3 ms·
Postgres is a nice open-source platform but the commercial engines are far more advanced in many areas and are not going to be disrupted anytime soon. If you ca
by manigandham 8y ago
Postgres is a nice open-source platform but the commercial engines are far more advanced in many areas and are not going to be disrupted anytime soon. If you can use Postgres for your needs today than it's likely that any relational system would've worked and you didn't need a commercial system in the first place.
SQL Server platform already has one of the most advanced optimizers and distributed planning with its use in Polybase, stretch tables, SQL MPP and Azure SQL DW.
- bdcravens 8y agoAt the end of the day, even if you find a better alternative, database engines are a bear to switch. Our database is still SQL Server, even though we've switched our development platform 3 times, and I'd love to be on Postgres.
- jhall1468 8y agoPostgres also has a number of commercial derivatives, so you get the advantage of starting small (open source) and moving to a more optimized derivative when it becomes necessary. Aster and Greenplum perform exceptionally for what they do. If SQL Server were better that's exactly what companies would use. Commercial engines are definitely more advanced. But some Postgres derivatives can absolutely out perform enterprise platforms in some uses-cases, and vice-versa.
- grumpydba 8y agoThey are more advanced in many levels, a bit behind on the dev friendliness side (postgres is kind there), that's true. However I manage roughly 2000 instances of commercial databases. I'd say maybe a 10th could not be hosted on postgres. It gives postgres a huge disruption potential and the management, in all the big firms I know, is actively looking at it.
- manigandham 8y agoYes, PG has many more usability features. Simple things like UPSERT make UX improvements if you don't need the advanced capabilities of MSSQL. I'm sure PG can take on more of the standard RDBMS workloads today but I don't think that's really making a big dent on SQL Server as the bulk of their revenue comes from the serious enterprise scenarios.
- grumpydba 8y ago> I don't think that's really making a big dent on SQL Server as the bulk of their revenue comes from the serious enterprise scenarios. I'd say postgres can take 90% of the revenues. I've worked in 3 of the top 10 european banks. Most of the SQL Server instances do not even need partitioning, for example. Let alone always on, hekaton and so forth. People mostly buy peace of mind. Until they are charged millions and start questioning the stupid expensive bills saying "do we need that ?" Really I have all the metrics to back this up: CPU usage/Availability requirements/data size... It is literally my job to collect those. Also, proper window support, great postgis, great json support, open source ecosystem support are not simple and huge cost savers.
- manigandham 8y agoRevenue for what? Postgres is free. If you mean the licensing and support from commercial distributions and vendors then, as you already recognize, these decisions will fall to which vendor they trust more. That will usually end up being Microsoft.
- grumpydba 8y agoYou do not trust a vendor who rips you off.
- manigandham 8y agoIt's not the fault of the vendor if you decide to buy their product. If you didn't need the product features then you shouldn't buy it but enterprise deals are rarely about the absolute price.
- grumpydba 8y agoIndeed. And as I said this reasoning might eat 90% of sql server's revenue.
- jeltz 8y ago
- nradov 8y agoWell that's exactly how technology market disruption works. The disruptive product sneaks into the low end of the market without any of the established competitors really noticing. Then the disruptive product gradually moves up market and eats everyone's lunch. 15 years ago Windows Server was far more advanced than any Linux distribution. What does the server OS market look like today?
- garyclarke27 8y agoI moved from SQL Server to Postgres 5 years ago, I would argue the opposite, ie PG is way better. eg pg supports UTF8, csv, jsonb It has way more standard SQL features and support, such as windows analytical and aggregation functions, string agg is particularly powerful. More SQL join options, join on using clause is v handy, lateral join is standard SQL and v powerful and much easier to understand than SQL servers obtuse version. It’s user defined functions are far more powerful, even with standard sql, let alone the umpteem other languages you can use such as Python or Javascript, functions can be chained for incredible power. Many Dev friendly features make you much more productive eg drop schema and the ability to easily use powerful editors such as sublime text or VSCode. With Recent parallel query improvements pg has mostly caught up on performance, it’s only glaring weakness vs Oracle now is lack of auto incremental mat view refresh - MS auto mat views had so many limitations they were almost pointless, last time I looked - pg mat views have no restrictions but are not that useful because they lack incremental refresh.
- manigandham 8y agoYes PG has more usability features but it is not a match on performance at all and lacks the advanced features that enterprises want. I don't know why this is so shocking to hear. These commercial engines have billions of dollars in research and engineering, they're not just standing still and aren't obsolete because PG finally got parallel query (which is only workable in v11 and still years behind the others). There's a reason why all the enterprise PG distributions add so many other features and even basics like connection pooling, because that's what it takes to compete in the enterprise space with its vast and complex requirements. As I said, "If you can use Postgres then you didn't need a commercial system in the first place."