Great article, though MS would argue their data layer is open(ish) to others by adopting Delta Lake across all/most of Fabric, which is most exciting feature to me.
I do wonder how it all cost to use in the real world, as it looks this isn't a pay as you go model and I feel Azure is the worst offender for cloud costs spiralling out of control by putting often essential features in its enterprise tier.
Yeah, I meant to include a footnote about that actually. On one hand, I think that's a pretty big selling point of the whole thing, where it can be a compute layer on top of data wherever it sits (which is starting to become a bit of the norm, with things like Starburst and Snowflake's external tables).
On the other hand, it's only a partial solution. The "integration tax" that they mentioned isn't just about data siloes; it's also about the various applications that sit on top. And that's the part where one vendor will have a really hard time building everything, because people want different products, difference experiences, etc. So Fabric still ends up trying to be all in one, in ways that I'm not sure it'll be able to pull off.
I'm not saying I have built the potato, but I'm not, not saying it either.
I'm at least attempting to build it and it feels like I have the right ingredients: table storage (the gravitational mass), open catalog interface (easy way to connect appendages and facial features), central governance to make sure it all works together in practice for businesses.
Benn I would love your take on what you think the critical potato elements are?
So that's the hard part. It's probably something similar to snowflake's app store honestly, but with more connectivity to other apps.
To me, I want motif to build their language and sisu to build their ML system, and neither of them to need to build bi for people to buy it. I'm not 100% sure how you get there, but it feels like an app store, with more connectivity, and something more like fabric underneath, where you can query across data storage services.
Microsoft is saying all the right things and you're right about their likely path to having a large slice of the market. This is certainly a conceptual improvement over Synapse.
I'm skeptical that anyone else has the chops or desire to go head-to-head. Oracle can, but so many people dislike how they go to market. Google doesn't seem to understand how to bring products to market - one trick pony.
Snowflake would be a good candidate to build an all encompassing platform around their core, but that doesn't seem to be how they think.
Google could (and has all the pieces to), but doesn't seem to be able to put it together. I think Snowflake or Databricks also *could,* but I'm not sure it'd be worth the risk for them. Given their scale compared to Microsoft, they couldn't afford to get much wrong, especially if they alienate partners in the attempt.
Ultimately though, that's sorta my point - I don't think anyone should chase Microsoft in an all-in-one arms race. Because if someone can pull of building the actual platform (ie, the potato), they can let the ecosystem beat Microsoft for them.
And I'd also add that Google kinda had a head start with the whole GCP thing, as my initial thought to this "all in one" idea, was that this feels like something Google has been trying to do for a number of years, but still not eating up much market share. The problem has always been Microsoft effectively owning the enterprise, hence the big £££, as you mentioned. I'd honestly love to see a stable, reliable, and useful potato in this game, but I tend to err on the side of cynicism.
Yeah, I'm not sure what to make of GCP's attempts here. On one hand, it's evidence of how hard this is to pull off; on the other hand, I'm not sure how much of that is just difficulties that GCP had that Microsoft is probably better at avoiding. I'd probably lean slightly towards the former, but, who knows?
Benn maybe I’ve missed it but surely you could market Google fabric with BigQuery, Datastudio, Looker whatever, Colab, Dataform, etc
AWS equally but even less elegantly, but they’ve pushed out Glue, among others.
Frankly my experience with Sharepoint leaves me ice cold with this launch, Microsoft cannot build decent products if Sharepoint (a core product) is so ridiculously bad.
AWS at least are consistent with their unique flavour of industrial, but Google has the best UX of the 3 and so I don’t really see what this is other than marketing.
Yes, or rather, a marketing driven "unification" of their products into a supposedly coherent platform when in reality it is a thin veneer over the regular gang of Spark, Jupyter and PowerBI.
I guess it is a thin veneer, but purely based on priors. Is it any good? I watched this video and it looks more unified than I had initially expected.
Ah, gotcha. And yeah, I had a similar reaction. That's what footnote 6 was about - I thought it'd basically be a wrapper around stuff, but it seems like maybe not? They keep emphasizing how it's "all one SaaS product" too, so they clearly don't want it to just be the veneer.
Great article, though MS would argue their data layer is open(ish) to others by adopting Delta Lake across all/most of Fabric, which is most exciting feature to me.
I do wonder how it all cost to use in the real world, as it looks this isn't a pay as you go model and I feel Azure is the worst offender for cloud costs spiralling out of control by putting often essential features in its enterprise tier.
Yeah, I meant to include a footnote about that actually. On one hand, I think that's a pretty big selling point of the whole thing, where it can be a compute layer on top of data wherever it sits (which is starting to become a bit of the norm, with things like Starburst and Snowflake's external tables).
On the other hand, it's only a partial solution. The "integration tax" that they mentioned isn't just about data siloes; it's also about the various applications that sit on top. And that's the part where one vendor will have a really hard time building everything, because people want different products, difference experiences, etc. So Fabric still ends up trying to be all in one, in ways that I'm not sure it'll be able to pull off.
this one’s a banger, Benn 🥔💥
your sign off called me to leave this here
https://open.spotify.com/track/3sSDCpbg9ju2bqKDfcPTzm?si=Wow6cLZTQ76o9UVaCvml-g
Ke$ha, underrated pop star of the 2010s.
a treasure of our generation
All time grammy performance (inspired by awful circumstances, unfortunately): https://www.youtube.com/watch?v=8FjzcPrab78
holy shhh… that was unreal. thanks for sharing i hadn’t seen this performance. Her new album is lit & super raw. Can’t stop listening.
I'm not saying I have built the potato, but I'm not, not saying it either.
I'm at least attempting to build it and it feels like I have the right ingredients: table storage (the gravitational mass), open catalog interface (easy way to connect appendages and facial features), central governance to make sure it all works together in practice for businesses.
Benn I would love your take on what you think the critical potato elements are?
So that's the hard part. It's probably something similar to snowflake's app store honestly, but with more connectivity to other apps.
To me, I want motif to build their language and sisu to build their ML system, and neither of them to need to build bi for people to buy it. I'm not 100% sure how you get there, but it feels like an app store, with more connectivity, and something more like fabric underneath, where you can query across data storage services.
1-step closer... and this still feels directionally accurate.
Wait, did microsoft release something like this?
Microsoft is saying all the right things and you're right about their likely path to having a large slice of the market. This is certainly a conceptual improvement over Synapse.
I'm skeptical that anyone else has the chops or desire to go head-to-head. Oracle can, but so many people dislike how they go to market. Google doesn't seem to understand how to bring products to market - one trick pony.
Snowflake would be a good candidate to build an all encompassing platform around their core, but that doesn't seem to be how they think.
Google could (and has all the pieces to), but doesn't seem to be able to put it together. I think Snowflake or Databricks also *could,* but I'm not sure it'd be worth the risk for them. Given their scale compared to Microsoft, they couldn't afford to get much wrong, especially if they alienate partners in the attempt.
Ultimately though, that's sorta my point - I don't think anyone should chase Microsoft in an all-in-one arms race. Because if someone can pull of building the actual platform (ie, the potato), they can let the ecosystem beat Microsoft for them.
And I'd also add that Google kinda had a head start with the whole GCP thing, as my initial thought to this "all in one" idea, was that this feels like something Google has been trying to do for a number of years, but still not eating up much market share. The problem has always been Microsoft effectively owning the enterprise, hence the big £££, as you mentioned. I'd honestly love to see a stable, reliable, and useful potato in this game, but I tend to err on the side of cynicism.
Yeah, I'm not sure what to make of GCP's attempts here. On one hand, it's evidence of how hard this is to pull off; on the other hand, I'm not sure how much of that is just difficulties that GCP had that Microsoft is probably better at avoiding. I'd probably lean slightly towards the former, but, who knows?
Benn maybe I’ve missed it but surely you could market Google fabric with BigQuery, Datastudio, Looker whatever, Colab, Dataform, etc
AWS equally but even less elegantly, but they’ve pushed out Glue, among others.
Frankly my experience with Sharepoint leaves me ice cold with this launch, Microsoft cannot build decent products if Sharepoint (a core product) is so ridiculously bad.
AWS at least are consistent with their unique flavour of industrial, but Google has the best UX of the 3 and so I don’t really see what this is other than marketing.
Don't see what Fabric is other than marketing?
Yes, or rather, a marketing driven "unification" of their products into a supposedly coherent platform when in reality it is a thin veneer over the regular gang of Spark, Jupyter and PowerBI.
I guess it is a thin veneer, but purely based on priors. Is it any good? I watched this video and it looks more unified than I had initially expected.
https://www.youtube.com/watch?v=oxMUJT3p9f4
Ah, gotcha. And yeah, I had a similar reaction. That's what footnote 6 was about - I thought it'd basically be a wrapper around stuff, but it seems like maybe not? They keep emphasizing how it's "all one SaaS product" too, so they clearly don't want it to just be the veneer.
The video indicates that at the very least it is a single login screen. I have my doubts, fool me once etc.
Fair, I'm skeptical too, and will probably remain so until Microsoft bulldozes me with their AGI supercomputer.
I pity those that get handed MS solutions without a say in the matter.
This echoed my fears for Fabric:
https://twitter.com/pdrmnvd/status/1663737808643829760
Google could, but they don't know how.
"IBM® ILOG® CPLEX® Optimization Studio"
What?! They didn't call it IBM® ILOG® CPLEX® Optimization Studio Cloud Pak for Optimization?
They've lost their edge. Sad to see it.