Microsoft Future Decoded, November 1st 2017, Day 2 - Tech Deep Dives
These are my notes from the day. It’s not a detailed summary of the event, just notes for things I found interesting or want to remember to try out or do.
Not much that was relevant to us. Mostly vendors helping people migrate to the cloud.
Focused on Mixed Reality, DNA storage, blockchain in Azure. All very interesting but nothing immediately actionable.
- We should add some validation onto all the entities we store in CosmosDB that runs before we save/update. Probably using FluentValidation. This should include anything used to generate the
- When we go live, start by over-provisioning RUs and then once we can see our usage through the portal, dial it down.
- Find out more about the different consistency levels as there could be performance benefits, particularly when importing data. This is configurable per-request.
- Think about how we want to configure indexing - lazy?
- We can use Azure Functions with CosmosDB triggers to trigger copying of data to SQL Server
- Endjin Cloud Adoption Risk Model poster
IndexingPolicyis configurable, we need to work out what we need
- Even if we’re not partitioning to begin with, we need to set a
PartitionKeyand make sure it’s a good one
- Although the SDK will handle and retry errors, e.g.
429- we need to decide what to do if this still fails. Endjin use a dead letter queue.
- We could provide a UI for editing RU configuration without needing to re-deploy - though the Azure Portal should be able to handle most requirements for this
- Look into per minute request units configuration to give some flexibility for bursty traffic
- Cross partition queries are expensive so avoid using them where possible
- Sign up for http://azureweekly.info/ a summary of the week’s top news in the Microsoft Azure ecosystem, as well as interesting historic content that’s well worth reading.
- Since we’re migrating from RavenDB, check for gaps in CosmosDB tooling, e.g. how easy is it to manually edit a single document?
- Once you select the API for accessing CosmosDB you can’t then use a different API. For example you can’t mix the table storage API with DocumentDB API.
- Check the current document size limit and make sure none of the data we’re migrating is approaching it
- Make sure we’re logging the RUs used by all requests. See: Cosmos DB Request Units and the .NET SDK
- Look at how to use the VSTS Release feature alongside Gitflow/Hubflow. There must be some documentation on it.
- PowerBI can be hooked up to VSTS so you can generate dashboards and reports from VSTS activity. This could be a good way of getting an overview of all builds/releases. See: https://docs.microsoft.com/en-us/vsts/report/powerbi/
- Find about the Microsoft Assessment and Planning Toolkit (not relevant to us as designed for cloud migrations)
- A classification, e.g. “address”, “name”, “email” can be assigned to columns in SQL Server to make it easier to identify personal data
- Some example code used the
FORMSOFfunction - could be useful one day. See: Querying SQL Server Using Full-Text Search
- Try out SQL Server Vulnerability Assessment Tools
- There was an interesting demo of temporal tables showing how it can be used to a) expire data after a specified date b) restore data to a previous version (without doing a full database restore). See: Temporal Tables Temporal tables
- It was implied that it’s okay to keep backups containing data of individuals that have asked to have their data deleted so long as you also keep a list of these individuals and make sure that after restoring any backups, there’s a process to run through this list and re-delete the data of those individuals.
I didn’t stay for this but the video is available online.