4 ms·

I work at AWS and I have questions for the author about this post: TL;DR -- I urge all readers to take this post with a grain/boulder of salt: The author is an

by ranman 10y ago

I work at AWS and I have questions for the author about this post: TL;DR -- I urge all readers to take this post with a grain/boulder of salt: The author is anonymous, prone to hyperbole and error, and makes multiple unverified claims. I'm sure he's a nice guy/girl and if we met in real life I'd be happy to converse over a beer -- but I find the language and misinformation in the post overly polemic and disingenuous. If this stuff interests you guys tune into the reinvent livestream next week: https://reinvent.awsevents.com/live-streaming/ https://reinvent.awsevents.com/live-streaming/ All numbers are from EU-WEST-1 (Does anyone know why GCE doesn't have a eu-west-1a? only b,c,d? I'd be curious to know the story there... not trolling -- just curious.) Graph 2: c4.large has 2 CPUs... n1-standard-1 has 1... comparison on price is strange? Your point below about optimizing for manageability doesn't really make sense to me as manageability would stem from config/deploy/etc. -- not from instance type? Perhaps your language isn't clear. You're anonymous but I'm guessing english is a second language for you (judging from sentences in post) so it's possible we're missing your point there. Could you clarify please? >"an ancient virtualization technology" source? details? KVM came out in 2007. Xen came out in 2003. I don't consider either of those dates particularly ancient but whatever it's your post, use the language you want. Graph 3: A c4.4xlarge has 16 CPUs, and 30 gigs of RAM. An n1-highcpu-4 has 4 CPUs and <4 gigs of RAM. Comparing the two on price is disingenuous. Your claims that these are the "production instances" don't make sense to me. >Network Heavy: So this is a test between two t2.micros in different AZs: https://s3.amazonaws.com/ranman-code/2016-11-20+00.58.23.gif https://s3.amazonaws.com/ranman-code/2016-11-20+00.58.23.gif It shows 1gbps... I ran it literally a minute ago... So that's 1gbit... Do you mean inbound from public internet or outbound to public internet? I created an n1-highcpu-4 and had it talk to a c4.xlarge at 1gpbs in (both in eu-west) so your claim that you need a c4.4xlarge for 1gbps seems dubious. Test Complete. Summary Results: [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-60.00 sec 7.40 GBytes 1.06 Gbits/sec 1297597 sender [ 4] 0.00-60.00 sec 7.40 GBytes 1.06 Gbits/sec receiver It actually seems like Google's network was the limiting factor there because when I ran a similar test on a 10gpbs instance I couldn't get faster than 1gpbs. Which is fine because that's what's advertised -- just pointing out that the need for a c4.4xlarge is wrong. On a c4.large I got 900mbps. >C4/m4/r3 have a hard cap at 220 mbits This one is just false. What am I missing here? Where did you get this info? From eu-west-1 to us-east-1 I get faster than that using public routes and t2.micros. Internally across availability zones I get much faster than that. Local SSD #s: Use PIOPS volumes, tunable size, lower cost, can attach to any instance type. Also consider the numerous other instance types that offer ephemeral SSDs. >It is quite flexible. For instance, we could recreate any instance from AWS on Google Cloud. I don't think you mean any instance... but ok sure! >Amazon does everything wrong, and Google does everything right, A message by an employee from Amazon than Google, not directly relevant but still a good read. ^ I don't think you read the above post when listing it as a reference. The title is ironic. Steve Yegge quit Google. Famously... on stage... He also goes on to say that Google can't do platforms. I'm sure that's changed in the past few years though. In the end I'm super excited about both AWS and GCP they both have awesome products. I encourage people to continue researching and writing posts around these topics. It's hard to not take some of the criticisms personally when you're passionate and invested in your work (at least it is for me). I've got a google employee DM-ing on twitter calling me "petulant child" among other things. We don't need that. It's not helpful for our companies and it's not helpful for our customers. I'll encourage everyone to tune in to the reinvent livestream on the 29th, 30th, 1st: https://reinvent.awsevents.com/live-streaming/ https://reinvent.awsevents.com/live-streaming/

logmeout 10y ago
Until bandwidth pricing is fixed rather than nickel and dimeing us to death; a lot of us will choose fixed pricing alternatives to AWS, GCP and Rackspace.
jsolson 10y ago
Note... I work on GCE. That said, numbers I'm posting here for GCE are from plugging custom instance and volume types into the GCE console on my personal account rather than anything formal (I'm on my personal laptop). While you can nitpick the OP's points, our custom instance types make it possible to do something closer to an apples-to-apples comparison, and by my math below (which may be wrong; the risk of posting at 1am) the 50% number still holds up if you ignore reserved instances, assuming the OP's prices for AWS were correct. > Does anyone know why GCE doesn't have a eu-west-1a? only b,c,d? I'd be curious to know the story there... not trolling -- just curious. Just guessing (I've worked on GCE for a bit over three and a half years, but I don't remember the history here) -- I suspect that it was a zone that existed prior to GCE going generally available that was turned down and there was a desire to not re-use the name. > c4.large has 2 CPUs... n1-standard-1 has 1... comparison on price is strange? Well, ok, but a custom 2 vCPU instance with 3.75 GB of RAM is $45/month (including a 10 GB PD root volume), so still nearly 50% lower than the c4.large price. Similarly, if I build a c4.4xlarge equivalent custom instance type (16 vCPUs and 30 GB RAM) it's $358/month, again putting it at just about 50% of the rack rate for a c4.4xlarge. I'm not sure where the OPs claims were coming from for networking. Certainly you can get 1 Gbps out of relatively small EC2 instance types, although at a virtual NIC level GCE is considerably more generous, providing virtual 2 Gbps/s/vCPU link speeds (so up to 8 Gbps for a 4 vCPU instance versus 750 Mbps for a c4.xlarge). Bandwidth-to-the-edge will be lower. That said, when trying to find an upper bound on bandwidth it's really best to test (regardless of which platform you're evaluating) with multiple streams ideally to multiple peers. Single stream can end up limited by a number of factors including RTT and congestion window scaling bounds as well as getting unlucky in terms of ECMP path selection for multi-path links. > Local SSD #s: Use PIOPS volumes, tunable size, lower cost, can attach to any instance type. PIOPS volumes are EBS-SSD, not local, aren't they (looking at https://aws.amazon.com/ebs/pricing/ https://aws.amazon.com/ebs/pricing/)? Our analog would be PD-SSD. GCE PD (including PD-SSD) IOPS scale with volume size — let's say you needed 20 GB of storage at 10000 IOPS. On EC2 you'd be paying $0.125×20+$0.065×10000 = $652.5. On GCE you'd need to buy a bigger volume, at 30 IOPS/GB for PD-SSD you'd need a ~334 GB volume. However, at $.17/GB you'd get the same 10000 IOPS at $57/month — for this example EC2 is more than 10x as expensive, and if we scale the storage up so you get the same 334 GB on io1 EBS PIOPS SSD it's more than 12x as expensive. I picked these numbers somewhat arbitrarily; EC2 would've come out looking better for very large volumes with very low performance requirements, but with PD-SSD priced at $.17/GB and io1 volumes priced at $.125/GB + .065×1IOPS = $0.19, io1 volumes are still more expensive even at the lowest IOPS bound (above 1GB they get cheaper, but only below 3 IOPS). General purpose SSD is cheaper than PD-SSD, so I suppose for users who can live without a performance SLA it might be a good choice, although at that point I'd be tempted to benchmark it against our vanilla PD offering. You can also attach GCE's truly local SSD (as opposed to PD-SSD) to any instance type (with really fantastic IO performance -- up to 680,000 IOPS read perf with multiple volumes), although I'll grant you that the 375 GB granularity is a little... coarse. > source? details? KVM came out in 2007. Xen came out in 2003. To be fair, GCE doesn't exactly run on stock KVM (nor, I assume, is EC2 running on stock Xen), and our userland is not qemu. I don't think it's meaningful to try and assess these technologies from the outside[0]. > Steve Yegge quit Google. Famously... on stage... He didn't, he just switched projects... he's still at Google. He went on to say that Google couldn't do platforms; one of my favorite things about working here is the company's ability to be introspective and recognize when a rant like Steve's has more than a grain of truth to it. Google adapts. I think comparing the GCP APIs to Google's historical API efforts makes it abundantly clear that Google has improved by leaps and bounds at building a cohesive platforms. [0]: I do lead the team that owns our virtual NIC and a chunk of our host dataplane in addition to prior involvement in broader aspects of our hypervisor architecture, so I'll admit to more than a little bias on this front :)

by ranman 10y ago

logmeout 10y ago

Until bandwidth pricing is fixed rather than nickel and dimeing us to death; a lot of us will choose fixed pricing alternatives to AWS, GCP and Rackspace.

jsolson 10y ago

Note... I work on GCE. That said, numbers I'm posting here for GCE are from plugging custom instance and volume types into the GCE console on my personal account rather than anything formal (I'm on my personal laptop). While you can nitpick the OP's points, our custom instance types make it possible to do something closer to an apples-to-apples comparison, and by my math below (which may be wrong; the risk of posting at 1am) the 50% number still holds up if you ignore reserved instances, assuming the OP's prices for AWS were correct. > Does anyone know why GCE doesn't have a eu-west-1a? only b,c,d? I'd be curious to know the story there... not trolling -- just curious. Just guessing (I've worked on GCE for a bit over three and a half years, but I don't remember the history here) -- I suspect that it was a zone that existed prior to GCE going generally available that was turned down and there was a desire to not re-use the name. > c4.large has 2 CPUs... n1-standard-1 has 1... comparison on price is strange? Well, ok, but a custom 2 vCPU instance with 3.75 GB of RAM is $45/month (including a 10 GB PD root volume), so still nearly 50% lower than the c4.large price. Similarly, if I build a c4.4xlarge equivalent custom instance type (16 vCPUs and 30 GB RAM) it's $358/month, again putting it at just about 50% of the rack rate for a c4.4xlarge. I'm not sure where the OPs claims were coming from for networking. Certainly you can get 1 Gbps out of relatively small EC2 instance types, although at a virtual NIC level GCE is considerably more generous, providing virtual 2 Gbps/s/vCPU link speeds (so up to 8 Gbps for a 4 vCPU instance versus 750 Mbps for a c4.xlarge). Bandwidth-to-the-edge will be lower. That said, when trying to find an upper bound on bandwidth it's really best to test (regardless of which platform you're evaluating) with multiple streams ideally to multiple peers. Single stream can end up limited by a number of factors including RTT and congestion window scaling bounds as well as getting unlucky in terms of ECMP path selection for multi-path links. > Local SSD #s: Use PIOPS volumes, tunable size, lower cost, can attach to any instance type. PIOPS volumes are EBS-SSD, not local, aren't they (looking at https://aws.amazon.com/ebs/pricing/ https://aws.amazon.com/ebs/pricing/)? Our analog would be PD-SSD. GCE PD (including PD-SSD) IOPS scale with volume size — let's say you needed 20 GB of storage at 10000 IOPS. On EC2 you'd be paying $0.125×20+$0.065×10000 = $652.5. On GCE you'd need to buy a bigger volume, at 30 IOPS/GB for PD-SSD you'd need a ~334 GB volume. However, at $.17/GB you'd get the same 10000 IOPS at $57/month — for this example EC2 is more than 10x as expensive, and if we scale the storage up so you get the same 334 GB on io1 EBS PIOPS SSD it's more than 12x as expensive. I picked these numbers somewhat arbitrarily; EC2 would've come out looking better for very large volumes with very low performance requirements, but with PD-SSD priced at $.17/GB and io1 volumes priced at $.125/GB + .065×1IOPS = $0.19, io1 volumes are still more expensive even at the lowest IOPS bound (above 1GB they get cheaper, but only below 3 IOPS). General purpose SSD is cheaper than PD-SSD, so I suppose for users who can live without a performance SLA it might be a good choice, although at that point I'd be tempted to benchmark it against our vanilla PD offering. You can also attach GCE's truly local SSD (as opposed to PD-SSD) to any instance type (with really fantastic IO performance -- up to 680,000 IOPS read perf with multiple volumes), although I'll grant you that the 375 GB granularity is a little... coarse. > source? details? KVM came out in 2007. Xen came out in 2003. To be fair, GCE doesn't exactly run on stock KVM (nor, I assume, is EC2 running on stock Xen), and our userland is not qemu. I don't think it's meaningful to try and assess these technologies from the outside[0]. > Steve Yegge quit Google. Famously... on stage... He didn't, he just switched projects... he's still at Google. He went on to say that Google couldn't do platforms; one of my favorite things about working here is the company's ability to be introspective and recognize when a rant like Steve's has more than a grain of truth to it. Google adapts. I think comparing the GCP APIs to Google's historical API efforts makes it abundantly clear that Google has improved by leaps and bounds at building a cohesive platforms. [0]: I do lead the team that owns our virtual NIC and a chunk of our host dataplane in addition to prior involvement in broader aspects of our hypervisor architecture, so I'll admit to more than a little bias on this front :)