FYI, this is going to be a rather long article. In fact, I think it
could be more thought of as a white paper. However I feel that a white
paper would block my creativity and will not allow me to tell my not so
unbiased opinion. So whatever opinion I express here is my personal
opinion, not per se the opinion of my current employer.
So the question I am often facing is "Is Veeam enterprise ready?". Honestly I think it is in a modern environment. However sometimes it is too painful to be honest to customers. So I'll just discuss some statements I hear while talking to customers and what are my answers to them. However before I do that, let me start with a small issue I am facing today in my daily life.
In my daily life I drive a Volkswagen Golf. In 99% of the time it is the perfect car for me. The reasons are simple:
Veeam and Physical machines? :
Answer is simple, Veeam doesn't do them. Please start by virtualizing all your servers today! There are no real good arguments not to do it. Often the argument is performance. Well, honestly, performance is not the issue. What is the issue is that you are over-provisioning your virtual environment and mostly don't have the I/O on your storage to allow for good performance. Sometimes I get the argument that local disks have lower latency. True, it's all about physics again but then again, there are great modern solutions like Nutanix, VSAN or PernixData that are taking away that pain by doing server side caching.
So virtualise them all! Often I hear people want to set up private / public clouds (cause it is the big buzz word). Well IaaS is the only cross-platform, cross-application solution that can offer true cloud solutions. So if you want to be cloud ready, this is the way to go. Decouple your machine from physical boundaries. Not only will you free up wasted resources but more importantly:
But I have this SQL physical server :
Let it dump a backup on a CIFS / NFS share that is hosted by a virtual machine. Even better, stop duct taping your infrastructure and virtualize that load already. You are trying to patch an old setup instead of letting it evolve.
But I have this SQL physical server and I don't want to workaround:
Use file to tape to copy your MDF/LDF files or your backup file to tape on a daily basis. In fact this is a feature not a lot of people know exists. Veeam actually allows you to backup files to tape from (physical) machines. Even with VSS intergration
However, tape is not made to handle a lot of small files but rather likes big files to stream nicely to tape. It is why Veeam doesn't backup VM's directly to tape but rather put backup files on tape. So if you want to use it for a physical server, only do it for your big files that contain the data you care about. Those Windows DLL's are really not that useful.
Most painful aspect of this approach? What if the server crashes? You will need to reinstall it, reinstall the application, import the data and hope it all works! Personally I wish you good luck!
Or you could just virtualize the load and even test if your backups can be successfully restored with Surebackup. In fact, for some enterprise customers, Surebackup is one of the main reasons they switch to Veeam. They have the requirement to do recovery tests whether it is by law or company policies. However they don't have the storage nor the man power to do it on a monthly basis. Well guess what, Veeam can do it automatically for you without requiring any extra storage. That is what I call being enterprise ready.
When you will release a Veeam Explorer for Exchange 2003 (and other legacy applications questions):
Well this is the bad thing about virtualization. It has allowed for legacy applications to stay around for waaaaaaaaaay too long. With physical servers, migrating the workload was often combined with executing an upgrade (add a new node to the domain, then demote the old node). Unfortunately, vMotion has made it so easy to migrate that people are just moving the VM's to newer physical servers.
Even worse, when people start virtualizing they use VMware Convertor to P2V. Honestly these are the worst migrations. You have no idea if it will work correctly afterwards, you get stuck with IDE drives and you are duct taping again instead of evolving. I always tell customers, to do a clean installation of the OS and application. Then export and import the data (or add and demote). The one time effort will be bigger but later on you will reap the benefits.
So now once in a while I get the "is it supported" question about Exchange 2003 or SQL 2000. What about them? Why doesn't Veeam support it? Well actually the real question is, why are you still running it? Even Microsoft doesn't support the application itself, so why should a third party tool support it?:
The worst argument I heard was "Yes but our application requires SQL 2000". Please replace your application then and stop duct taping! If this vendor doesn't support 2008 or 2012, it is not worth your time. I'm sorry but it's true!
What about tape?
We have it in v7. Did I get excited when Veeam announced tape support? Not really but a lot of customers where asking for it cause they need cheap long time retention media or because they have "company requirements that state ...". Well sometimes you need to evolve your requirements. But anyway so Veeam has it, so making the product more enterprise ready!
Scalability?
Sometimes I see customers complaining that Veeam doesn't scale well. If I ask them about proxies, they have never heard of it or only installed one. If I ask them about backup storage they have a cheap NAS solution with 5 disks in RAID 6. Well, I'm sorry to tell you, your backup storage is the bottleneck, not Veeam.
Veeam is an easy solution but in enterprise environments, you really need to think about architecture, even for Veeam. If you want fast backups, you need the hardware to support it.
Besides in many Enterprise environments that use legacy agents based backups, Veeam will drastically reduce backup windows. There are even extreme examples of environments going from 48 backup windows (which doesn't fit in 1 day) to 2 hours. That's what I call enterprise ready.
Mixed environments: What about my AS400 or HPUX?
So this arguments, I see a lot. Often a customer has one legacy device running a core application. Somebody set it up 20 years ago, nobody knows how it works but it is really critical. Well maybe it is time to migrate that badboy to an x86 VM. Yes I know it is expensive but you will have to do it one day.
Essentially, this is what you are doing, you are ignoring the problem and making it worse. Sure add more duct tape and it will hold a bit more. But one day it will fail and it might be the end of your company. So before all COBOL experts are dead, please migrate those critical applications while you still can.
Mixed environments: But I don't want to use 2 solutions
Well this argument is the worse. Often people will argue they have to install 2 applications. Well lets say you buy product x which has a plugin for VM's. Well in this case you will have to install and configure the main application and the plugin. Veeam is so easy to install and configure that I bet you it is easier to setup then the plugin itself. In this case you will end up with:
In fact, Veeam can be configured to seamlessly backup all your VM's in your environment. Because instead of selecting individual VM's, you can just backup resource pools, host and folders or datastores. When new VM's are created they are automatically being backed up by an existing job. What automagically?! You heard that right, it means you can refocus your time on innovative stuff instead of duct taping your daily backups. In environments with hundreds VM this is really a game changer.
So daily cost (opex/managemet) is not the issue. But what about licensing?
Well all backup products are capacity based in some way (node, application, socket, tb, ...). Well because you only need to license 5% of your old legacy environment, 95% of the budget will be free to buy Veeam licenses. And boy are you in luck! Veeam is affordable and doesn't have any hidden costs. You have 3 ESXi hosts? You license the ESXi sockets! That's it.
What is the reality? CxO level persons look at "requirement" and they rather buy the "truck/van" instead of buying the golf. The result? A lot of frustration and OPEX costs afterwards. RFP and Tenders are the worst tools ever for IT decision makers because everybody can offer some form of "instant recovery" like bare metal recoveries. The reality is that these features are most of the time horribly implemented and are just not working. Test driving Veeam really helps selling the product, because people just can't believe that it actually does what it promises.
Finally because Veeam is cut out for the job, it will do fast backups and will offer advanced functionality that is just not possible if you do things via an agent based approach. The reason is simple, Veeam is not a duct tapped backup solution but rather a modern data protection product. It can keep up to speed with VMware (ESX) and Microsoft (HV) release cycles because that is the main focus.
So don't buy a product that focuses 95% of its effort on 5% of your environment, but rather buy one that focuses 100% on 95% of your environment and try to evolve that other 5%.
So is Veeam Enterprise Ready?
Well Veeam is Enterprise ready! If it can't fit in your environment, it means you need to rethink your architecture and evolve it so that your architecture is up to par with the latest standards (Or at least not running 14 year old technology)
So stop duct taping today and get ready for some real modern data protection!
So the question I am often facing is "Is Veeam enterprise ready?". Honestly I think it is in a modern environment. However sometimes it is too painful to be honest to customers. So I'll just discuss some statements I hear while talking to customers and what are my answers to them. However before I do that, let me start with a small issue I am facing today in my daily life.
In my daily life I drive a Volkswagen Golf. In 99% of the time it is the perfect car for me. The reasons are simple:
- I drive a lot alone so don't need a lot of space
- It can drive on highways and it can get me everywhere
- It is reliable
- It is not too expensive and since I am not a car freak, I kinda like it. It's that good balance between "I need to compensate for something" and "I don't care what it looks like, if it drives, it drives"
- It can seat 4 people comfortably
- It can drive me
- It can do highways
- It should be reliable (I hope)
- It can fit more then 4 people
- Once a year I can go camping in this baby! I love camping!
- I can move stuff really easily
- Mileage per gallon will be horrible. Damn I hate physics and friction!
- Taxes in Belgium are awful, imagine what this will cost me per year.
- Although it will get me everywhere, visiting my customers in Brussels will be painful. Imagine having to park this monster.
Veeam and Physical machines? :
Answer is simple, Veeam doesn't do them. Please start by virtualizing all your servers today! There are no real good arguments not to do it. Often the argument is performance. Well, honestly, performance is not the issue. What is the issue is that you are over-provisioning your virtual environment and mostly don't have the I/O on your storage to allow for good performance. Sometimes I get the argument that local disks have lower latency. True, it's all about physics again but then again, there are great modern solutions like Nutanix, VSAN or PernixData that are taking away that pain by doing server side caching.
So virtualise them all! Often I hear people want to set up private / public clouds (cause it is the big buzz word). Well IaaS is the only cross-platform, cross-application solution that can offer true cloud solutions. So if you want to be cloud ready, this is the way to go. Decouple your machine from physical boundaries. Not only will you free up wasted resources but more importantly:
- You don't have to set up horrible difficult cluster solutions to have HA. In fact I recently had this discussion with a partner. Most of the times clusters are so badly managed and configured that they are causing more downtime then that they offer high availability. Also they bring a lot of nasty requirements that makes administrating them a hell. On a side node, stop using Physical RDMs. These arguments are invalid:
- Clusters: Read above!
- Performance:VMware released a white paper 5 years ago (ESXi 3.0) saying you don't gain any performance by using RDMs.
- VMDK Size: 62 TB not big enough?
- Your disaster recovery plan will be so much easier cause you are not locked down to a physical server
- Your physical migrations projects will be so much easier. Nothing easier than storage vMotioning a VM to that brand new SAN box you bought.
But I have this SQL physical server :
Let it dump a backup on a CIFS / NFS share that is hosted by a virtual machine. Even better, stop duct taping your infrastructure and virtualize that load already. You are trying to patch an old setup instead of letting it evolve.
But I have this SQL physical server and I don't want to workaround:
Use file to tape to copy your MDF/LDF files or your backup file to tape on a daily basis. In fact this is a feature not a lot of people know exists. Veeam actually allows you to backup files to tape from (physical) machines. Even with VSS intergration
However, tape is not made to handle a lot of small files but rather likes big files to stream nicely to tape. It is why Veeam doesn't backup VM's directly to tape but rather put backup files on tape. So if you want to use it for a physical server, only do it for your big files that contain the data you care about. Those Windows DLL's are really not that useful.
Most painful aspect of this approach? What if the server crashes? You will need to reinstall it, reinstall the application, import the data and hope it all works! Personally I wish you good luck!
Or you could just virtualize the load and even test if your backups can be successfully restored with Surebackup. In fact, for some enterprise customers, Surebackup is one of the main reasons they switch to Veeam. They have the requirement to do recovery tests whether it is by law or company policies. However they don't have the storage nor the man power to do it on a monthly basis. Well guess what, Veeam can do it automatically for you without requiring any extra storage. That is what I call being enterprise ready.
When you will release a Veeam Explorer for Exchange 2003 (and other legacy applications questions):
Well this is the bad thing about virtualization. It has allowed for legacy applications to stay around for waaaaaaaaaay too long. With physical servers, migrating the workload was often combined with executing an upgrade (add a new node to the domain, then demote the old node). Unfortunately, vMotion has made it so easy to migrate that people are just moving the VM's to newer physical servers.
Even worse, when people start virtualizing they use VMware Convertor to P2V. Honestly these are the worst migrations. You have no idea if it will work correctly afterwards, you get stuck with IDE drives and you are duct taping again instead of evolving. I always tell customers, to do a clean installation of the OS and application. Then export and import the data (or add and demote). The one time effort will be bigger but later on you will reap the benefits.
So now once in a while I get the "is it supported" question about Exchange 2003 or SQL 2000. What about them? Why doesn't Veeam support it? Well actually the real question is, why are you still running it? Even Microsoft doesn't support the application itself, so why should a third party tool support it?:
- Exchange 2003 (Article from 2011!) : http://technet.microsoft.com/en-us/library/hh145599%28v=exchg.141%29.aspx
- SQL 2000 : http://support.microsoft.com/lifecycle/search/?sort=PN&alpha=sql
The worst argument I heard was "Yes but our application requires SQL 2000". Please replace your application then and stop duct taping! If this vendor doesn't support 2008 or 2012, it is not worth your time. I'm sorry but it's true!
What about tape?
We have it in v7. Did I get excited when Veeam announced tape support? Not really but a lot of customers where asking for it cause they need cheap long time retention media or because they have "company requirements that state ...". Well sometimes you need to evolve your requirements. But anyway so Veeam has it, so making the product more enterprise ready!
Scalability?
Sometimes I see customers complaining that Veeam doesn't scale well. If I ask them about proxies, they have never heard of it or only installed one. If I ask them about backup storage they have a cheap NAS solution with 5 disks in RAID 6. Well, I'm sorry to tell you, your backup storage is the bottleneck, not Veeam.
Veeam is an easy solution but in enterprise environments, you really need to think about architecture, even for Veeam. If you want fast backups, you need the hardware to support it.
Besides in many Enterprise environments that use legacy agents based backups, Veeam will drastically reduce backup windows. There are even extreme examples of environments going from 48 backup windows (which doesn't fit in 1 day) to 2 hours. That's what I call enterprise ready.
Mixed environments: What about my AS400 or HPUX?
So this arguments, I see a lot. Often a customer has one legacy device running a core application. Somebody set it up 20 years ago, nobody knows how it works but it is really critical. Well maybe it is time to migrate that badboy to an x86 VM. Yes I know it is expensive but you will have to do it one day.
Essentially, this is what you are doing, you are ignoring the problem and making it worse. Sure add more duct tape and it will hold a bit more. But one day it will fail and it might be the end of your company. So before all COBOL experts are dead, please migrate those critical applications while you still can.
Mixed environments: But I don't want to use 2 solutions
Well this argument is the worse. Often people will argue they have to install 2 applications. Well lets say you buy product x which has a plugin for VM's. Well in this case you will have to install and configure the main application and the plugin. Veeam is so easy to install and configure that I bet you it is easier to setup then the plugin itself. In this case you will end up with:
- Veeam for 95% of your machines
- Product x for that remaining 5% of your physical legacy machines, which you will migrate in the end anyway and not having to configure any plugin
In fact, Veeam can be configured to seamlessly backup all your VM's in your environment. Because instead of selecting individual VM's, you can just backup resource pools, host and folders or datastores. When new VM's are created they are automatically being backed up by an existing job. What automagically?! You heard that right, it means you can refocus your time on innovative stuff instead of duct taping your daily backups. In environments with hundreds VM this is really a game changer.
So daily cost (opex/managemet) is not the issue. But what about licensing?
Well all backup products are capacity based in some way (node, application, socket, tb, ...). Well because you only need to license 5% of your old legacy environment, 95% of the budget will be free to buy Veeam licenses. And boy are you in luck! Veeam is affordable and doesn't have any hidden costs. You have 3 ESXi hosts? You license the ESXi sockets! That's it.
- If you want to run 1 VM or 100 VM's or 1000 VM's on those hosts, Veeam doesn't care.
- If you need to deploy new VM's to meet new business challenges, you don't have to worry about backup licenses, you will always be compliant.
- You want to deploy 5 SQL servers instead of 1 to split the load? Well no need to count and report those nodes, cause it doesn't matter. You have Exchange and/or Sharepoint? It's okay, you don't need to license any applications separately/
What is the reality? CxO level persons look at "requirement" and they rather buy the "truck/van" instead of buying the golf. The result? A lot of frustration and OPEX costs afterwards. RFP and Tenders are the worst tools ever for IT decision makers because everybody can offer some form of "instant recovery" like bare metal recoveries. The reality is that these features are most of the time horribly implemented and are just not working. Test driving Veeam really helps selling the product, because people just can't believe that it actually does what it promises.
Finally because Veeam is cut out for the job, it will do fast backups and will offer advanced functionality that is just not possible if you do things via an agent based approach. The reason is simple, Veeam is not a duct tapped backup solution but rather a modern data protection product. It can keep up to speed with VMware (ESX) and Microsoft (HV) release cycles because that is the main focus.
So don't buy a product that focuses 95% of its effort on 5% of your environment, but rather buy one that focuses 100% on 95% of your environment and try to evolve that other 5%.
So is Veeam Enterprise Ready?
Well Veeam is Enterprise ready! If it can't fit in your environment, it means you need to rethink your architecture and evolve it so that your architecture is up to par with the latest standards (Or at least not running 14 year old technology)
So stop duct taping today and get ready for some real modern data protection!
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.