Reltio Connect

Dear Reltio Product Management,
I started a reindex job this morning to try and solve a problem we're seeing in the tenant data. There are 10M records in the tenant (not really a big number by today's standards) and the job continues to run after almost 5 hours and yes I started it as a distributed job (although one wonders why I'm even asked to make the decision. Just do it automatically if you think it'll speed things up!) I've no idea when the job is going to finish; I can only speculate and sweat while I have a hard deadline to get things working for the client by later this evening to meet some high-focus, user-facing activities. Anxiety is running high.

Being an avid and frequent user of services on AWS, I ask you to study their model closely as you think about the Reltio Console and the instrumentation it offers. With AWS I'm in complete control of the resources. I can create and destroy any service at will. I can pick the type of EC2 servers I wish to use and their horsepower; I can create load balancers and auto-scaling groups and set thresholds that automatically add resources to meet my requirements for getting work done at the rate I choose.  If I want to spend an extra $10 or $100 to cut the time for a job in half, I can do that with a few clicks of my mouse which is precisely what I would be doing right now in the Reltio Console if it were offered. It is this kind of instrumentation and control that helped propel AWS to the global standard it is today. But I need not tell you guys this because you use AWS (and GCP and Azure) for all the same reasons mentioned.

As I sit here now, writing this post and periodically refreshing the Jobs screen in the Reltio Console, I look forward to the kind of innovation in a future release of Reltio that feels more and more like AWS in terms of complete self-service control. 

Interested in others' comments and thoughts on this.
Curt


------------------------------
Curt Pearlman
PwC
Agoura Hills CA
------------------------------

Hi Curt,

Good to hear from you and thank you for your suggestion. Broadly speaking, we agree with your thinking. We admire not only Amazon/Google/Microsoft, but also Snowflake, Salesforce and others who have made enterprise software "just work" at scale.

We are approaching this from two dimensions. 

First, every component will come with enhanced observability and status/error reporting. This will allow our customers to know more definitively the status and progress of tasks in progress. You will start showing up in our enhanced data loader and other components we are currently working on. In a big data system there will invariably be longer running tasks and a key problem for us to solve is how to provide accurate estimates for completion time for these jobs. 

Second, we do expose a of powerful knobs and dials allowing customers to configure the system behavior but it does come at an increased cognitive cost to the operators. To strike the right balance we intend to set a number of these dials at smart defaults or hide them entirely. 

Specifically, we will re-examine the re-index task based on your feedback. Do you have any other pointed asks of this nature ? Would love to get a list that we can chase down.

Cheers,

-ansh

------------------------------
Ansh Kanwar
VP, Product Management
Platform and Infrastructure
Reltio
------------------------------

Original Message:
Sent: 07-26-2022 18:48
From: Curt Pearlman
Subject: AWS has set the bar. Let's follow suit..

Dear Reltio Product Management,
I started a reindex job this morning to try and solve a problem we're seeing in the tenant data. There are 10M records in the tenant (not really a big number by today's standards) and the job continues to run after almost 5 hours and yes I started it as a distributed job (although one wonders why I'm even asked to make the decision. Just do it automatically if you think it'll speed things up!) I've no idea when the job is going to finish; I can only speculate and sweat while I have a hard deadline to get things working for the client by later this evening to meet some high-focus, user-facing activities. Anxiety is running high.

Being an avid and frequent user of services on AWS, I ask you to study their model closely as you think about the Reltio Console and the instrumentation it offers. With AWS I'm in complete control of the resources. I can create and destroy any service at will. I can pick the type of EC2 servers I wish to use and their horsepower; I can create load balancers and auto-scaling groups and set thresholds that automatically add resources to meet my requirements for getting work done at the rate I choose.  If I want to spend an extra $10 or $100 to cut the time for a job in half, I can do that with a few clicks of my mouse which is precisely what I would be doing right now in the Reltio Console if it were offered. It is this kind of instrumentation and control that helped propel AWS to the global standard it is today. But I need not tell you guys this because you use AWS (and GCP and Azure) for all the same reasons mentioned.

As I sit here now, writing this post and periodically refreshing the Jobs screen in the Reltio Console, I look forward to the kind of innovation in a future release of Reltio that feels more and more like AWS in terms of complete self-service control. 

Interested in others' comments and thoughts on this.
Curt


------------------------------
Curt Pearlman
PwC
Agoura Hills CA
------------------------------

Ansh, here are some thoughts in response.
The overall theme is self-service control to reduce my reliance on inefficient, human support functions (Reltio Support, Pro Serve, Customer Success,..) and insight to help me make better decisions while exercising self-service control. 

For any Job I might want to start (Reindex, Consistency Check, Bulk Update,..):

• Predict its duration, even if you have to spend a minute while I watch a spinner, evaluating some things so I can decide if now is the right time or not to do this.
• Explain the prediction and advise if there is anything abnormal about the config etc.. that is behind that prediction. (E.g. "Your index job will is estimated to take 12 hours to complete. Primary Factor: Your tokenization scheme is inefficient and generating many more tokens than expected. Run the "Analyzer Match Rules" job for more information. )
• While the job is running, show me current status with an updated prediction of its completion time
• Let me add resources to it if I wish while its running, and advise me of the cost of those add'l resources. (This assumes I have a pre-arranged "account with Reltio" to which I can accrue discretionary charges. Pls create this ability. Perhaps another role, Customer_admin_with_spend_authority)
• Let me kill the job if necessary but advise of any negative effects If I proceed. 

Again, its all about insight and self-service instrumentation, without any need to contact Reltio Support or a POSTman or python developer. 
There are numerous jobs today that can only be invoked via a Support Ticket, and these have been well-documented and prioritized by the Support team for the needed conversion to self-service control. These include (but not limited to):

• Clone tenant (e.g for a month for add'l testing prior to a go-live)
• Copy tenant (e.g. from Prod to Test)
• Backup Tenant
• Disable / Enable Streaming
• Clear the Tenant of all data 
  • Clear the tenant of all data for a specific source
• (and more..)

Lastly, the Jobs that exist today on the "Create Job" screen have parameters that in many cases are quite opaque and there is no documentation to explain why I would set any one of them a certain way, nor does the little abstract under the parameter help much. E.g. from Reindex Job

Delete Old Column Families

This parameter when turned on will delete all Column Families.

No one outside of Reltio engineering is going to have a clue what to do with the parameter above. And it sounds pretty dangerous frankly. 

Hope this helps
Curt

------------------------------
Curt Pearlman
PwC
Agoura Hills CA
------------------------------

Original Message:
Sent: 07-29-2022 10:56
From: Ansh Kanwar
Subject: AWS has set the bar. Let's follow suit..

Hi Curt,

Good to hear from you and thank you for your suggestion. Broadly speaking, we agree with your thinking. We admire not only Amazon/Google/Microsoft, but also Snowflake, Salesforce and others who have made enterprise software "just work" at scale.

We are approaching this from two dimensions. 

First, every component will come with enhanced observability and status/error reporting. This will allow our customers to know more definitively the status and progress of tasks in progress. You will start showing up in our enhanced data loader and other components we are currently working on. In a big data system there will invariably be longer running tasks and a key problem for us to solve is how to provide accurate estimates for completion time for these jobs. 

Second, we do expose a of powerful knobs and dials allowing customers to configure the system behavior but it does come at an increased cognitive cost to the operators. To strike the right balance we intend to set a number of these dials at smart defaults or hide them entirely. 

Specifically, we will re-examine the re-index task based on your feedback. Do you have any other pointed asks of this nature ? Would love to get a list that we can chase down.

Cheers,

-ansh

------------------------------
Ansh Kanwar
VP, Product Management
Platform and Infrastructure
Reltio

Original Message:
Sent: 07-26-2022 18:48
From: Curt Pearlman
Subject: AWS has set the bar. Let's follow suit..

Dear Reltio Product Management,
I started a reindex job this morning to try and solve a problem we're seeing in the tenant data. There are 10M records in the tenant (not really a big number by today's standards) and the job continues to run after almost 5 hours and yes I started it as a distributed job (although one wonders why I'm even asked to make the decision. Just do it automatically if you think it'll speed things up!) I've no idea when the job is going to finish; I can only speculate and sweat while I have a hard deadline to get things working for the client by later this evening to meet some high-focus, user-facing activities. Anxiety is running high.

Being an avid and frequent user of services on AWS, I ask you to study their model closely as you think about the Reltio Console and the instrumentation it offers. With AWS I'm in complete control of the resources. I can create and destroy any service at will. I can pick the type of EC2 servers I wish to use and their horsepower; I can create load balancers and auto-scaling groups and set thresholds that automatically add resources to meet my requirements for getting work done at the rate I choose.  If I want to spend an extra $10 or $100 to cut the time for a job in half, I can do that with a few clicks of my mouse which is precisely what I would be doing right now in the Reltio Console if it were offered. It is this kind of instrumentation and control that helped propel AWS to the global standard it is today. But I need not tell you guys this because you use AWS (and GCP and Azure) for all the same reasons mentioned.

As I sit here now, writing this post and periodically refreshing the Jobs screen in the Reltio Console, I look forward to the kind of innovation in a future release of Reltio that feels more and more like AWS in terms of complete self-service control. 

Interested in others' comments and thoughts on this.
Curt


------------------------------
Curt Pearlman
PwC
Agoura Hills CA
------------------------------