What does it mean for Cisco, their manufacturer partners, channel partners, and most importantly customers?

Whiptail is an SSD storage company. But before trying to gain some perspective on the acquisition and potentially answer some questions, its important to recognize that the next-gen storage market is broadly divided into two categories:

The first category of products are aimed at accelerating performance and latency sensitive applications with (relatively) low TB and (relatively) high IOPS requirements. Examples would be Whiptail, EMC’s XtremIO, Violin, Texas Memory Systems and of course the PCIe flash solutions like FusionIO. These companies are laser focused around use cases such as VDI, very high performance OLTP databases, HPC workloads, data analytics processing and so on. They are not targeted to replace general purpose storage arrays such as an EMC Unified array (VNX) or Netapp FAS which are swiss army knives that cut through balanced performance & capacity requirements with the same platform and also offer robust data services.

The other category of  next-gen arrays are designed to do exactly what the former doesn’t: leverage flash along with various efficiency techniques to compete with traditional mid-range storage arrays to serve up the general purpose capacity & performance and do it with a lower TCO. Comparatively speaking they target balanced TB and IOPS requirements. Examples would be: Pure Storage, Nimble, Tegile and others. These arrays are not targeting microsecond response times and millions of IOPS, but rather their purpose is to serve capacity & performance in a much more efficient manner than legacy arrays and without the management overheard of traditional storage constructs (I.E. RAID). Essentially they are looking to go after the EMC VNXs, Netapp FASs of the world by offering new architectures that can serve general purpose storage but be more perfomant (not extreme performance, just enough to remove it as a worry), be easier to manage due a ground up storage architecture design free of legacy storage constructs, and offer a lower TCO.

Whiptail makes references on its website to it being an “application acceleration” platform again and again, so its fairly clear where they are positioning themselves. Every use case on their website is to accelerate a tier1 application(s) with extreme performance requirements and thus they fall squarely into that category first category mentioned above.

Now lets hold that thought for a minute and explore how storage architectures are changing. Its a forgone conclusion that flash in the server is going to become ubiquitous as a “design principle”; the question is not IF but WHEN. We are seeing this take shape in the form of PCIe flash cards and their respective software stacks that offer caching, and DAS flash capabilities. This is something we are on the forefront of today, but I believe there will come a time when every server simply must have flash onboard just like today every server must have an ethernet port onboard. Today, flash in the server is also targeted towards accelerating tier1 workloads / performance bound workloads. It isn’t necessarily part of an overall data storage architecture, and the main reason is, with the exception of PernixData and vFRC from VMware, there just aren’t many software stacks to decouple the capacity & performance requirements. If you want to use PCIe flash today you have to be extremely tactical with where you use it. Imagine if operating systems didn’t have a memory subsystem that paged in / out memory segments but rather we had to decide manually which parts of process memory at what time needed to be in RAM versus swapped out to disk. Thats essentially where we are with server based flash architectures today (notable exceptions above). Clearly, this is not the way of the future. Server based flash needs to simply be another default component and the software drivers should included in operating systems like they are for ethernet cards. That is to say, server based flash shouldn’t be thought of as a “optional storage architecture decision” but as much of a no brainer as dual power supplies; of course, this will take time.

How about the way Cisco entered the “storage market”? If we compare this to how they entered the compute market (UCS) or are entering the SDN market (Insieme) we can see how Cisco behaves when they really want to disrupt a market. This acquisition by comparison doesn’t have that same feel to it, and IMO, if they really wanted to disrupt the enterprise storage market they could have gone after some other storage players that would have a much bigger impact on eating away at the enterprise storage TAM. And this relates to the discussion around the two types of next-gen array products on the market today and where they play: focused high-performance use case (lower TAM) versus more general purpose (higher TAM). However, lets also keep in mind that Cisco is a sales company, and sales reps will do what they are paid to — sell everything they possibly can to a customer. So we will have a bit of that which naturally will create some tension between Cisco and their storage partners in the field. Although it’s critical to understand use cases when positioning any solution with customers, what happens in the field is complex and not always so black & white. The “best” or “most correct” technical solution doesn’t always win, thus we will always have SOME overlap. Even with that said, and acknowledging that there will be SOME overlap, I don’t think the folks in the board rooms of Cisco and their storage partners are preparing for battle. What goes on in these board rooms and the resulting partnerships & relationships are complex and not as simple as technology A vs technology B on a line card.

Based on the above, it would stand to reason that the primary reason Cisco went and purchased Whiptail is to accelerate & augment UCS sales, at least initially. Let’s just take VDI as an example. The compute (UCS) portion of a VDI deal can mean big $$$ for Cisco. However, storage is still most expensive component of VDI when “legacy” storage is positioned, and in my experience it’s what tends to hold up the deals the majority of the time. The Whiptail acquisition will allow Cisco to remove that roadblock and using the IP package up either server based flash or a purpose built appliance just for the VDI environment and be 100% self-sufficient and control the deal. I’m guessing Cisco got sick of millions of dollars of UCS being held up due to traditional storage costs delaying projects.

If we look at the penetration of PCIe flash into UCS with the current offerings, it hasn’t been great. Although we are in the beginning stages, Cisco has to rely on its OEM partners heavily to integrate this into their architectures. If we work on the premise that server based flash is a forgone conclusion, does it makes sense for Cisco to rely on an outside force to execute on that vision on their flagship UCS platform? I don’t think so– UCS is a big bet for Cisco and they can’t afford to let anything get in the way of that. No more will server based flash be a bolt-on option, but just “part of UCS” and a competitive differentiator. And what’s huge about this is that they have acquired the software IP to make it a reality — weather they source the hardware from elsewhere or not (Whiptail is COTS hardware after all), the software is the differentiator and the most difficult part of the problem to solve. As I’ve also seen in the field, Cisco struggles with how to architect solutions with PCIe flash in their UCS systems, and now they have acquired some DNA and brought that entire knowledge “in house” to help their customers. If server based flash is going to become ubiquitous, they need that DNA in house — this is especially true as customers look to leverage VMware’s VSAN or EMC’s ScaleIO and other next-gen storage architectures and Cisco will need to differentiate with UCS versus other server platforms to avoid being 100% commoditized. Similarly, if we start to look at other performance use cases like HPC, Cisco can walk into a customer with a much more holistic and differentiated solution that is going to meet all their performance requirements without being hamstrung by other vendor partners for storage acceleration.

How exactly will Cisco build a server flash tier into their UCS system remains to be seen — they could implement it blade by blade, or perhaps a dedicated blade, or even a dedicated chassis — but there will be an all flash memory tier in UCS, someway, somehow. It wouldn’t be far fetched for them to offer an option of UCS similar to a Simplivity or Nutanix style architecture as they are seeing pressures from these “hyperconverged” players. Although they would need a ton of software help to create the solution. the design options are endless now that they own the IP. I think they are well positioned to leverage the flash to target next-gen applications and architectures as well as use it to bridge current legacy application designs while building a bridge to the future for customers; all in one platform.

From a customer perspective, it’s all about simplicity and ease of support. Customers today are more and more demanding vertically integrated solutions, especially when it comes to VDI or HPC projects with large scale. The success of VCE is a great example of that. Today, for a VDI project, a customer has to buy UCS, then figure out weather to include server based flash or not plus some combination of external storage with flash or HDD or a combination and then go on to worry about interoperability and performance benchmarking across a variety of vendors….and we haven’t even got to the actual application architecture of VDI yet! With the Whiptail acquisition Cisco can walk in with an “all Cisco VDI solution” that is ready to go and supported 100% by them and control the quality of the customer experience in a much tighter fashion.

From a competition perspective, certainly Cisco has a tight partnership with VCE and they too offer a VDI in a box in the form of Vblock, and now Cisco could offer its own without EMC’s or VCE’s help. And unless Cisco decides to specifically NOT sell the Whiptail gear as a pure storage array and only use the IP to integrate flash storage into UCS, I don’t see how it won’t be competitive with EMC’s XtremIO from a positioning perspective. But in terms of general purpose enterprise storage? I think that’s business as usual…..for now. To think that field conflict is not going to occur is being oblivious to the realities of sales. But in the end, regardless of what battles the vendors engage in, the customers will end up winners.

Regardless of what may or may not transpire in the next 12-18 months and beyond, its a great time to be innovative in the storage business.

Categories: Cisco, storage, Whiptail

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: