You can find out more about Bart on his popular blog Dirty Cache.
Kay: I’m joined today by Bart
Bart: Hi Kay.
Kay: Can you just start off just by introducing yourself, please? What your background is.
Bart: So, you know, I’m Bart
Bart: So and as a financial company, of course, Oracle was mission critical. So I did some work on security with Oracle. I did a lot of work on performance tuning. I did a lot of work on establishing disaster recovery, backup restore, upgrades, back fixing all that stuff. So obviously, you start to work with Oracle DBAs, Oracle developers. I got the opportunity to hire external consultants and I wasn’t just letting them go to do their work. I was just peeking over their shoulders to figure out what they were doing so I could learn. So that’s basically my Oracle history.
I joined EMC and actually the first day I worked for EMC I met an old colleague and he asked me “Hey, are you working for EMC now?” I said, “Yeah.” He
Bart: Well, that’s an interesting point. The main reason to start my blog is mainly because I am kind of like a — I’m obsessed with efficiency so I always want to get my customers the biggest bang for the buck. I want them to spend as few dollars, or euros, or pounds as possible and get the maximum return on investment. At the same time, you know, look at things like service levels, reliability, availability and all these things. So I think this is one of the reasons why I joined EMC because it’s kind of like the same
I’m obsessed with efficiency so I always want to get my customers the biggest bang for the buck. I want them to spend as few dollars, or euros, or pounds as possible and get the maximum return on investment.
So talking about Oracle, Oracle, not unlike EMC, is a very big company. I think Oracle is depending on how you count the profits it’s roughly twice in size. For example, if you look at
So in summary, because of my old days I like Oracle. It’s rock solid, it works, it’s reliable, there’s a whole ecosystem of tools, features, options and there is a lot of performance tuning tools and all that stuff. So as a foundation for
If you compare that to how EMC does things, we typically build in our hardware things like hardware redundancy, things like data integrity protection that are built in the chips instead of in software in your part of the system, so I think we have an advantage there. But that being said, it’s not always an easy discussion with customers, especially if you talk to Oracle DBAs. Oracle is very good in building kind of like an ecosystem with DBAs and developers, they are very loyal and you have to really convince Oracle-minded people that there is an alternative to what Oracle is shipping them. So, it’s probably some kind of, on the corporate level classic coopetition. We partner with Oracle as well as competing with them depending on which area, which side you look at.
For myself, it’s probably some kind of love-hate relationship. I am very into Oracle database, I like to work with it, I like to play with it. But if you look at the other options, like the Oracle virtualization and the Oracle ZFS file systems and all that stuff, I’m not such a big fan of that. Not so much because it has an Oracle logo, but mainly based on the experiences I hear from my customers and then you start investigating, figuring out what it’s all about. So it’s true, it’s not 100% in support of Oracle. I’m critical, and I don’t like all the products. That being said, I’m probably biased by the fact that we compete. That’s just the way it is.
Kay: So you’ve worked closely with Oracle in the time you’ve been with EMC. Oracle’s changed a lot since 2000 as has IT and Cloud is now gaining some traction. What’s your opinion on Oracle Cloud?
Bart: Well, to start with Cloud is something that is sometimes over-hyped, as well by vendors, even our own consultants and maybe even customers. What, I see a tendency in my customer base that they are a slightly turning away from the buzzword “Cloud” because they think it’s too much high level stuff without actually delivering some new things. EMC is fundamentally different in that area that we believe in Cloud. We have a whole strategy around it, but we may define our Cloud strategy a little bit differently than other vendors.
So in particular, looking at Oracle, a couple of years ago they ignored Cloud and they said, basically Cloud is nothing else than a different name for an old model which is called Application Service Provider, right? Now if you go hosting your applications with Oracle then you land on the Oracle Cloud with Oracle tools, with Oracle features and with Oracle pricing, etc, etc. fundamentally you have to make a choice. Am I going to run on Oracle Cloud, or am I going to run on something else? (whatever that may be). EMC’s philosophy is more about choice, we don’t like to force customers in one direction or the other. So if you look at EMC’s Cloud model you will see that we are trying to build a platform where customers can build their own Cloud and then extend it into a service provider and then move applications, ideally online, between on-premise and off-premise. At the same time everything is still private. You still control the operating systems, the applications, everything in the Cloud instead of handing it over to somebody else and then hopefully your service levels are guarded and well-maintained. Obviously, you only find out if there is some trouble and then you find out what the real protection of your data was. You cannot really test it upfront because you don’t have control over that field.
Now if you go hosting your applications with Oracle then you land on the Oracle Cloud with Oracle tools, Oracle features and with Oracle pricing, etc, etc. And fundamentally you have to make a choice. Am I going to run on Oracle Cloud, or am I going to run on something else, whatever that may be? And EMC, EMC’s philosophy is more all about choice
Also we support multiple platforms, many people know that VMware is partly owned by EMC so VMware is of course a big foundation in our strategy. But that being said, we support other hypervisors. We do some work with Microsoft Hyper-V, we even support Oracle OVM. We have our preferences and our pros and cons for all of these approaches, but we are not restricting customers to one certain direction. If a customer wants to deploy on Hyper-V that’s fine. If they want to go with OVM, well there may be some technical challenges, but we support it and we don’t restrict it. So we like to retain our customers because customers are happy with our solutions and service rather than locking them in and not giving them an opportunity to run on something else if they please. I think that’s one of the fundamental differences between the Oracle and the EMC approach.
Kay: So the customers you see move into the Cloud, does that present a challenge or an opportunity for your customers, or both?
Bart: I think both. What I see from my customer base at least, and that being said I typically talk to high-end customers, large banks, insurance companies, manufacturers and every now and then also a mid-size, maybe government or maybe smaller companies, so I see the customer’s views on this. What I typically see, the customers are very very conservative with their Oracle databases and I don’t blame them because mostly the Oracle database is actually the foundation for their mission critical apps. So if I’m an IT consultant, or CIO, or something like that, then I would kick the tires a bit with Cloud, for example with email, or with some online document storage and see how it behaves. But my mission critical apps? that’s probably the last thing I would ever put in the Cloud. If ever. This is what I see with customers and I think they are trying it out a little bit. Maybe do some testing there, but really going flat out on the Cloud with the mission critical apps, without having control over it anymore, I don’t see it happen that much. The EMC model is little bit more adaptive because even if you go into a Cloud conforming EMC architecture you still have control. You still manage the security levels, the high availability capabilities. So even though it’s not in your own data center anymore you still have control. That being said, Oracle is still not a big chunk of the pie in that area. Customers typically outsource or out-Cloud, if you will, other apps first.
Kay: So, Bart, in general what sort of challenges do you see for Oracle customers at the moment, and what do you do to help them?
Bart: Well, so there’s technical and non-technical challenges. Technical challenges is mostly around performance. Performance is still the whole thing and it’s actually part of my story when I talk to customers. I say well listen, this Oracle module is well over 40 years old and still valid. So how come we have a doubling of CPUs and cores and transactions every eighteen months and we still have performance issues? Well, probably because our workloads are growing, our data sets are growing and we need to run more different applications. So performance is still a big issue, that being said that’s also where EMC comes in because we have a lot of these new flash based systems, either in hybrid or all flash array formats. That helps customers remove their yoke on their neck and then they can actually use the faster CPUs instead of being restricted. So that’s one thing. I see customers having issues with upgrades. I, you know, to my big surprise every now and then I still see customers running Oracle 7, which is an exception but it happens. A little bit more Oracle 8 and a lot of customers still have Oracle 9, and Oracle 9 is long out of support. So my typical question is, why don’t you upgrade? And the typical answer, is it’s too risky, too complicated. Sometimes the interpretation, not so far with new Oracle versions, or it’s just too, too much work, too much trouble and they would rather stick with the old environments. Which is, by the way, something we try to help with. We cannot make Oracle upgrades easier, but we can facilitate easy data copies of entire application landscape that way you can actually do testing without influencing production. So we do have some side tools and methods to make it a little bit easier. So that’s a challenge.
On the non-technical side, I would say, I think the most important these days is customers being concerned about licencing, are we compliant? do we have enough licenses? do we have enough options? are we not using something we haven’t paid for? What happens if we get more of audits? What happens if we go virtual? Which DBAs typically don’t approve, but the cost savings are so obvious that they are trying to figure it out anyway. So even though I am an engineer and actually a techie I like to do hands on work, a lot of my discussions with customers is all around support, licensing, cost reduction, efficiency and all of that stuff. So that’s also a big challenge these days.
Kay: Bart, what do you think that Oracle could do better for customers?
Bart: So, well, again a technical and a non-technical side of this. The non-technical they should treat their customers a little bit better in terms of licensing and audits. I think they are building a very negative image these days which may lead to a little bit higher revenue or profit, in short term but in the long term they are losing confidence. So that’s something, if I was Oracle I would try to change that in different direction.
On the technical side, and also by the way customer support. They are very expensive support contracts but then if you really need support and what have you from customers it’s not always that responsive and high quality. So Oracle support model is probably a big cash cow and they try to avoid investing too much in it and that’s a big missed opportunity, I think. EMC is a little bit different in that respect. We try to help customers even though we know that an issue is not our issue. We still try to help customers and get it solved and we get very high ranks for that sometimes.
On the technical side, still the Oracle database is rock solid. I’m not so much an expert on applications so I cannot comment on them, but I of course I look with a special eye on everything that is industry latest. So what I see Oracle doing is gluing a couple of existing components together. For example, they have an operating system called Solaris and they have a File system called ZFS and they glue it together, they hook up a bunch of this and they call it a storage system, then they are starting to compete with EMC. On a technical level, I have no trouble at all competing with that. The problem is, of course, bundling it in with licenses and contracts and end-user systems so this is typically not so easy. That being said, if I was Oracle I would try to put little bit more effort in quality versus breaking the benchmarks with speeds and feeds, and doing a little bit more on the robustness and quality side. Marketing -wise, Oracle are story tellers, but most of the things they say I would verify, at least to make sure it’s true. Because it’s not always the reality they are promoting
Kay: Bart, what are your views on engineered systems?
Bart: Well to start with a comment that I am very biased, because if a customer buys mainly Exadata, well Exadata comes with internal storage. Right? and EMC has a lot of different products, but storage is still one of the foundations in the EMC infrastructure. So if a customer goes Exadata or engineered systems, then they are probably not buying that much of EMC’s technology. So, of course, we are competing with that a full hundred percent. I’m making no mistake about that. Everybody is free to understand that. So I always recommend everybody to check for themselves what I am saying, but, that being said, especially looking at Exadata, there is a lot of marketing hype but there is lot of marketing numbers about bandwidth and the IOPs and all that stuff, but what I miss is the quality of features on the backend.
So the front end is just the standard Oracle RAC cluster and, okay they have
Disaster recovery is a bit of an issue because they only feature they have is dataguard which is IP based and can only replicate the database. There is nothing there for applications or multi-database consistency. Virtualization is a big thing and I am a strong believer in virtualization to drive up the utilization of systems. Physically deployed systems have poor utilization, so I am a strong believer in driving it up from an average of 10 to 20% all the way up to well hopefully 50 or more. Because that saves big time on licensing. Until the beginning of this year Oracle did not support it on engineered systems, or at least not on Exadata and I think the pressure from customers has now become so high that with the introduction of X5 they finally started to support Oracle VM on Exadata. I still have my doubts whether this is a viable solution for critical systems, but at least you can see the sign on the wall that they are getting hammered by customers because they are asking for support and threatening to go VMware if they don’t.
But I still think that’s kind of a limitation. Of course, the obvious way from Oracle to go is to do pluggable database and I have issues with both on the cost side. I am fanatical about cost saving. So why would you spend the license for pluggable database, which is almost as expensive as Oracle, just to save a few bucks by consolidating databases? I think that the virtual machine approach is much more efficient. Also, if you’re running pluggable database, you cannot do a live migration from a database from one container to the other. You have to stop the database, unplug it, plug it into something else. Well, Mr. Oracle we are running 24 by 7 these days. You cannot stop a database and unplug it. You have to do everything online. So it’s expensive and it’s limited in functionality. Still that’s Oracle’s consolidation strategy, next to the OVM thing which is also a little bit limited in features and robustness and stuff. I have tested it myself and I was shocked by, not so much by VM itself, but how the way you have to manage it. It’s all, I don’t like it as much. But again, maybe I am biased and maybe customers are, some customers are running it happily without any issues. I have my doubts but, who knows? So the other engineered systems, is Sparc for example, interesting platform. That being said, I do not see that many customers that implement new green fields and environments in Sparc. So I think Sparc is replacing business for all service and that includes super cluster. Also if you compare the benchmarks of Sparc CPUs against the latest generation Intel, it’s not even close. The fastest Sparc CPU per core, and this by the way, is what you pay licenses on typically, the fastest Sparc CPU per core is only half the speed as the fastest Intel. So why would I deploy on Sparc unless I have to do this because of legacy issues? So I see Sparc dying a long and slow, painful death, but it will probably take much longer than I expect. It may be another 10 years, but…
So I see Sparc dying a long and slow, painful death, but it will probably take much longer than I expect. It may be another 10 years
Bart: But those days are over. I don’t see it happening that they ever regain the market. So you’ve had ZFS appliances and they also have new storage boxes based on Pillar, and they say it’s an all flash array but it can also do spinning discs so they are a little bit conflicting in their own messaging. If you use all flash array then you should not support spinning disc, if you do then it’s a hybrid. So they are trying to offer the best of both worlds, but you really can’t. So again it’s based on the old pillar and customers had some reliability issues with this thing so I don’t see many customers buying this. Oracle has some work to do, I think, in their portfolio. Again, database rock solid. Middleware probably okay; applications I cannot tell; but infrastructure, I don’t think–the only reason they typically sell it is because they sell it along with the applications in the database not because they have best of breed…
Kay: Bart, you mentioned virtualization. The latest version of VMware seems better at helping customers who are challenged by compliance. Can you talk about that?
Bart: I’m not a VMware deep expert but of course I check whatever they do. So a couple of comments on the latest VMware. I know that there are a couple of improvements on doing things like tracking virtual machine movements so the whole thing is a little bit better. This was one of the concerns that customers running Oracle and VMware would have. Hey, I’m running a cluster of, say, of 20 physical servers with a bunch of virtual machines and if I’m running more than just Oracle, then Oracle was claiming you have to license the whole cluster. Because if Oracle is installed all over the place and/or running and VM can feed motion to every CPU so you have to license everything. Well the official policy says if you don’t have Oracle installed and/or running then you don’t have to pay license. So if you can prove it, if you can fence off these virtual machine movements between certain number of physical servers instead of the whole farm, you can prove it. Then Oracle has no case in asking for the whole thing to be licensed. So that’s an improvement. That being said, I know and I cannot blame customers for still being scared because Oracle plays hardball on this thing. So even though the new features escape my typical recommendation to customers, which is architecturally and efficiency-wise maybe not the best thing, but for the political and compliance regions I still recommend, if you are going to run Oracle and VMware, build a database cluster, completely isolate it with your own mention of framework and make no possible way that Oracle can even run outside of this cluster and then you’re fenced off.
On the other side, I could introduce a policy all the way from CIO downwards and I can tell my DBAs “Listen, you’re not allowed Oracle anywhere outside of this cluster”. So now I’m suddenly in control of my Oracle environment. Where in the old days, anybody could install an Oracle database in a server under their desk and when you get an audit you’re suddenly in deep trouble. So you can flip the thing around and use VMware or a private Cloud strategy to your own advantage. So that’s one thing on the other thing.
Oracle introduced, I believe with VMware 5.5, the capability of moving a virtual machine across different clusters. So if you have two VMware clusters that’s managed by one V center and granted that all the pre-conditions are met like shared access and all those things, you could actually move a virtual machine from one cluster to the other up and running, so a live migration of V-motion, this is a very interesting technological feature. However, my opinion is that VMware shot themselves in the foot a little bit on that one because now suddenly Oracle customers are required to run their own separate V-center because otherwise Oracle comes in there and requires them to license the whole thing.
So, if I was VMware I probably would have introduced a feature to also limit that V-motion capability. What I do now is to recommend customers “Make sure that you block any live migration by not allowing to have other servers access the storage area network data of the database”. So you make it physically impossible to do a migration even though the framework would allow it. That way you have a second line of defense for being out of compliance. It’s not ideal and you have to be very careful what you do. But if you understand all the metrics you can use it in your own advantage.
Kay: Right. Okay, but what is the future of storage? I mean in Oracle we’re hearing about software and silicon. What’s coming down the stream at EMC that will make life better for people on the storage side?
Bart: So, EMC is of course doing a lot of things. So we have a diversity of products. What’s happening within the industry you see a slow but steadily consolidation of products and some features are merged from different platforms into one. Because EMC is all about choice, but too many choices that have too much overlap is not a good thing. But this is happening when you have a large company with a lot of products and you do acquisitions and new developments and you also have to support your legacy. So we’re constantly trying to consolidate the number of platforms, the number of architectures and trying to see if we can have a very good product portfolio. So that’s one thing.
Second thing, EMC is much like Oracle. We have a flag ship product from the old days which is called Symmetrixs or better known as V-max these days, which is our high end, big consolidation machine. We also have an all flash array which is typically priced a little bit lower but also in performance is a little bit better. You may lack some of the high end features and snapping and cloning capabilities, but for customers this is a very attractive scenario. So EMC is actually cannibalizing their own high end market by offering something like XtremIO. Which is quite interesting because most customers will hang on to their cash cow for as long as possible and I see Oracle and companies do that. In EMC we are daring to bring that old mainstay and even go after our own market if we have to do something new.
Another example of that is we have a product called Scale IO. Scale IO is basically a full software product that allows you to glue a number of standard servers with just the number of internal discs shoved together and create enterprise storage out of that. Of course, you cannot sell a customer a high end physical piece of hardware if he is going for that approach and only deploying software and then using just a bunch of discs much like Amazon does to create high end storage capabilities. So, we cannibalize our own market. Sometimes our revenue is under pressure, but the long term we know we have to make this move because this is where a big portion of the world is going. Especially if you consider big data analytics, enterprise data lakes as we like to call this, we have to go in this direction. So that’s coming from EMC.
On the Oracle integration side you will see improvements in things like cloning, disaster recovery. Another thing we introduced not so long ago and this is actually a high end feature, some customers have monster databases. I know customers who run Oracle databases that are tens of terabytes, maybe even a hundred terabytes. How do you back it up? You cannot back up a hundred terabyte database in an hour, can you? So either you have very long back up windows that spend multiple days and this impacts your performance. Not even to mention how do you restore this thing? So we have a method to use EMC cloning capabilities in our high end storage to back up directly to another product of ours called Datadomain. Because this is integrated in high end storage, you can actually do this backup in a matter of minutes no matter how big the database is and for the restore it’s exactly the same. You can restore at primary storage speeds even if your database is very very large.
You may expect more and more features from EMC going into the future and we will have better integration with VMware. We’ll have more options in terms of storage tiers. You’ll also see integration of this all flash array extreme IO into our high end platforms because some customers request both the ultimate in availability, double redundancy for them is not good enough. They want 3+1 or 5+1 or n+1 one redundancy, but they also have requirements for very high performance. So you will see a tendency to use extreme IO as one of the storage tiers used in symmetrics. So there is a lot of things on the ropes. When possible and when applicable to Oracle, I will probably blog about it or make some references…So stay tuned for that.
Kay: Will do. Really interesting, Bart. Thank you for that. A question that I ask everybody that I interview, is if you were Larry Ellison what would you do tomorrow?
Bart: I would certainly change the whole way you harass customers with audits, legal and licensing stuff, is the first thing I would change. Second thing, this more on the public side and maybe they are already doing this, I don’t know, but I would probably try to build a better Cloud strategy. One that is really open and flexible and real Cloud instead of the application services provider model. I would probably and I cannot look into their heads is whether this is profitable or not, but customers asking for more integration and cross vendor support. Customers like to run all VMware with EMC or maybe another storage vendor with different servers and they would still like to enjoy things like hybrid compression or smart scans or any other of the features Oracle is offering. So I would try to become more of the open Oracle that they were 10 years ago or more. Instead of the vendor mode they are in now. Now you can completely buy into the Oracle story, buy everything from Oracle. That’s probably selling your soul to the devil a bit because you will have a very hard time ever leaving that model and going to something else. I think even though Oracle may get some good revenue for the short term, but in the end, even for Oracle themselves, I don’t think that is sustainable. Also as Oracle, I would look more into things like big data, analytics and Clouds and not hang onto Oracle database as THE platform to make this work, but also look more into new ways of doing things like scale outs, and multi-node databases and things I like Hadoop or something similar.
Kay: Okay. Bart, is there anything else you would like to add to the interview?
Bart: Well, I’ve probably talked enough for now. So, no, I’m good. Thank you.
Kay: That’s great. Bart, it’s been really interesting interview. It’s been great to speak to you. Thank you very much.
Bart: You too. Thanks for the opportunity.