Upgrading to DX APM 20.2 - need sizings for small installation on premises

[RyanJ]

Hi,

We are upgrading an existing DX APM 9.x installation (on premises) and are looking for realistic sizings for a small number of agents and metrics for a DX APM 20.2 on premises install.  Documentation is very focused on large scale installations supporting thousands of agents/metrics. Any assistance or feedback from folks who have done this already would be much appreciated.

Thanks

Ryan

------------------------------
Director
Response Systems / Facet Consulting
------------------------------

[Priyesh Patel]

How many agents are we talking here?
If it is in few hundreds, is a customer open for DX SaaS (comes with DX APM and more)?

------------------------------
Enterprise Solutions Architect
Broadcom
------------------------------

[Jörg Mertin]

For our test environments/trainings on DX APM only, we use 4 servers with each server running CentOS 7.7 OS  on 32Gb RAM +16 CPUs + 300 GB HD.
Make sure the disks are fast though.
This for a APM only workload! No AXA and no OI!

Can't say what load the system will really be able to handle, as this is for testing only we use it. I would however say that this is the real minimum install you can base your sizing on.


------------------------------
Customer Solution Engineering
Broadcom Inc.
------------------------------

[Osvan Widyowibowo]

Hi Jorg,

I have the same question about the sizing for only less then 100 agents for our customer. And the company's policy does not allowed to opt for SaaS. In your small sizing recommendation, there are 4 servers. I assume those are for app layer, elastic search layer, master layer, and NFS storage system. My question is, for less <100 agentrs, can NFS storage be deployed in Master Layer? The DX APM installation guide in the link below says that "For small deployments, the NFS server can be deployed on the master node." Can we use your server sizing (32GB RAM +16 CPU) for master layer and NFS storage? 

https://techdocs.broadcom.com/us/en/ca-enterprise-software/it-operations-management/dx-platform-on-premise/20-2/installing/set-up-the-environment/Kubernetes---Pre-installation-Tasks.html

Your assistance would be much appreciated. 

Thank you.

Osvan W.

[Jörg Mertin]

You need to know that the setup you will build is very small, and the required resources will be high compared to the gain.

You can use the master as NFS storage. That  won't really matter on a small setup.

I have however recently setup a OnPrem instance 20.x and It required 7 hosts sized from 8vCPU + 32GB ram to have the installer install a small instance with no redundancy. Note that this is for APM, AXA and OI.

For 4 machines, you will need to disable the error-checks on the installer side. And please remember that this is a non-supported setup!

You should really take into account about what Priyesh mentioned below.

------------------------------
Customer Solution Engineering
Broadcom Inc.
------------------------------

[Danny Saro]

Classification: Confidential

Hi Osvan,

 

If you install with less than 3 Hot Elastic Search nodes that is considered a small installation, the sizing for a small installation is not in the documentation (I guess because it is not supported for production environments?) but the installation script shows below table of required CPU and RAM for the one ES Hot node and the Application nodes.

 

Specify the part you want to install(You can choose more than one):

* 1: DX Application Performance Monitoring

  2: DX Operational Intelligence

  3: DX App Experience Analytics

Enter your choices as comma-separated values: 1

========================= Hardware overview =============================

========== Elasticsearch Layer     - Required / Available / NodeName ============

[ PASSED ] Elasticsearch 1         -   1 CPU  / ... CPU   / ...

[ PASSED ] Elasticsearch 1         -  11 RAM  / ... RAM   / ...

========== Application Layer       - Required / Available / NodeNames =====================

[ PASSED ] Application Layer Nodes -  17 CPU  / ... CPU   / [..., ..., ... ]

[ PASSED ] Application Layer Nodes - 105 RAM  / ... RAM   / [..., ..., ... ]

==========================================================================

 

I did initially run into OOM issues during execution of install script as long as my numbers were lower than the above. It installed when using 5 x (8 CPU / 32GB RAM) with one ES Hot and with The master node tainted so that it also run Application Layer pods (this is a test environment).

 

Hope this helps,

 

Cheers,

Danny

 

::DISCLAIMER::

---

The contents of this e-mail and any attachment(s) are confidential and intended for the named recipient(s) only. E-mail transmission is not guaranteed to be secure or error-free as information could be intercepted, corrupted, lost, destroyed, arrive late or incomplete, or may contain viruses in transmission. The e mail and its contents (with or without referred errors) shall therefore not attach any liability on the originator or HCL or its affiliates. Views or opinions, if any, presented in this email are solely those of the author and may not necessarily reflect the views or opinions of HCL or its affiliates. Any form of reproduction, dissemination, copying, disclosure, modification, distribution and / or publication of this message without the prior written consent of authorized representative of HCL is strictly prohibited. If you have received this email in error please delete it and notify the sender immediately. Before opening any email and/or attachments, please check them for viruses and other defects.

---

[Priyesh Patel]

G'day Ryan,

I received your reply to my question.

Per your reply, required APM Agent quantities is very small (below hundred) and it ain't worth building an on-premise set up just for APM for this customer. But, this answer is not the solution you are looking for, hence let's look at this as an Opportunity to establish strategic foot with this customer. And that begins with expanding beyond Silos, meaning, explore with customer beyond just APM. The potential other areas can be Digital Experience Monitoring, Network Monitoring, Infra Monitoring, AIOps & Automation. 

You may ask but why not just APM on-premise for such a small estate, the rationale is, our solution is aligned to offer AIOps, Monitoring and Automation in a unified cloud agonistic platform (can be built on K8s or Openshift) and with it customer can optimize their spend with multiple point tools and benefit from a single unified platform solution which gives the collective intelligence across Digital customer experience (Mobile & Web), Modern APM, Infra Monitoring and NetOps with integrated Automation capabilities (self healing use cases) and have access to Service Analytics, Alarm Analytics, Performance Analytics and so on. Even, Industry analyst like Gartner too recommend having collective observability which comes via Domain agnostic AIOps solution which has data ingested via all possible Monitoring tools. If this conversation generates more interest with customer, your game elevates to next level which is APM + other solutions and that's an opportunity.

You will then have two ways to deliver it. Here are few Options for your consideration to bounce with customer!

Option1: Revalidate the exact reason with customer, why not DX SaaS?
Let's try to understand, is this customer strictly no-no for any SaaS solution? Meaning, customer doesn't even us any SaaS solution, e.g. Salesforce, ServiceNow? If that's the case we don't win with Option 1 however if there is any technical concern, we can definitely work towards addressing them and if that goes well, Option 1 is best suited for such small size customers.

Option2: Assuming customer, has skills to manage K8s or Openshift environments and willing to expand beyond just APM with other areas like AIOps and more, it makes sense to build an on-premise set up. Based on actual information, sizing can be determined.

Option3: Assuming customer is not willing to Expand beyond APM but open to a shared service platform in local region, that's when Your consulting firm can build a multitenant platform and on-board multiple small size customers in them. Based on actual information, sizing can be determined.


​

------------------------------
Enterprise Solutions Architect
Broadcom
------------------------------

[evelina franco]

We are upgrading an existing DX APM 9.x installation (on premises) and are looking for realistic sizings for a small number of agents and metrics for a DX APM 20.2 on premises install.  Documentation is very focused on large scale installations supporting thousands of agents/metrics. Any assistance or feedback from folks who have done this already would be much appreciated.

[Haruhiko Davis]

There is no upgrade path from any on premise version to DX APM, due to the architecture of the solution.
What can be done is ease the migration that older agents can still report to DX APM via Cloud Proxy agent, which acts like a collector to forward agent metrics to DX APM. Once you've upgraded your agents, you can continue to use Cloud Proxy or remove it entirely and point the agents directly at DX APM's ingestion URL.

------------------------------
Solution Engineering
Broadcom
------------------------------