On my ZDnet blog I had some fun with the Cisco announcement. In Cisco’s bong-sized cloud I opined that the 32 server limit of VMware’s VMFS means that
The platform they’re promoting is tiny by cloud infrastructure standards. Perfect for a dorm room – but not anywhere near close to what an enterprise hoping for economies of scale or management would want.
Scale: pay me now or pay me later. VMware chose later – and now “later” is demanding payment.
Interesting comments
The post got a couple of interesting comments that touch on Cisco’s messaging to their “coopetition” and users. Caspianhiro said, in effect, that Cisco’s message to a coopetitor is that this is just a logical brand extension:
I work for one of their partners with a competing blade. Cisco tells us that they are making the blade for a niche set of customers, especially in telecom, that want an all Cisco solution set for particular applications. The volume supported the business case. We don’t expect it to be price or feature competitive for business use, and it’s probably not going to fly anyway. But, if you are Cisco, and you know you can sell 100,000 of them a year without a lot of work, then why not?
Also, they specifically mentioned that HP is stealing market share from Cisco in the networking space, and they wanted a way to fight back.
AFAIK telco’s were not called out in the announcement conference. That said, this less ambitious strategy seems more realistic than a frontal assault on HP’s and IBM’s blade business.
Rpwillia0 said he heard a different message from Cisco.
Cisco in a recent presentation was talking like they were the only ones who could really figure out how to leverage the Nehalem processor and use all of that cheap memory.
That has the ring of authenticity about it. Engineers always just know they can do it better, and Cisco has plenty of smart engineers. Of course, that’s before they get into the details.
The StorageMojo take
Combined with the inexplicable purchase of Pure Digital Video for $590 million – Kodak’s market cap is less than $1B, Cisco could own a well-recognized brand AND have salable assets for 2x the Pure deal – one of 2 conditions hold:
- Either Chambers is way-y-y smarter than everyone else, or
- He’s desperate and hoping he’ll get lucky.
I’m rooting for the first, but thinking it’s the second.
Courteous comments welcome, of course.
In Dec 2008, Cisco was talking about “medianet”, where they believe in the near future, video will use up a significant portion of the Net. (Extrapolate from youtube’s inception?).
To prepare for this, Cisco tells us we need better tools and equipment not only to switch, route and regulate, but also to (dynamically) transcode, and post-process.
Pure Digital Video could be the investment to beef up Cisco’s ability in their belief in video.
VMFS is probably being improved in 4.0, and you don’t have to use VMFS anyway. In fact, some storage vendors would probably prefer that you use RDM mode since the moves the storage “smarts” back into the array instead of ESX.
The PureDigital purchase can only really be explained by hoping that people will start shooting everything in HD and upload it to YouTube thereby flooding the networks and forcing carrier upgrades.
Ivan, it seems like a very expensive and poorly thought-out plan. I would rewrite Robin’s second condition as:
2. A clever underling convinced Chambers it was a brilliant idea. In 12-18 months, after Chamber’s realizes Cisco need not be in the cam business to own video, that underling will quietly “go on to pursue other business opportunities”.
Robin,
You are right … it does not take much to do an Intel ‘blade’. It all now depends on how the virtualization and the long awaited VDC-OS deal with storage. To do this properly Cisco will need to provide native FCoE interface to storage … but they don’t own storage. So, all they can do is to complicate the issue with very expensive 10G Enet to FC bridging within their switch. How does this eliminate FC to deliver the hype about this ‘unified’ fabric? If they let their FC ‘partners’ to lead then the FCoE will cost as much as the FC connection does today.
One would think that at least Cisco could provide a complete single point management and provisioning of virtualized resources i.e. the VDC-OS essentials. Instead, they went to basic VMware and to another party for their system management software. In doing that they have created a lot of potential for finger-pointing with no single party system responsibility. This will be a big slow-down factor.
Perhaps HP or IBM are awake and fix all of that ahead of Cisco. Cisco will need to buy EMC or Net App and VMWare to get it right. This will be very expensive and will take time … by then it may be too late.
Kodak? Chuckle. Flash cubes? 110 format? Disc camera? Relabeled Chinon junk?
Kodak has been on the foot-dragging loser side of every battle since the Land camera? Kodamatic anyone?
Sorry, I see more baggage than bragging rights.
Wes, storage vendors prefer that you buy arrays for VM storage? Who’da thunkit?
John, somebody floated the idea that the PD buy is related to mobile tele-presence. I get seasick just thinking about it.
Richard, there’s certainly a case to be made that Cisco has banded together with a group worried about being marginalized as the shift to cluster architectures accelerates. We who are about to die salute you.
RC, Kodak was the long-time leader in digital photography, building the earliest solid state sensors in the ’70s, the first “pro” digital camera in the mid-80s, the first consumer digital camera – the Apple QuickTake 100 – in the mid-90s, and a leadership position in the US consumer digital camera market today and a strong position worldwide. With the right focus – which Cisco would be unlikely to provide, granted – they could take on Canon and make a strong move in the prosumer and pro markets.
What has Pure Digital done?
Robin
Pure Digital is more like Linksys products. Where does the public see the Flip Video Camera? On top of a pallet display at Costco.
The Flip looks like something Apple would make. For anyone serious about cameras, it is about as attractive as a Fisher-Price PixelVision. But if the future is about putting stupid videos on YouTube, what equivalent product does Kodak offer?
I’ve seen cameras that tout YouTube ready video, but I’ve had good results with a series of Nikon Coolpix. Thanks for the blog. Interesting stuff.
RC,
Kodak Zi6. Removable batteries – AA – SD card, 720 HD, tripod mount, USB flip out connector, digital zoom that works – most don’t – in Pink and Black, $150 online. Shoots in native QuickTime so it goes great with a Mac as well as a PC.
My DTrace video was shot with it. Watch it on YouTube HD and you’ll get a pretty good idea of the quality. Even the built in mic isn’t too bad. I got it because I got tired of schlepping my HV20, 2 mics, Gitzo tripod, wide-angle lens, tapes, cables and such around.
I can import into the MacBook Air, edit in Final Cut Express, and upload to YouTube. The MBA could use more RAM and MHz for HD, but for short pieces – all I ever do – it works.
Robin
Robin – I did a quick post to clarify the VMFS-3 limits (it is NOT limited to 32 VMs in ANY WAY SHAPE OR FORM) here:
http://virtualgeek.typepad.com/virtual_geek/2009/03/vmfs-best-practices-and-counter-fud.html
@Wes – speaking as a storage vendor – RDMs do indeed have a place, but they are relatively rare (W2K3/W2K8 clusters, Symm Gatekeeper LUNs if you virtualize Solutions Enabler, extremely large database log/db storage objects – not because RDM is faster, but because you can’t accidentally oversubscribe it). The primary storage for customers is VMFS, NFS or (for the smartest ones) a bit of both. Array snapshot use cases generally don’t need RDMs anymore.
@Robin – BTW – I certainly DO love my Kodak Zi6. It’s how I do most of the videos on my blog, along with Camtaisa. I find the 720p mode sucks it in indoors conditions, but the VGA is great, and the 720p is great outdoors in bright light.
Chad, appreciate the update and more info on VMFS. Just for the record though I didn’t say VMFS was limited to 32 VMs – just 32 servers. Which, IMHO, doesn’t make for a very large cloud – even for high-powered servers.
Robin
True Robin – so then bear with me a second longer – that’s also not a VMFS restriction. The thing that governs VMware cluster size is VM HA (which is linked to autostart or AAM). This is limit is being increased to 64 in the vSphere release. More importantly, in the cloud models, you don’t manage VMFS or even ESX clusters – there’s a higher level abstraction created by the vCloud API model, which is what you manage (which in turn manages many clusters).
All cloud models require some type of containerization – the advantage is that if the containerization model is a standard (OVF for example), you could take your container, run it in many clouds, including the one in your own datacenter.
Chad, thanks for mentioning NFS; I had forgotten about that.
As for the 64-server limit in ESX clusters, I don’t see the benefit of two-level clusters of clusters; this sounds like unneeded complexity.