6 ms·
Google Cloud SQL now available with an SLA, 500 GB databases, and encryption
- dsr12 13y agoAs all data is encrypted when stored, I think it will have serious performance degradation for large databases. If it performs well then it will be interesting to know how Google is achieving that.
- omh 13y agoWhy would there be a significant performance degradation? I'd expect this to be implemented by hardware encryption, probably at the storage layer, with minimal speed impact.
- thrownaway2424 13y agoThat seeks like by far the least likely way it is implemented.
- lallysingh 13y agoWhy? It's also the cheapest (in CPU utilization).
- thrownaway2424 13y agoA notorious cheapskate like Google is not going to pay to put hardware crypto modules in huge number of servers. Those things are expensive, power hungry, unreliable, instantly obsolete, full of bugs, often associated with binary drivers or HALs, and barely or not at all faster than CPU cryptography. I've never seen a case where a real server with a real CPU could benefit from hardware crypto.
- lallysingh 13y agoAES-NI - http://software.intel.com/en-us/articles/intel-advanced-encryption-standard-instructions-aes-ni http://software.intel.com/en-us/articles/intel-advanced-encr...
- omh 13y agoAny modern CPU has dedicated hardware for encryption. And a huge number of storage devices have it too - either storage arrays (which perhaps Google aren't using) or standard hard disks, especially SSDs.
- zurn 13y agoIntel CPUs got hardware AES instructions in 2010 or so. Even before that software AES went at a good clip, a 200+ MB/s. On top of that, the fragmented (seek dominated) nature of typical database IO, and the fact this only happens only on cache misses...
- pg_is_a_butt 13y agoi've been building cloud SQL apps for a long time... one major issue is that the cloud SQL instances are "transient" which means at any time google can decide to optimize your app by moving your DB to a different datacenter... while that is happening, connection attempts will respond with "MySQL server has gone away". so the 99.95% uptime shouldn't be seen as "1 day every 6 years the DB will be down" or "half a day every 3 years"... it should more been seen as "1 connection in every 2000 will fail"
- omh 13y agoWhat is the purpose of encryption in these cases? Is it just a compliance tick box? The keys are managed by Google and obviously this won't protect against them getting at your data. And any external attacker who gets into Google's infrastructure could (in theory at least) get the keys as well. It presumably protects against someone stealing the hard drives from the data centre, but that doesn't seem like a significant threat.
- corresation 13y agoThe various internal encryption statements seem to be speaking to the various NSA issues.
- omh 13y agoDoes anyone who is concerned about the NSA issues seriously think that encryption (in the way it's implemented here) would be any help?
- pg_is_a_butt 13y agothe data is often transferred between google datacenters, as the DB is "transient"... that transfer is probably also encrypted, but the NSA may already have those links compromised... encrypting the data and the transport can't hurt...
- michaelt 13y agoThe encryption that has been proposed does nothing to address the NSA issues. Amazon has a similar feature on S3 - they will encrypt your data on upload, and automatically decrypt it when you read it [1]. From a security perspective, this appears to be useless [2]. I suppose it might be useful if you're at an organisation with a rule saying "all data must be encrypted at rest" where the rule cannot be changed, but does not have to be implemented in an effective or useful manner. [1] http://aws.typepad.com/aws/2011/10/new-amazon-s3-server-side-encryption.html http://aws.typepad.com/aws/2011/10/new-amazon-s3-server-side... [2] http://security.stackexchange.com/questions/8765/what-does-amazons-s3-server-side-encryption-protect-against http://security.stackexchange.com/questions/8765/what-does-a...
- tsumnia 13y agoHas anyone had the opportunity to try Google's Cloud in comparison to AWS?
- curiousDog 13y agoI'd say both Azure and AWS are better right now in terms of features, availability, maturity and support. App engine support is terrible
- tsumnia 13y agoI just switched from GAE to AWS due to not handling OpenCV; but I had someone suggest I could do the image handling via Google Cloud. I'd already made the switch, but was still curious since the original site was written for GAE.
- curiousDog 13y agoAnyone know how they handle failover? I'd imagine they won't be physically copying 500GB databases between machines. But then the size limitation indicates they're constrained to what a single machine can hold.
- RyanZAG 13y agoThey're almost definitely copying 500GB databases between machines. More than likely your 500GB database is running across 3 machines, and if 1 fails they bring a new one up and copy your database over. They wouldn't be able to provide that SLA otherwise. 500GB copies reasonable quickly on the in-datacenter links Google runs their servers on, afaik.
- Lewisham 13y agoBasically, Cloud SQL databases are written out to Google's storage system, which automagically replicates across several datacenters. If the primary datacenter that your database is in goes down, it's still replicated across other datacenters and the data at rest isn't at risk. When a connection comes in, we see that the database isn't up right now, and spin in up into a new datacenter (this happens automatically if your database is always on). We offer two different ways of working with Cloud SQL: synchronous and asynchronous writes. With the synchronous writes, you'll get an OK back when you update the database, which lets you know that the write completed successfully and is replicated. If something bad happens, you'll get an exception back and can keep the data in memory until the database is live and accepting connections again. A faster method is asynchronous, which performs the writes every second or so, so you're not waiting for replication to complete. If something dies during that second, you could lose data during that period and not know it. Does this help? Disclaimer: I'm a developer on Cloud SQL.
- curiousDog 13y agoAh I see, so the database sits in a shared storage network and you spin-up a VM with MySQL when needed? If yes, is latency a concern? Also, if my database goes cold, will you spin down the VM?
- chiph 13y agoPricing info here: https://developers.google.com/cloud-sql/pricing https://developers.google.com/cloud-sql/pricing You can buy a package (reserved capacity?) or pay per-use.
- bpicolo 13y agoDoesn't seem that page includes the 500GB package yet (or anything even close)
- Lewisham 13y agoThe 500GB bit is listed under "Storage Size", the column in the table is the amount of storage you get without needing to pay. This is a bit confusing, I'll file a bug to get it cleaned up. Thanks for the spot.
- sandGorgon 13y agoWhy do most of these "cloud database" offerings, seem to be built on MySQL ? RDS, Rackspace and Google all are MySQL first - only Heroku seems to be Postgres. Does Mysql lend itself innately to scalability at Amazon/Google scale that is hard to do with Postgres ?
- derengel 13y agoJust guessing but maybe cause MySQL is(was?) more popular? the PHP and Java crowd is popular for using MySQL
- res0nat0r 13y agoPostgres support was added to RDS a few months ago FYI. http://aws.amazon.com/rds/postgresql/details/ http://aws.amazon.com/rds/postgresql/details/ Also I think it is just because MySQL is generally more popular than PG, so it is deployed first in most cases, with PG support being added later.
- sandGorgon 13y agoyes I am aware of that - PG came after SQL Server and Oracle added to RDS first. I dont know why the situation is so bad. Rackspace, for all its much-vaunted "fanatical support", refuses to touch Postgres.
- res0nat0r 13y agoMainly because to provide 'fanatical support' of PG means adding an entire new software component and everything that comes with it, and ensuring they can keep delivering high quality service. They probably don't want to take on the operational burden adding a 2nd DB would entail vs. the added revenue it might generate.
- Lewisham 13y agoMySQL provides the underpinnings for many other hosted apps, like Wordpress. It's the lingua franca of web databases, so that's why we're working with it. Disclaimer: I'm a developer on Cloud SQL.
- cwyers 13y agoIt's frustrating how the docs constantly refer to MySQL, but make no mention of version number. This page includes a version, seemingly on accident, but because it's part of an example of using the MySQL connector I have no idea if it reflects the version currently in use or just the one that was in use when this page was written: https://developers.google.com/cloud-sql/docs/mysql-client https://developers.google.com/cloud-sql/docs/mysql-client It's showing 5.5 -- 5.6 went General Availability over a year ago, and it's not even a recent build of 5.5 at that, so you're seemingly missing out on nearly a year's worth of bugfixes and security patches. I know Google is applying a lot of custom patches to MySQL here, but I would really worry that the work they're doing to customize MySQL is causing them to lag behind the official version in terms of features and fixes.
- tnuc 13y agoPurely on pricing Azure easily wins. http://www.cleardb.com/store/azure http://www.cleardb.com/store/azure On an ability to plug things in and get things working Azure still wins. I would stick to AWS or Azure. I get the feeling Google is only offering the same as its in house services to the public so they can get free troubleshooting/bug fixing and stop their support staff from growing complacent.
- Lewisham 13y agoI would frame it the other way which is that Google has had to dogfood the inhouse services so has spent a bunch of time troubleshooting/bug fixing internally ;) I know that Google as a whole has a rep for support which isn't great. However, I also know that Cloud is trying really hard to show that isn't the case with Cloud products (as is the company as a whole, my personal experience with support phone lines at Google Play, for example, have been really good). We have Stack Overflow (which we as devs check in on to see if there are problems we can help with e.g. [1]), we have support packages which have real live human beings on the other end [2]. I know everyone here in my dev team is 110% behind support being great, but it's one of those things where we'll be judged by execution rather than words (as it should be). [1] http://stackoverflow.com/questions/21774738/google-cloud-sql-authorized-network-cant-connect http://stackoverflow.com/questions/21774738/google-cloud-sql... [2] https://cloud.google.com/support/ https://cloud.google.com/support/
- jfoster 13y agoThe support experience is going to vary depending on what you need. I once called Google Wallet developer support to let them know that their sandbox was down. I pointed them toward some samples on their documentation that had broken as a result. They didn't seem able to grasp that it wasn't a problem on my end, and I'm not convinced they knew what a sandbox is. In the end, I reached out to a Google Wallet product manager via Google+ who confirmed that it was down and that they were working on it. It seems a waste of talent to have developers picking up the phone when I call developer support, but it would be nice if that team could run stuff by developers when they need to. They also need better training on typical things that developers might call about. (Eg. "Sandbox seems to be down, can you confirm and provide an ETA on the fix?")
- russell_h 13y agoThe SLA seems odd: - "Downtime" means more than a twenty percent Error Rate. Downtime is measured based on server side Error Rate. - "Error Rate" means the number of Valid Requests to open a connection that fail to open a connection, divided by the total number of Valid Requests during that period. So does a network outage not count as downtime?
- mark_l_watson 13y agoI tried this a year or two ago when it was in beta, seemed OK. I used to run several web apps on AppEngine, but more or less stopped when they dropped Wave as a product - I basically had a negative emotional response to a cancelled service, and walked away from AppEngine. I have been thinking of giving AppEngine, etc. another chance but I have been spoiled by very good customer support at RimuHosting, and the knowledge that some support for AWS would be available if I ever had any problems. Google needs to crank up their customer support efforts. I guess that the fact that Google makes relatively little money from PaaS services worries me. All that said, I really enjoyed using Google internal infrastructure during a brief consulting gig at Google, so using AppEngine, etc. is appealing to me out of nostalgia - for for the general public they might not be a great choice.
- jpalomaki 13y agoDangerous to compare these by just looking at features or pricing. Based on my limited experience from these kind of shared services I would say the devil is in the details and in performance. In shared services one interesting topic is how the various limits are implemented. Is there some hard timeouts that prevent complex queries? Is there throttling? Do you get some kind of penalties for running queries that are not optimal? Do you need to implement some specific retry-logic?