Disassemble the advertising benefits of cloud hosting: the answer to the cloud provider
Some time ago, a critical article “ Twisted in the clouds: dismantling the advertising advantages of cloud hosting ” appeared on the virtual server selection service PoiskVPS.ru. The author of the article and service habrauzer leonid239 is quite skeptical about the topic of the popularity of cloud servers, mentioning in its comparison including Cloud4Y . We just could not get past such an enchanting arrangement of points, and decided to publish our answer, sorting out the points already the analysis itself.
We decided to answer point by point, highlighting the abstracts of the original article:
Advertising says: fast scaling.
')
Fast scaling is one of the features of cloud hosting. The fastest scaling is a special case of it - the Autoscale service (autoscale), which has one feature — in automatic mode, additional resources are connected to the peak load, for which you naturally have to pay. Due to the flexible setting and setting of maximum resource allocation thresholds, a situation with a greatly increased invoice for services can be avoided.
Advertising says: clouds are cheap
The question of price is a question of the value of the price itself. That is what the client receives for a particular amount. What is the level of iron, what are the guarantees, the possibility of placing abroad, which software is based on the SaaS model, which payment schemes, which guaranteed SLA, etc ... There are a lot of examples here. The main thing is one thing - high quality and reliability cannot be cheap. You have to choose: either cheaply - or efficiently and reliably.
Advertising says: provide a free trial period. In fact, it turns out that the free test is provided only for the minimum configurations, which are not enough if I want to test, for example, the operation of SSD disks or CPUs over 4.
We provide test access, including for large configurations, if you need to test in real time, for example, performance, fast SSD drives - we give this opportunity. You can test not only CPU, RAM, HDD, channel speed, but also the configuration of 1C Enterprise, as well as 1C Bitrix: Corporate portal. At the end of the test period, if you choose us as a supplier, there is no need to re-establish the infrastructure when switching to a commercial basis: all data remains and you continue to use the cloud on a commercial basis using a postpaid scheme.
Advertising says: flexible tariffs in the clouds.
When we say that we offer flexible tariffs - we really offer flexible tariffs :) On the main page of our website there is a price configurator, where you can choose resources for your requirements, be it 1CPU and 5 GB RAM or 4CPU and 2 GB RAM The same applies to other resources, for example, we offer HDDs of different speeds (4 options) from slow to store static files to fast SSDs with guaranteed IOPS.
The test can get up to 32 CPU, 96 GB of RAM and 300 GB HDD. As a rule, this resource pool is sufficient for testing most tasks. If necessary, the pool of test resources can be increased.
Advertising says: pay only for what you need.
Because of the possibility to tune the server for themselves, hosters lose part of the profits if the client takes exactly how much he needs, and not with a margin, as if he took a regular VPS with fixed tariffs. However, no one likes to lose profits, so this loss is already recorded in the cost of resources, and ultimately the cloud server with parameters identical to the virtual is more expensive.
One of our main concepts is that the client pays only for what he really needs and according to the postpaid scheme. Perhaps, if a client took resources with a margin - then our profit would be higher, but in this case we move away from the cloud concept and deprive the client of the convenience for which he chose us. And this in turn leads to the fact that there can be no talk of any long-term cooperation.
And most importantly - the clouds are completely insured from earthly troubles: if the light gets cut out in the data center or the tractor driver breaks off the optics, then the customers will be completely indifferent, their server is cloudy or virtual.
We had a similar “experience” with a tractor driver;) We were interrupted by the optics with a tractor bucket, unfortunately, no one is immune from this. But since our infrastructure has an optical ring without points of failure with duplicated channels, the clients did not notice anything.
And if any hoster broke one server out of 100, then only 1% of clients would have problems.
In case of failure of any client server, technologies like VMware High Availability are applied, that is, automatic restart of virtual machines on another server (including another data center) within 5-10 minutes.
Indeed, all companies have problems, Amazon is not an exception, but it is important to understand what a company has taken to fix problems and what it did to avoid them in the future.
We openly publish incidents, record and comment on each of them. In particular, after a major DDoS attack at the end of last year, we published comments on the incident and provided information on the prevention of such incidents, namely:
Advertising says: in the clouds, the availability level of SLA services is 99.9%.
I will not be the first to say that in life everything happens and sooner or later providers have problems with accessibility. At one time, 3 days at Amazon solved the problems with the availability of EC2 and RDS. Nevertheless, I want some guarantees, which providers simply do not have.
We provide availability at the level of 99.9% in the basic version - that is, for all customers, regardless of the size of the company. For some customers, an SLA of 99.99% is possible. These guarantees are fixed legally in the "Service Level Agreement". For review on the site is available a full list of documents.
SLA at the level of 99.9% - we consider it very responsible and make every effort to justify such statements.
For this reason, we use:
Nevertheless, we agree with the author’s conclusion: if the clouds are used, it means that people need them. Our turnover is constantly growing. And we will be happy if this post also helps someone to realize that the time of the clouds for him is now.
We decided to answer point by point, highlighting the abstracts of the original article:
Advertising says: fast scaling.
')
Fast scaling is one of the features of cloud hosting. The fastest scaling is a special case of it - the Autoscale service (autoscale), which has one feature — in automatic mode, additional resources are connected to the peak load, for which you naturally have to pay. Due to the flexible setting and setting of maximum resource allocation thresholds, a situation with a greatly increased invoice for services can be avoided.
Advertising says: clouds are cheap
The question of price is a question of the value of the price itself. That is what the client receives for a particular amount. What is the level of iron, what are the guarantees, the possibility of placing abroad, which software is based on the SaaS model, which payment schemes, which guaranteed SLA, etc ... There are a lot of examples here. The main thing is one thing - high quality and reliability cannot be cheap. You have to choose: either cheaply - or efficiently and reliably.
Advertising says: provide a free trial period. In fact, it turns out that the free test is provided only for the minimum configurations, which are not enough if I want to test, for example, the operation of SSD disks or CPUs over 4.
We provide test access, including for large configurations, if you need to test in real time, for example, performance, fast SSD drives - we give this opportunity. You can test not only CPU, RAM, HDD, channel speed, but also the configuration of 1C Enterprise, as well as 1C Bitrix: Corporate portal. At the end of the test period, if you choose us as a supplier, there is no need to re-establish the infrastructure when switching to a commercial basis: all data remains and you continue to use the cloud on a commercial basis using a postpaid scheme.
Advertising says: flexible tariffs in the clouds.
When we say that we offer flexible tariffs - we really offer flexible tariffs :) On the main page of our website there is a price configurator, where you can choose resources for your requirements, be it 1CPU and 5 GB RAM or 4CPU and 2 GB RAM The same applies to other resources, for example, we offer HDDs of different speeds (4 options) from slow to store static files to fast SSDs with guaranteed IOPS.
The test can get up to 32 CPU, 96 GB of RAM and 300 GB HDD. As a rule, this resource pool is sufficient for testing most tasks. If necessary, the pool of test resources can be increased.
Advertising says: pay only for what you need.
Because of the possibility to tune the server for themselves, hosters lose part of the profits if the client takes exactly how much he needs, and not with a margin, as if he took a regular VPS with fixed tariffs. However, no one likes to lose profits, so this loss is already recorded in the cost of resources, and ultimately the cloud server with parameters identical to the virtual is more expensive.
One of our main concepts is that the client pays only for what he really needs and according to the postpaid scheme. Perhaps, if a client took resources with a margin - then our profit would be higher, but in this case we move away from the cloud concept and deprive the client of the convenience for which he chose us. And this in turn leads to the fact that there can be no talk of any long-term cooperation.
And most importantly - the clouds are completely insured from earthly troubles: if the light gets cut out in the data center or the tractor driver breaks off the optics, then the customers will be completely indifferent, their server is cloudy or virtual.
We had a similar “experience” with a tractor driver;) We were interrupted by the optics with a tractor bucket, unfortunately, no one is immune from this. But since our infrastructure has an optical ring without points of failure with duplicated channels, the clients did not notice anything.
And if any hoster broke one server out of 100, then only 1% of clients would have problems.
In case of failure of any client server, technologies like VMware High Availability are applied, that is, automatic restart of virtual machines on another server (including another data center) within 5-10 minutes.
Indeed, all companies have problems, Amazon is not an exception, but it is important to understand what a company has taken to fix problems and what it did to avoid them in the future.
We openly publish incidents, record and comment on each of them. In particular, after a major DDoS attack at the end of last year, we published comments on the incident and provided information on the prevention of such incidents, namely:
- introduced additional communication channels
- introduced traffic filtering
- increased number of uplinks
- added a secure peer, added the ability to transfer traffic through Voxility - the largest provider of Anti-DDoS solutions
Advertising says: in the clouds, the availability level of SLA services is 99.9%.
I will not be the first to say that in life everything happens and sooner or later providers have problems with accessibility. At one time, 3 days at Amazon solved the problems with the availability of EC2 and RDS. Nevertheless, I want some guarantees, which providers simply do not have.
We provide availability at the level of 99.9% in the basic version - that is, for all customers, regardless of the size of the company. For some customers, an SLA of 99.99% is possible. These guarantees are fixed legally in the "Service Level Agreement". For review on the site is available a full list of documents.
SLA at the level of 99.9% - we consider it very responsible and make every effort to justify such statements.
For this reason, we use:
- own high availability ring without a single point of failure, with a capacity of 180 Gbps, with expansion up to 800 Gbps;
- IBM Blade Server, NEC with Duplicated Power, Management;
- Hypervisor Clustering (highability)
- NetApp storage with redundant controllers and a fail-safe RAID array;
- ExtremeNetworks 40g switches in a stack;
- CISCO routers in failover configuration (Failover);
- Duplication: paths to devices; Internet channels of large telecom operators; trunk routes between data centers;
- Wide data center network: 2 in Moscow, 1 in St. Petersburg, 1 in the Netherlands and 1 in Germany
Nevertheless, we agree with the author’s conclusion: if the clouds are used, it means that people need them. Our turnover is constantly growing. And we will be happy if this post also helps someone to realize that the time of the clouds for him is now.
Source: https://habr.com/ru/post/231619/
All Articles