What's the best way to trial hyper-converged (HCI) solutions?

27

We all know that it's important to conduct a trial and/or proof-of-concept as part of the buying process. 

Do you have any advice for the community about the best way to conduct a trial or POC? How do you conduct a trial effectively? 

Are there any mistakes to avoid?

Guest
As seen in
27 Answers
Pinkesh PatelUser

The best way to conduct a trial or POC of any solution is to first try it for certain applications for a subset of users from a few strategically selected locations. I would recommend the following steps :

* Gather Requirements
* Size the Solution
* Establish Success Criteria
* Prepare for Install
* Product Installation and Configuration
* Health Check
* Present Results

16 August 18
Jefe de Infraestructura Tecnológica at a non-tech company with 1,001-5,000 employeesReal User

If you want to do a proof of concept about HCI, then I recommend that you do it and that you try all the hypervisors of your choice, with any of them you will find very good results.

Of course, performing a proof of concept in HCI equipment similar to your work load in production would be the best, as it has an almost real test.

In my case, I have migrated VMs that have databases like MSQL, postgres, mysql and in all of them I have better response times in read and write operations and zero data corruption.

The cloning of VM is very fast and the simplicity of the HCI operation helps me to concentrate on other activities.

The operation in HCI is very simple.

26 July 18
Jon SeetReal UserTOP 10

Nutanix provides a test drive of their solution via: https://www.nutanix.com/test-drive-hyperconverged-infrastructure/ . It can be personalised in minutes.

26 July 18
Bob WhitcombeReal UserTOP 5LEADERBOARD

Selecting an HCI path is pretty straightforward and it goes through the cloud. You first select your workloads and what performance is needed for success. Since the key differentiation across HCI platforms today is software - you should be able to construct a target load of the apps you want to test and run them in a vendors cloud sandbox. You want to align your hardware solutions so you can leverage your existing support models and contracts, but you are testing software platforms for usability, performance and adaptability to your current operations model.

Once your workload homework is complete and your have selected an application type, VDI, OLTP, Data Warehouse etc, and determined worst case response times, you can throw a target workload to the cloud for evaluation. At this point you are looking for hiccups and deployment gotchas. HCI and cloud processes may be new to you - so you may need to stretch beyond your deployment models. This is a good thing. Recognize HCI is a leading edge trend and is one step removed from the cloud - which is where you will be in 5-10 years.

You want to look for key software features that lower the cost and complexity to manage this installation. But for a corner case or three, most applications will fit squarely in the middle of the "good" zone for today's SSD based HCI solutions.

With cloud testing of a target HCI platform you should learn how your applications perform, see key features you really really want and satisfy yourself that these systems can be managed without significant incremental effort by your current staff.

Then you do the grid - is the target aligned with my current hardware vendor; endorsements from people running similar applications; killer features and a drop dead signing bonus that justifies adding this platform to my portfolio of aging IT equipment? If and only If you come down to a near tie between two vendors should you go to the trouble of a full meal deal on-site PoC. They may not provide any more information than the version in the cloud, require physical hosting on your site, need an assigned project manager and then you get to deal with the loser - who may very well be your current vendor - and what a joy that will be.

25 July 18
John StraubUser

In general, it is difficult to test HCI because it requires the purchase or lease of at least one HCI node on which to install a significant part of the primary database. The best way may be to ask the vendor to install the data on one of their HCI nodes, then test HCI performance on that node, either remotely or at the vendor's lab. ALIGN's partner, Experis, routinely allows potential customers to test their data set on their HCI equipment.

Conversion to HCI may require a complete conversion of legacy hardware to the HCI platform, a fact that complicates test installations.

25 July 18
Roden PinedaUser

Gathering crucial information about your existing environment and requirements and comparing them with what the technology has to offer.
Contacting multiple Vendors is very important to know which can best deliver your needs.
Support level is very crucial as well and not just at the point of sales so looking at the profile of the vendor would allow us to foresee any underlying issues for that matter.

25 July 18
MohamedMostafa1Real UserTOP 5

There are several ways to evaluate HCI Solutions before buying, Customers need to contact HCI Vendors or one of the local resellers who propose the same technology.
Both of HCI Vendors and Resellers will be able to demonstrate the technology in Three Different scenarios like :

1 – Conduct Cloud-Based Demo, in which the presenter will illustrate product features and characteristics based on a ready-made environment and the presenter will be able to demonstrate also daily administration activities and reports as well.

2 – Conduct a Hosted POC, in which the presenter will work with the customer in building a dedicated environment for him and simulate his current infrastructure components.

3 – Conduct Live POC, in which the presenter has to ship appliances to customer’s data center and deploy the solution and migrate/create VMs for testing purpose and evaluate performance, manageability & Reporting.

If the vendor or a qualified reseller is doing the POC, there should be no mistakes because it’s a straightforward procedure.

25 July 18
Jiten KamothiReal UserTOP 20

Please keep in mind that POC is proof of concept not the exact solution of the your objective. Define everything (objective, challenges, resources to manage scenario after POC etc) as points and find the product accordingly. There will always be obstacles / challenges on COST in non-i.t. industry. For example, we have ERP running on SQL 12 R2 database, which was licensed for physical cpu core. So, when I planned for HyperV or VMware, the MS license was the major costing element even in doing POC. Finally, i drafted all brain inputs on paper and started designing new flowchart for setting up ERP server, near DR server, Report Server and Database servers. I finished these without buying any licenses i.e. Win-12 Ent or SQL-12 R2. In short, the plan in mind must be drawn on soft form and discuss the drawing with internal/external team to come to conclusion.

25 July 18
AndreNagelConsultant

If I read the feedback in the market there are many success stories. Most vendors will provide platforms for POC’s. I recommend that you engage with them.

If your applications are virtualized already it should not have any issues. I recommend a POC with two tests. 1- Simple single server solution to ensure that the migration method is tested. 2. A 3 tier application with Application servers and DBase servers.

25 July 18
Henrik VaageReal User

Hi,

Here is our Best Practices for performing a POC on an All Flash HCI Cluster

1. Initialize the disks
• This is more realistic - Applications don’t read data they haven’t written
2. Use longer Test Duration:
• SSDs / All flash systems show very high performance initially – which isn’t realistic
• Tests should be long running (12 – 24 hours depending on write load)
3. Larger Working set size
• Ability to support larger working set size (several TBs / node) is a key differentiator for all flash
• For smaller working set size hybrid will perform similar to all flash (as data is coming from flash cache)
4. Enable Dedupe and compression on the whole cluster
• Most enterprise workloads see significant storage efficiency savings – as such this is not optional
• Expecting users to choose this on a workload by workload basis is not realistic with 100s of VMs
5. Workload Read / Write Mix:
• Reasonable write percentage (30%/35% at least) in the workload
• 100% - 95% read workloads are not realistic in most environments
6. Similar hardware configuration:
• Pay careful attention to the BOM – CPU, Memory and SSD (type and count for cache/WL and persistent data SSDs)
• Additional performance may come at a trade-off: cost or manageability
7. Similar software configuration:
• Pay careful attention to the resiliency setting – that needs to be identical
• Comparing Replication Factor = 2 performance with Replication Factor = 3 performance is not comparing apples to apples

Performance Results / Comparison
1. What to look for in performance results:
• IOPS / Throughput
• Latency – not all workloads show sub millisecond latencies (even on all flash!)
• Consistent Latency / standard deviation of latency
2. Variance in performance seen across VMs
• Many vendors typically have huge differences in IOPS delivered to different VMs based upon where the VM is located (vs client), after VMotions, network latency etc.
3. Compare Cost:
• Especially important when comparing different hardware configurations
• Metrics: $/GB Effective, $/IOPS and $/VM (with given latency minimum to VM)

25 July 18
Kirtikar OjhaUser

Conducting the POC is always desirable. Point to be taken care off is at your application side. The Licenses policies for your OS is they are proprietary in nature as this aspect can derail your successful POC. Problem arises when you are migrating some workloads on the HCI. Nutanix to my mind is the most suited environment in the HCI platforms.

25 July 18
JohanVan AardeVendor

When you do a trial or PoC, Make sure you look at things like:
First of all, determine what your requirements are for HCI. Then start to look at the various vendors that play in that area.

1) Use 3rd party tools for the performance testing, and not vendor specific tools
2) Do hard crash failure testing. For example, just yank the power cables out of the POC kit while busy with the PoC. Return the same start up and see the stability of the environment and also assess for any data loss.
3) How many Management consoles do you require to run the environment? (Monitor, Manage, Performance etc)
4) Consider the actual hardware size. Power & Cooling cost $$$
5) Do you require additional and/or different hardware to add features like File Shares, Block Storage, Containers etc.
6) The HCI environment MUST be able to support multi Hypervisors
7) HA & Fault Tolerance should be supported out of the box.
8) Upgrading should be seem-less with little to NO down time.
9) Where is the HCI vendor on the Gartner Magic Quadrant
10) Is Micro Segmentation part of the solution or not?
11) Expansion should be seem-less and ALSO allow for either Storage Heavy, Compute Heavy or a combination of the same. You do not want to sit in a situation that you have to procure massive amounts of compute just because you require additional storoage
12) Expansion should be done in minutes, not days, weeks or months.
13) If a disk fail, ONLY that disk should be excluded from the storage pool, and nothing more...

25 July 18
Vicheth TorUser

What's the best way to trial hyper-converged (HCI) solutions? Exactly

We all know that it's important to conduct a trial and/or proof-of-concept as part of the buying process. Yes, before purchased my team to do POC one month and through one by one of feature and function.

Do you have any advice for the community about the best way to conduct a trial or POC? It’s good view and know more details of products before decide to use it.

How do you conduct a trial effectively? I think there is a perspective on conducting experiments in HCI that ... and experience but basically they can be understood as tests or trials of ideas.

Are there any mistakes to avoid? No.

25 July 18
Phak YunUser

I would recommend to do POC to see the performance and workload before purchasing any new product as the marketing competition of HCI. HEP Simplicity, VxRail, Nutanix, CISCO Hyper flex...

25 July 18
AmitSingh5Consultant

Can we know what the user is looking for what his requirements are??

As there is no such simulation program for HCI, however, the user can go thru the test cases provided by multiple vendors to see if HCI is compatible or not.

Moreover, based on his requirement we can suggest do's and dont's for HCI

25 July 18
anush santhanamConsultant

Hi,

When evaluating HCI, it is absolutely essential to run a trial/POC to evaluate the system against candidate workloads it will be expected to run in production. However, there are quite a few things to watch out for. Here is a short list:

1. Remember that most HCI depend on a distributed architecture which means it is NOT the same as a standard storage array. What that means is that, if you want to do any performance benchmarking with tools such as IOMeter, you need to be extremely careful in the way you create your test VMs and how you provision disks. Guys such as Nutanix have their own tool X-Ray. However I would still stick to a more traditional approach.
2. Look at the list of apps you will be looking to run. If you are going to go for a KVM type of a hypervisor solution, you need to see if the apps are certified. More importantly, keep an eye out on OS certification. While HCI vendors will claim they will and can run anything and everything, you need the certification to come from the app/OS OEM.
3. Use industry standard benchmarking tools. Remember unless you are using a less “standard” type of a hypervisor such as KVM or Xen, you really don’t need to be wasting your time with the hypervisor part as VMWare is the same anywhere.
4. Your primary interest should be the storage layer without question and the distributed architecture. Remember with HCI, the computer does not change and hypervisor (assuming VMWare) does not change. What changes is the storage. Next there are the ancillary elements such as management and monitoring and other integration pieces. Look at these closely.
5. Use workload specific testing tools. Examples include LoginVSI, jMeter, Paessler/Bad boy for web server benchmarking etc.
6. Finally, remember to look at the best practices on a per-app basis. The reason I suggest this is because of the following. You may have been running an app like Oracle in your environment for ages in a monolithic way. However when you try the same app out in HCI it may not give you the performance you want. This has to do with the way the app has been configured/deployed. So looking at app best practices is something to note.
7. If you are looking at DR/backup etc, then evaluate your approaches. Are you using any native backup or replication capability or are you using any external tool. Evaluate these accordingly. Remember your RTO/RPO. Not all HCI will support sync replication.
8. Finally if you are looking at looking at native HCI capabilities around data efficiency etc (inline de-dupe and compression), you will need to design testing for these carefully.
9. Lastly, if you are looking at multiple HCI products, ensure you use a common approach across products. Otherwise your comparison will be like looking at oranges and apples.

Hope this helps.

25 July 18
Bubot AlumbroUserTOP 20

Do you have any advice for the community about the best way to conduct a trial or POC?
One way is for the supplier to bring in a test server with 1 or 2 workstations. It should have significant amount of test data so that the user will be able to try the features. It would be best if test data to highlight features can be tested.

25 July 18
Oshri SharonReal User

Nutanix provides cloud based demo platform scheduled upon needs. Best value with Nutanix to my opinion is moving to Acropolis HV (AHV). Going this direction will require to check compatibility for Backup solution to support AHV, Virtual Appliances if any, Mission Critical Application support for AHV. To my opinion, infrastructure & performance are at most cases not an issue today because all HCI technologies provides much more simplicity and performance than legacy solutions. Major topics to test besides Hypervisor alternative is Cloud integration and Automation level of the platform to the application level.

25 July 18
Peter NormanUser

There are obviously a number of HCI solutions available to the market at the moment. The market is getting quite crowded and the vendors all are keen to get their products in customer’s data centers. This generally means that they are open to a ‘Try before you Buy’ type arrangement. So, for example: if you have chosen the HCI platform to be VXRail and have come to an agreement on the benchmark it needs to reach and the capabilities that will be tested; then the vendor will generally provide the system for a pilot which the customer will buy if it meets all the agreed criteria. Or send it back with no penalty if it does not.

In this model, the implementation services will need to be paid for, which cannot be refunded.

There are numerous resources on the Internet and from the HCI vendors on running pilots. Suggesting the types of tests and configuration best suited to their products, including the load generation tools to use.

Testing should always include what is important to your business. This may be:

* High availability of system (simulating multiple subsystem failures)
* Performance
* User experience
* Administering the system
* Non-disruptive upgrades
* Expansion of nodes (Scale-out)
* 3rd party integration
* Vendor support

All this can be run by a competent pilot group representing IT and the business. The alternative is to engage a consultancy group (such as TDLogicalis) to Consult, Define and Run the pilot on the customers' behalf.

25 July 18
NileshThakareUser

Would recommended take HCI POC from nutanix as web based user interface. HP Simplivity gives on site hardware for POC test ...DELL Vx rail they dont have option to provide POC as per understanding..... But best HCI as experience is NUTANIX ...

25 July 18
Jimmy-ChangUser

Suppose you want to try hyper-fusion, you can recommend starting with cisco hyperflex. POC or testing can be done according to cisco's poc manual. In my experience, hyperflex's dedupe and clone are very efficient, and can support external storage and use FI to centrally manage hardware and software settings. The hyperflex setting is very simple, but you need to focus on the pre-emptive network vlan and ip planning, because it will take about forty ip.

25 July 18
Carlos Roberto Condori FloresUser

If you are going to run your own App, you must ensure the sizing of resources correctly as it happens that many seek to save and in the long term can cause headaches.

If only to observe the interface, each manufacturer offers a demo environment and this leads to VMware, Acropolis, hyper-v "are the most used."

24 July 18
KevinWrightUserTOP 20

My best advice for conducting a trial/POC of an HCI solution is to maintain a great relationship with the vendor(s) that you are considering for your HCI solution.

The vendor Account Managers can assist with setting up POCs and get you the equipment as well as Professional Services that will assist in ensuring that you are getting a solution to meet your requirements.

24 July 18
DanAndersonUser

Ask the right questions of your provider:
Accessing your data off local drives is the secret sauce that makes HCI dominate over traditional three tier architectures. Select a provider that always guarantees data locality ( Nutanix does).

Beware converged traditional solutions that masquerade as HCI... if they are selling a storage array and compute it is NOT HCI!

Ask your provider about their vision to use/migrate to the cloud, does it align with yours?

Do you want the option of moving away from your current hypervisor to save money?

24 July 18
Chris CurtinConsultant

The best way to trial HCI is:

Determine objectives (desired end state) and goals that moving to HCI will meet for the business and IT
Identify workload(s) and success metrics
Work with the business to identify the business value of moving to HCI (this is more than just a simple TCO analysis)

24 July 18
Ram NangunooriUser

Hi
First do some home work, in terms of what HCI means?
Once you convinced that is the approach you want to take, release RFP.
Identify potential vendors. Reduce to the potentials. Prepare your questionnaire. Setup lab.
Call them for presentations
Ask them to provide the POC in your lab.

24 July 18
Sufyan KhanUserTOP 20

Hi,
The buying is the last step:
I suggest to follow below strategy:
1. The best way is to have its live demonstration in production environment (if any opportunity is available). Connect with those HCI partners who already deployed HCI solutions.
2. specify your environment, so that a particular instruction set can be furnished?
3. If you are planning to have latest tech refresh then specify which hyper-visor maps your complete required solution.
4. are you planning VMware/Hyper-V/KVM or else?
5. Trials can be recommended openly but need some information like what kind of workload you are going to execute, how much compute and storage will be required etc.

24 July 18
As seen in

Sign Up with Email