Data Center Strategies

Scale-Up Intel Architectures

2009-11-05T06:50:05-08:00

Announcements this week from ScaleMP and 3leaf Systems offered new ways to build very large Intel-based systems from commodity building blocks. Both approaches are similar in concept, in that they create a single system from multiple x86 servers by tying them together on an InfiniBand fabric. Once the system is built, it behaves as a single system so in you can build 8-way, 16-way, or even 32-way x86 servers with terabytes of memory, running a single operating system image.

Where the two companies differ is in their implementation and go-to-market strategies. ScaleMP’s approach uses InfiniBand cards in PCIe slots, so all the inter-node traffic (i.e. memory reads and writes) has to be moved from local memory and sent over the PCIe bus. The big advantage of this approach is that it will work with any commodity server, and ScaleMP already has some level of partnership with all the major server OEMs as well as a number of channel partners (mostly in the high-performance compute (HPC) space).

Meanwhile 3Leaf Systems has built a hardware solution that puts an InfiniBand interface directly into a CPU socket on custom made 3-socket motherboards that use AMD processors. Technically this approach has advantages because the latency for inter-node traffic will be reduced by eliminating the PCIe bus from the equation. The downside is that can’t be used with the off-the-shelf systems, so 3Leaf will have to find hardware partners willing to build suitable servers.

The interesting question is whether these systems have a market outside HPC that can attract the server OEMs. Candidates for possible new markets include:

Large database server market: Currently dominated by RISC/UNIX platforms. The problem here is that a large server constructed from multiple commodity boxes will not have the availability characteristics of the large RISCV/UNIX platforms, so for now at least, this is a non-starter.
Hosting for large numbers of virtual machines: Today, x86 server virtualization products are typically run on 2-socket and 4-socket commodity servers. This places an upper limit on the number of VMs hosted on each server, resulting in the need for multiple physical servers in clusters. A cost-effective way to build 8-, 16-, and 32-socket servers would allow for much higher consolidation ratios. However, many customers would be wary of hosting hundreds of virtual machines this way because of the complexity involved in the event of hardware problems.
Flexible public/private cloud infrastructure: A cloud built from 2-socket servers can’t host a workload that needs more than a 2-socket server can deliver. You can get around this by having servers of different sizes in the cloud (i.e. a mix of 2- and 4-socket servers) but this mean predicting what the customers will ask for and will increase hardware costs. But a cloud that can reconfigure the 2-socket servers into larger systems has much more flexibility. For example a cloud consisting of sixteen 2-socket servers can be presented to customers as four socket servers, two –8-socket servers, two 4-socket and four 2-socket server and so on.

Of these markets, I think the last one is the most interesting, and it will be interesting to see if the idea catches on with hardware infrastructure as a service (HIaaS) providers.

Posted by: Nik Simpson

Clustered Commodity Storage (CCS) hows that sound?

2009-11-04T13:32:50-08:00

Cloud storage discussions are instigating a rethinking of storage infrastructures inside a datacenter. The thought pattern goes like this: As a datacenter owner, I’d like to use an external cloud storage supplier to drive down my storage costs and simplify my life. But I just don’t trust the external vendor to handle my data for any number of reasons. hmmm, maybe I can implement my own cloud storage environment, keep it inside my datacenter and avoid all the issues associated with using an external supplier. After all, why can’t my datacenter implement its storage environment by borrowing from the cloud storage vendor’s ideas?

Thus enters the idea of a storage infrastructure built around low cost commodity storage hardware, a distributed file system, simple management, multi-access points, multi-protocol, failure resilient, virtualized, extreme scalability, do-it-yourself storage.

Now, that’s a mouth full. How can I give an elevator pitch with that? It better be a tall building. The industry needs a name. A name I don’t have to take several breaths while saying.

Clustered Commodity Storage. Now that’s simple. Even simpler is: CCS.

I’d suggest CCStor but Symantec’s CommandCentral already uses it. And unfortunately, Scale Computing uses “Commodity based Clustered Storage” or CCS in some of their literature. But they are just getting started as a company. Perhaps a slight modification to their (one and only?) white paper could be made; consider that using the word “based” is redundant if one switches the word order. What do you say Scale Computing?

Here’s a simple diagram for what I mean by CCS:

Clustered Commodity Storage. Easy to say and brings home the concept.

Let’s walk through the definition.

Clustered: Cluster implies many similar things joined together for a purpose. Many nodes of storage are necessary to build out a storage system; certainly “cluster” applies well. The dictionary defines cluster as: “A group of the same or similar elements gathered or occurring closely together; a bunch”. Works for me. Btw, “Grid” is also a possibility but we’ve already been there.

Commodity: Commodity implies cheap and easily attainable hardware. Individual storage nodes are made from commodity hardware. Nodes can come from more than one vendor but may need to be homogenous in a single system. The storage hardware does not have the utmost in performance or reliability but is “good enough”. This makes the hardware cheap.

Storage: The necessary noun. Strip the adjectives away and generically we have storage. Not compute, not network. The collective storage, not the individual pieces making up the system. The sum of the parts, the entire system that contains data. The logical storage device that a compute device sees.

Clustered Commodity Storage. CCS.

I think that works.

Yes there are a few other definitions for CCS. But so be it. Context is everything.

Can the industry rally around the term? Entrants in this nascent market include Parascale, IBM (soFileSystem), Symantec (FIleMover), Gluster, IBrix, Scale Computing. Any others? I’ve certainly missed a few. What do you folks think? Keep your own branding but classify your products as CCS.

Can we have a harmonic convergence on terminology? Any trademark or copyright problems?

Let me know.

Posted by Gene Ruth

Burton Group Institute - Data Center Workshops

2009-11-04T09:25:49-08:00

Last month Burton Group launched, Burton Group Institute, a new on-demand training resource for enterprise technologists. The collection includes data center operations workshops that feature real world experiences from Burton Group analysts and consultants for enterprise technologists to utilize technologies challenging IT. Topics span several technologies, but the short list of data center training courses includes:

The training library is a collection of Burton Group workshop styled presentations focusing on a range of technologies and implementation strategies that include: identity management, security and risk management, networks and telecom, enterprise social networking, and content and collaboration. As an extra benefit iaap (International Association of Privacy Professionals) credits are available for completion of select Burton Group Institute workshops.

The workshops are delivered in a video format with integrated slides and audio and include presenter/audience interaction for further details on the topic. An example of the workshop can be viewed featuring Chris Wolf and Richard Jones presenting “Advanced Server Virtualization Workshop.”

As a reader of the Data Center Blog, Burton Group would like to offer a 20% discount with the coupon code DCB38.

Which Reminds Me

2009-11-03T06:22:09-08:00

I was just reading an article about IT’s desire to better understand costs (see here). The article reminded me that back in July, I posted (here) an invitation to a survey that Burton Group was conducting as part of my ongoing research on data center cost models. At the time I promised to share some of the results on the blog. Before I go into the details of the survey I’d like to link to some additional information that may be helpful.

If you are Burton Group customer, you can download “Counting the Cost of the Elephant in the Data Center” which contains a more detailed look at cost modeling and also includes a sample spreadsheet that can be used a starting point for looking at infrastructure costs (i.e. what does it cost each year to run a server).
There is a recording of my 1/2 day workshop on data center economics from Catalyst 2009 which is available to anybody with a browser. In the interests of full disclosure, this is $199 to purchase.

And while I’ve got your attention, if anybody is interested in talking to me about data center cost models and how they are used in your business, please get in touch via email to nsimpson@burtongroup.com.

Anyway, on the results of the survey…

The 80+ respondents to the study were well distributed across industry segments and company size:

Participants in all major IT sectors including; education, government, pharmaceuticals, aerospace, energy and consulting.

Companies ranging in size from less than $50 million to greater than $10 billion in annual revenues.

Out of all the companies surveyed, less than 25% had a comprehensive model for their data center costs. A similar number had no cost model of any kind. The rest of the companies surveyed had some form of cost model that was limited in scope and application.

Of the companies that did not maintain a cost model or only maintained a limited model, the key problems were either lack of motivation (i.e. costs are not a major part of the decision making process) or difficulty in collecting the necessary information because of it’s spread across multiple fiefdoms within the organization (i.e. facilities, purchasing, human resources, etc.)

As to the contents of the data center cost model, at least 25% of those using a cost model did not include any information about power, cooling, or facilities costs. This represents a serious shortcoming as these costs are a significant part of the overall IT costs.

Another point that emerged is that the old idea that a data center should be expected to last twenty years is no longer valid. Of the survey respondents, more than 85% had built or upgraded a data center since the year 2000, with 45% having done the work in the last year to 18 months. Yet 65% of the respondents expect to outgrow their current facility within five years. The major factors dictating data center life were limits on floor space and power distribution, with cooling coming in a distant third.

The biggest gap in terms of data required for a data center cost model was in the area of power. Only about 25% of respondents could track power consumption down to the individual piece of IT equipment, rest had varying levels of insight into their power usage. Interestingly this 25% figure was mirrored almost perfectly by the number of respondents who had calculated PUE for their data center.

Another weakness common to the majority of respondents was the availability of detailed purchase price information for equipment in the data center. Roughly 20% of respondents had the purchase price information in the corresponding Configuration and Management Database (CMDB) entry for the equipment. Another 23% had a spreadsheet that has to be maintained by somebody in the IT organization. But over 50% of respondents would have to track down the information from other sources in the company such as a purchasing department.

Overall, the survey indicated that many IT organizations are not in a position to fully understand the costs of running their data center. The lack of a complete understanding of data center costs creates a potentially dangerous hole in the IT decision making process, especially because IT decisions are increasingly cost driven.

Posted by: Nik Simpson

Red Hat and Microsoft: The Odd Couple?

2009-10-14T17:30:45-07:00

Red Hat and Microsoft announced last week that they have completed the cross-certification of each other’s server and hypervisor products. I have been blogging about the need this vendor interoperability for more than a year now and am glad to see that this has finally been completed – another quest that I can now mark completed!

There are some interesting aspects behind this announcement with the biggest being that Red Hat submitted the KVM hypervisor technology included in RHEL 5.4 as the third party hypervisor to be certified in the Microsoft server virtualization validation program (SVVP). RHEL Xen is not included in any of the certifications. As a result, this certification doesn’t help existing RHEL Xen deployments at all. I interpret this move two ways: First that Red Hat is telling its virtualization customers to move to KVM based virtualization from RHEL Xen, and second that Red Hat is done with Xen – they do not want to invest any more than the bare minimum to meet their RHEL 5 product lifecycle support obligations.

I had blogged last year indicating that Red Hat needed to come out and tell the world clearly where it is going with virtualization. Red Hat has not publically talked about Xen since it acquired Qumranet last year. They have not clearly articulated their plans for Xen, but when asked, they indicate that it will be supported throughout the RHEL 5 product lifecycle and then change the subject to the benefits and greatness of KVM. RHEL 5 first shipped March 14, 2007 meaning that it will end of life near the same day in 2014.

But Red Hat’s actions tell the story loud and clear: Red Hat virtualization customers: 2010 should be your year to move from Xen to KVM if you desire to stick with Red Hat virtualization for the long term. This story will become even more clear once Red Hat releases the Red Hat Enterprise Virtualization (RHEV) platform sometime later this year or early next year. RHEV consists of a set of virtual infrastructure management tools and cloud enabling components all built around the KVM hypervisor core. It will also include a stand-alone hypervisor that is effectively a stripped down RHEL 5.4 distribution. RHEV hypervisor is planned to have RHEL 5.4 kernel parameters tuned, such as thread scheduling and dispatching, for KVM based virtual machines as opposed to natively hosted applications.

The interesting part of this certification for Microsoft is that they now have two Linux distributions certified on Hyper-V: Novell SLES and Red Hat RHEL. The Red Hat distribution versions certified on Hyper-V include RHEL 5.2 through 5.4. Preparing for this work, Microsoft submitted paravirtualized Linux guest device drivers for Hyper-V to the mainline Linux kernel, and those were accepted this past summer – another huge step forward in collaboration between Linux and Microsoft.

But Microsoft still has some way to go with Hyper-V and Linux. Even with the latest Hyper-V R2 release coming soon, Hyper-V does not support virtual SMP for Linux guest operating systems. This limits Hyper-V’s ability to penetrate into heterogeneous environments.

[Posted by: Richard Jones]

Oracles ExaData 2 Results and $10M Challenge

2009-10-13T19:42:16-07:00

Last month I covered two ads that Oracle had published about their commitment to Sun’s hardware line and its performance vs. IBM as a database server (see here.) The second ad showed an IBM server at 6 million transaction/sec side-by-side with an unannounced SUN configuration at “XX” million transactions/second. I’m a literal kind of guy, so I assumed that “XX” implied a double-digit result greater than 10 million transactions/second. So I was disappointed with the actual Sun result released on Monday (see here). The SPARC-based ExaData 2 cluster achieved 7.7 million transactions/second using the latest hardware, scale-out clustering, and a boat load of FLASH memory instead of high-performance disk. This clustered result represents ~25% improvement over IBM’s single system result from 18 months ago which didn’t use FLASH at all.

The 25% improvement brings me to Oracle’s performance challenge, “offering a 10 million dollar reward if your application isn’t twice as fast on Sun hardware”. If the best that Sun/Oracle’s engineers could manage was a 25% improvement, it sounds like it should be pretty easy to win. But before you start spending the money, read the conditions of the benchmark. First, if you aren’t using IBM storage for your database, you can’t even play (see rule 2-B). Second, you have to prove that the application performance is limited by the database, rather than other way around, which may be difficult. If that doesn’t turn you off, how about section 5-D which states “Sponsor, in its reasonable discretion, reserves the right to limit the participation of any entity or person in the Challenge, amend or interpret these Official Rules”.

Posted by: Nik Simpson

A degraded world awaits you

2009-10-12T13:28:59-07:00

Anyone have a petabyte scale storage environment? Maybe hundreds of terabytes? Raise your hands. For those with hands raised (you can put them down now), no doubt your life is occupied replacing failed disks. Keeping such a large storage system running can be like me plugging leaks in my roof during a Texas down pour (everything is bigger in Texas after all!). Hardware failures, like my roof leaks during a Texas down pour, are inevitable, continual and just part of life. The bigger the system is, the more likely that at any moment a disk rebuild is underway and someone is madly hunting for the failed disk to replace it, like me hunting for the hole in my roof. And woe to the datacenter that lets its temperature rise – disk failures accelerate as ambient temperature goes up.

Let’s look at an example of a Texas gully washer size system: In 2005, IBM built a supercomputer system for the government (Lawrence Livermore National Laboratory) called Purple/C. Here’s some specs:

1536 node 100 teraflop
2 PB GPFS (general parallel file system)
500 RAID controllers, 11000 disk drives, 4 disks per RAID 5 set
126GB/s parallel IO to a single file

Now that’s a big system. Lots of disks. IBM tells me that at any given time, up to 20% of the disks are in a rebuild mode. Since these are SATA disks, that’s not surprising. A rebuild can happen not just due to an outright disk failure but also from a single block corruption. So calculating just on the basis of a disk’s MBTF is misleading. One must also account for data reliability for each disk. The more active a disk is the more likely a bad sector will be encountered and thus require a RAID rebuild. Even with a background disk scrubbing routine.

That’s all well and good (well, not so good), and certainly there are other large supercomputer-like systems dealing with the same laws of physics, but the lesson, for those not in the supercomputer business, is this:

Large enterprises moving to a cloud storage environment based on distributed commodity storage nodes should expect the same. That is, regular hard failures throughout the system. For some admins this may seem alarming. Disk failures are never a good thing. Grouping 1000s of disks (SATA or otherwise) into a system, however, ensures the presence of failed disks – there is no avoiding the laws of probability.

To make large implementations feasible requires a resilient distributed file and management system capable of living with constant disk failures. Some startups such as Parascale and Gluster are in that space as are the likes of Symantec with their just announced FileMover product, IBM’s SoFS system based on GPFS, IBRIX, and Isilon, etc…

So if you plan a cloud storage deployment within your IT, get use to and plan for disk failures. This means staging boxes of disk drives by the system, a rational way to identify failed disks, and then someone to replace the disk.

To simplify maintenance, maybe a robotic tape library can be repurposed to plug and unplug failed disks (tongue-in-cheek). Unfortunately, no such option exists for my roof.

Posted by Gene Ruth

Storage is too complex

2009-10-02T07:29:42-07:00

Advanced storage features that seem whimsically inviting in individual storage systems may prove to be sources of great complexity in a large and a diverse storage infrastructure. Little involvement with business process stakeholders, IT architects and administrators, enables the purchase of “coolest” storage arrays that collectively grow into an unruly management mess. The lack of proper oversight and adult supervision can result in an inefficient and dysfunctional storage environment so twisted that no team of rocket scientists can unravel it without pyrotechnics.(regards to Robyn Hitchcock – Uncorrected Personality Traits).

hey, cheer up!

In a past blog I touched on the importance of managing information across an enterprise, identifying its value and mapping it accordingly onto storage hardware. Ok, I’ll admit that is not easy, but anything is better than nothing. Even if all storage hardware performed the same with no differentiation, one would still need to ensure that information is properly secured, privatized, discoverable and discarded. None of those is easy to figure out, so let’s turn to another part of the problem:

Storage tier definition

I get asked a lot about the definition of a storage tier. The theory being that once armed with a clear storage tier definition, an organization can place their orders with for tier 1 through tier x. Oh, if only it was so easy, it is not. Storage tiers are in the eye of the beholder. Each organization must set its tiers of service and group storage hardware appropriately. Tier “one” storage to a health insurance records processor is tier “unusable” for a financial institution processing stock transactions.

complexity makes life interesting and ensures full employment.

One truism in the storage industry continues to be, well, ah, true: there is no shortage of storage product to pick from when defining a storage tier. Vendors and technology abound, innovation lives. But in that innovation lies the seeds of confusion and poor choices. Like the Cambrian explosion, many possible forms of storage products simultaneously coexist, but someday their diversity will diminish as market forces willow down the weak contenders.

I need to be careful here lest I rile anyone’s feathers but to test the “complexity” theory open the web page of your favorite top tier storage vendor. I mean companies like IBM, EMC, NetApp, HP, HDS and the like. Count the storage product offerings. On one site I counted 22 major product families and countless model numbers. Talk about covering your bases! Now of course all these models exist because someone is buying them – a testament to good selling techniques. But really, can’t things be simplified? Does the perversity of products help end customers of just allow vendors to inextricably infuse themselves into IT infrastructure?

Not to let the less-than-giant vendors off the hook, they too attempt to fill every nook and cranny hoping to either make it big (a few do) or get bought by bigger companies (win the lottery) or just fade away to oblivion. [quick side note: I did a briefing with KOM Networks recently which makes archive equipment. They’ve been in business for 40 years! amazing.] Again these “ less-than-giant vendors” slice and dice the storage market to the slimmest of differentiation.

Narrow your choices

Look at your data and service-level requirements to begin narrowing down the useful storage technologies. Is a unified Ethernet in the cards or do you need to carry along a legacy infrastructure? For most, considerations for an existing infrastructure must be made. Think of the future state of where a storage infrastructure needs to be. Plan out the obsolescence of existing equipment and its replacement candidates. If high transactions rates are called for, then a tricked out SSD product is in your future. But don’t put any service on high performance storage without a justification. Unless proven otherwise, services belong on “good enough” storage. Maybe that’s NAS, maybe that's iSCSI or maybe your server virtualization needs will drive the decision.

In any case, keep the “storage tier” count small. Drive down complexity by pushing a “one size fits all” approach. Avoid mixing vendors in a common tier – this can only lead to heart ache. Now that’s not likely to fly but at least the lunatic fringe case will surface and can be dealt with. As many vendors offer multiple and significant performance levels in a common frame consider these products to span as many service levels as possible

to simplify your life.

Backup and archive will loom large in any well thought out environment. Apply deduplication and recognize that replication across sites or systems (for disaster recovery) will often lock down a common vendor choice between systems to make the replication work efficiently.

Now you may be thinking, I know this, tell me something new. Well, that’s the trouble we know but do not act. Set a plan in motion to minimize the storage product count in your storage environment. And, don’t ever allow a application to be released into the IT wilds without an information and storage mapping plan. Taking this approach should make a CIO smile, knowing that tangible efforts are being taken to drive down costs and align IT infrastructure with business processes.

Posted by Gene Ruth

Its painful to watch Suns troubles

2009-09-30T14:03:05-07:00

The senior vice president of Sun’s North America Region Global Sales and Services organization, Randy Seidl moves over to a similar position in today’s announcement from HP. Good for Randy but its so sad to see Sun get dismantled piece by piece. Not only are Sun's competitors stealing away customers, but they are also taking key executives...especially in sales. Losing top executives does nothing but erode customer confidence. The Exadata V2 blew some life back into Sun’s corpus morte but Sun's customers need for Oracle to do more - never mind that the TPC counsel slapped oracle down for misleading claims. How will this end, will there be any customers left by the time the European Union works out their objections to the merger? I sure would hate to see a company with such a storied past disappear because of a botched merger.

Posted by Gene Ruth

Wheres The Beef?

2009-09-16T05:55:45-07:00

Yesterday, Oracle announced the first Oracle/Sun post-merger announcement of a an integrated hardware/software product, known as Exadata Version 2 (see here). The previous version of Exadata was built on HP hardware, but it’s no surprise that Oracle has turned to Sun for this, though the timing of the announcement makes it hard to tell whether the decision to use Sun hardware came before or after the merger announcement in April.

On the hardware side it’s not particularly exciting, standard Intel Xeon 55xx-based Sun servers (not SPARC in case you were wondering) combined with Sun’s FlashFire SSD storage subsystem. Interesting, but not earth shattering in it’s ingenuity or innovation. I expect that you’ll see FLASH memory heavily used in future database configurations from many vendors. Dell, HP, and IBM are working with Fusion-IO and Dell has already published record breaking TPC-H benchmark numbers using Fusion-IO technology (see here) so it’s only a matter of time before this becomes common practice.

Unfortunately, the information available from Oracle at this point has no performance data to back this claim, maybe the “teased” October 14th announcement (see here) will fill in the blanks. Personally, I think that people will need to see more innovation from Oracle/Sun if the hardware part of the business is to survive.

Posted by: Nik Simpson

Oracle - Were Serious About Hardware

2009-09-10T12:24:06-07:00

Based on recent announcements from Oracle, they are at least trying to dispel doubts over their intentions for the Sun hardware business, and poking a finger in IBM’s eye while they are it. This is a welcome sign from Oracle, especially with the additional doubts created by the EU decision to review the merger (see here), a solid statement of intent was increasingly critical to prevent continued hemorrhaging of Sun’s hardware business. The first ad (see below) can be found here,

On the face of it, this is a pretty clear statement of intent and I hope they mean it. But, I’m left wondering how they intend to make good on these promises and still make decent money on SPARC hardware.

The second ad is more direct competitive statement related to IBM’s DB2 business (see here), something that used to be a Sun stronghold.

Today, the king-of-the-hill for a single system database benchmark (TPC/C) is IBM hitting 6,085,166 tpmC, but Oracle seems to be implying that their result will be greater than 10,000,000 tpmC. The question is, will this be an apples to apples comparison of two single boxes or will this be a cluster to single box comparison? My guess is that Sun doesn’t have a single system capable of a result of this magnitude which leaves them with using a clustered version of Oracle for the benchmark.Whatever the case, it should be interesting to see how they do and what hardware they use. Anyway, I’m looking forward to October 14th to see just what they have to say.

Posted by: Nik Simpson

When You Think Its All Over, It Has Only Begun

2009-09-04T12:51:49-07:00

Over the last few weeks, you could be forgiven for thinking that the Sun/Oracle merger was a done deal. Recent decisions by Sun’s shareholders to approve the deal, and the US government’s decision not stand in the way seemed to signal the end of the saga. But earlier this week the European Union dropped a bombshell by deciding to investigate the deal, particularly with respect to the acquisition of MySQL. It’s unlikely now that the deal will go through before January 2010 at best, and that’s not good for anyone. There are three possible outcomes, onoe of which is particularly attractive if you are Sun or Oracle (or a Sun customer):

The EU eventually approves the deal, meanwhile Sun has been left twisting in the wind for the best part of a year, as it’s competitors steal one customer after another. IN just the past two quarters, Sun’s revenues have dropped 30% and 37% respectively which is little short of disastrous.
The EU rejects the merger (and they’ve done that before, just ask GE). A rejection would leave Sun in an untenable position, it’s been hemorrhaging market share over the last few months, and it’s value to possible suitors such as HP would be dramatically reduced by a failed Oracle merger.
The EU places conditions on the merger such as preventing Oracle from acquiring key software assets from Sun such as MySQL. Since it was the software that Oracle was after in the first place, this might make the whole deal unattractive to Oracle.

At best, this puts an ever bigger question mark over Sun’s future which can’t be good for their customers. So, whatever the final merits of the deal, I’m inclined to agree with Macbeth, “If it were done when 'tis done, then 'twere well It were done quickly.”

Posted by: Nik Simpson

Thoughts on the VMworld Day 2 Keynote

2009-09-03T06:57:10-07:00

I was very impressed by the information disseminated in the second VMworld keynote, led by CTO Steve Herrod. Here’s a summary of the thoughts I tweeted during the morning keynote (in chronological order).

Steve Herrod talked about a "people centric" approach. VMware's technology needs to understand desktop user behavior. The existing offline VDI model (requiring a manual “check-out”) is not people centric.
VMware’s announcement to OEM RTO Software’s Virtual Profiles was a good move. Burton Group considers profile virtualization a required element of enterprise desktop virtualization architecture.
VMware’s Steve Herrod and Mike Coleman discussed VMware’s software-based PC-over-IP (PCoIP) protocol. Feedback from Burton Group clients who were early PCoIP beta testers indicates that the protocol’s development is progressing well.
Herrod showed a picture of "hosted virtualization" for employee owned PCs on a MacBook. Is that a hint of a forthcoming announcement?
I would like to know if VMware’s Type I CVP client hypervisor will have VMsafe-like support in the 1.0 release. VMware has made few public statements regarding CVP architecture.
VMware’s CVP demo looked good, but it didn’t reach the “wow factor” achieved by Citrix when Citrix demoed a type I client hypervisor on a Mac at their Synergy conference.
The Wyse PocketCloud demonstration was impressive. PocketCloud is VMware’s first answer to the Citrix Receiver for iPhone.
VMware demonstrated the execution of a Google Android application on a Windows Mobile-based smart phone. Many opportunities exist for VMware and Google to collaborate in the user service and application delivery space.
Burton Group client experience backs VMware’s claims that vSphere 4.0 is a suitable platform for tier 1 applications. We recommend that x86 virtualization be the default platform for all newly deployed x86 applications, unless an application owner can justify why physical hardware is required (e.g., for a proprietary interface that is unsupported by virtualization).
To support tier 1 application dynamic load balancing, storage and network I/O must be included in the DRS VM placement calculations. It’s good to see that VMware is heading in that direction. DRS will also need to evaluate non-performance metrics such as vShield Zone membership as part of the VM placement metric (no word on this yet).
I would like to hear more from folks who have tested AppSpeed. Burton Group clients I have spoken with to date have not been impressed.
The DMTF needs to start doing more to evangelize the role of OVF as it pertains to cloud computing and service manifests.
I like vSphere’s VMsafe security API, but I want to see tighter integration with external management (exposed via the SDK), and better integration with VMware’s DRS and DPM services.
VMware talked about Lab Manager as a tool to promote user self-service for server VMs and applications, but I haven’t heard mention of a similar interface for desktop applications (like Citrix Dazzle). A user application service catalog is a missing part of VMware’s current virtual desktop architecture, and will need to be addressed by either VMware or a third party.
The data center on the show floor running 37,248 VMs on 776 physical servers would be more impressive if VMware disclosed the applications running on the VMs, along with the application workloads. Otherwise, the demonstration is really just a density science project.
I liked VMware’s coverage of virtual data centers. They are also defined in Burton Group’s internal cloud hardware infrastructure as a service (HIaaS) reference architecture.
Herrod mentioned forthcoming network L3 improvements that will make it easier to separate location and identity. This is something to follow.
Both Cisco and F5 are enablers for VMware’s long distance VMotion and are vendors to follow as this technology further matures.
VMware’s cloud layered architecture is very similar to the architecture defined in the Drue Reeves’ report “Cloud Computing: Transforming IT.”
Herrod did a great job articulating the importance of SpringSource to the VMware software solution. VMware needs an application platform to have a chance at holding off Microsoft long term, and SpringSource gives them that.

That’s it for my thoughts on day 2. As always, I’d love to hear your feedback. VMworld 2009 was a great conference. I enjoyed my time meeting with Burton Group clients as well as the several conversations that I had with many attendees. See you next year!

Citrix HDX 3D IHV support

2009-09-02T10:47:22-07:00

Citrix announced on Monday coming updates to their HDX remote desktop protocol including 3D rendering. Shortly thereafter, Brian Madded blogged on HDX 3D requirements, specifically that the back-end data center server that hosts the desktop VM requires an NVIDIA GPU that has CUDA capabilities. Fortunately, modern NVIDIA GPUs have the CUDA feature. Brian gives a good overview of HDX 3D in his blog is you need the background on what it is.

Dependency on NVIDIA GPU in the server. This point raised some questions: Servers don't do graphics, and to save $ - they don't include higher-end GPUs. Most include just a simple graphics chip on the motherboard. Since I'm at VMworld and had the opportunity to visit with Independent Hardware Vendors - I went to find out. Here's the answers I got from HP and Dell:

HP: HP will support you if you install an NVIDIA GPU adapter (PCIe) into a DL160 or DL165 rackmount 1U server only. While you can put an NVIDIA GPU in other HP rack mount server products, and it would most likely work, the company won't support you. That being said, however, HP will direct you to their workstation blade product line - a product specifically built for desktop virtualization purposes. Each blade in the workstation blade chassis includes an NVIDIA GPU pre-installed on a mezzanine board. But realize that for graphics intensive solutions, you are stuck to a 1:1 consolidation ratio of desktop to blade without virtualization as the GPU has yet to be virtualized by any hardware (I can suspect we will see this technology in the near future as the demand for desktop virtualization grows).

Dell: Dell is definitely behind the game here. Dell currently does not offer any rack or blade servers specifically designed for desktop virtualization - that include a GPU. The guys in the Dell booth at VMworld did tell me that they have a third party systems integrator that takes Dell blade products, adds the necessary GPU hardware and software and sells those to desktop virtualization customers, but the end product is supported by the partner and not Dell. The guys in the booth couldn't remember the name of the partner and 15 minutes of Googling didn't turn up any leads either.

Needless to say, the choices are indicative of a market in its infancy with HP in a clear lead. The next big step will be GPU virtualization.

[Posted by: Richard Jones]

VMworld Day 1 Keynote A Few Thoughts

2009-09-02T08:24:17-07:00

VMworld got off to a strong start yesterday, with the Paul Maritz keynote setting an early tone for the day’s events. I used twitter to add commentary to the keynote and plan to do the same today for Steve Herrod’s morning keynote. Below is a summary of my thoughts/tweets on the day 1 keynote.

VMware’s clear goal is to provide a complete platform for cloud-based computing. I’m hopeful that VMware will spend more time helping customers build internal clouds, which focuses on their immediate needs. Public clouds, while intriguing, is a future initiative for most Burton Group clients.
Paul Maritz and VMware talk about vSphere as an enabler for choice; however, “choice” is a subjective word. To VMware, choice means choice of cloud providers. To others, choice means choice of virtualization platforms.
vSphere is a foundation for internal cloud, but to keep it in perspective, it’s just the foundation at this point. VMware and its partners need to build the walls. Improvements such as tighter integration between DRS, DPM, orchestration, and security zoning (i.e., vShield Zones) need to be made. vSphere APIs need to further evolve to communicate logical and physical zoning restrictions from either vShield Zones or VMsafe-enabled appliances to management and automation tools.
I was glad to see Maritz bring up context switching and its impact on x86 virtualizaiton. Many virtualization consolidation planning tools miss analyzing applications with high context switching and don’t consider it when determining VM placement. This oversight in planning tools can cause post-deployment performance problems.
Martiz clearly demonstrated his command of the VMware product suite and partner integration avenues. He’s a very technical CEO and his knowledge resonates well with the VMware user base.
The IBM demonstration of vSphere integration with hardware-assisted virtualization active power management was impressive. Many of our clients are looking at exploiting the Enhanced Intel SpeedStep and Enhanced AMD PowerNow! hardware technologies to save on power and cooling costs.
Maritz highlighted the importance of chargeback and service catalogs in cloud computing, and I agree. I discussed the importance of both of these technologies in the blog post “Cloud and the Wal-Martification of IT.”
vCenter Chargeback is a good first step, but I think it falls short of giving IT the complete picture of available capacity. Awareness of factors such as security zoning restrictions is required before vCenter Chargeback will accurately forecast available capacity across an enterprise ESX infrastructure.
Paul Maritz discussed the concept of a virtual data center. Burton Group agrees with this, as we’ve had virtual data centers in our hardware infrastructure as a service (HIaaS) reference architecture for several months. You can learn more about it here.
VMware’s running tally of more than 1,000 service provider partners is impressive, but I would like to see increased guidance for customers on how to build out internal clouds.
VMware submitted the vCloud API to the DMTF – a first step toward making the vCloud API an industry standard. That being said, VMware’s competitors would need to adopt the API for it to be a true standard. A standard with one implementation isn’t a true standard, but rather is a proprietary technology, but with good marketing.
I liked the vCloud Express demo, but would also like to see a service interface on the front end. I like the option of provisioning a server, but would also like to see a demo of the vCloud Express showing an interface that makes sense for the application developer or business professional.
Rod Johnson did a tremendous job with the SpringSource demo. Giving application owners an interface to provision an app locally, or to an internal or external cloud was spot-on. IT service delivery requires IT operations to give application owners and individual business units interfaces that they understand. Hyperic is also a key part of an integrated VMware cloud stack and is a technology VMware shops should begin working with in their labs.

That’s it for day 1. I’ll post analysis of the Steve Herrod keynote next.