㈜그라운드원 등 2개 사업자는 비밀번호 관리 소홀 등으로 사용자 주민등록번호가 암호화되지 않은 상태로 유출됐고, 이에 따른 개인정보 유출 신고, 통지를 지연했다.
개인정보위는 법규에 명시된 개인정보 유출 신고와 통지 의무, 안전성 확보조치 의무 등을 위반한 6개사 모두에게 과태료를 부과했다. 이에 대해 접근통제를 위반하거나 주민등록번호를 암호화하지 않은 MS 등 3개사에는 과징금을 부과하고, 개인정보취급자 관리 감독을 소홀히 한 ㈜그라운드원 등 3개사에는 개선권고 처분을 내렸다.
MS는 개인정보처리 시스템 관리자 계정 접근통제 등을 하지 않았다. 이메일서비스 '아웃룩'의 계정이 세계에서 11만9432개 유출됐고, 이가운데 국내 이용자 계정 144개의 개인정보도 유출당했다. 영문으로 24시간 이내에 유출 통지가 이뤄졌으나 한국어로는 통지가 11일 지연돼, 개인정보 유출신고와 이용자 통지 등 조치에 늦은 것으로 판단됐다. 과징금 340만원과 과태료 1300만원을 부과받았다.
그라운드원과 이노베이션아카데미는 패스워드 관리 소홀 등으로 암호화를 하지 않은 상태의 주민등록번호를 유출당했다. 이에 따른 개인정보 유출 신고나 통지 조치에 늦었다. 개인정보 취급자에 대한 관리·감독에 소홀했다. 그라운드원은 과징금 2500만원과 과태료 600만원, 이노베이션아카데미는 과징금 2500만원과 과태료 300만원을 부과받았다.
넷기어 레디나스의 원격접속 (넷기어 자체 VPN) 이용이 원활하지 않은 경우가 있어, 해당 원격접속의 장애가 있을 경우를 대비하여 대체수단을 찾고 있다. BR200과 같은 VPN 장비를 설치하여 넷기어 자체 VPN을 대체하는 것이 좋을 같아서 검토중인데, 큐냅이나 시놀러지와 같이 자체에 VPN 서버를 설치할 수 있지 않을까 하여 확인해 보니, 역시 설치할 수 있었다.
Balancing requests is a great way to share workload across the cluster pool and avoid overloading a single server instance, keeping the overall health of the system. There's also the possibility to direct VIP customers to a dedicated cluster pool, guaranteeing them the best user experience.
By using a director group, Varnish will manage and spread the incoming requests to those servers included in it. Having the servers constantly checked, Varnish can take care of the sick servers and maintain everything as if there was no problem at all.
Getting ready
There are six types of director groups to be configured: random, client, hash, round-robin, DNS, and fallback. While the random director is self-explanatory (randomly distributes requests), a DNS director can be used to spread to an entire network of servers. The hash director will always choose a backend based on the hash of the incoming URL and the fallback director can be used in emergency cases when servers behave oddly.
The two most common directors are the round-robin and client directors.
The round-robin can be used to spread requests one by one to the entire cluster of servers no matter what is requested, and the client director can be used to create a sticky-session based on unique information provided by the client, such as an IP address.
In this recipe we will create both the client and round-robin balancers to spread requests across application servers.
How to do it...
Group the backend servers to load balance requests by using the following code snippet:In the preceding example, we have declared that our director nameddr1is around-robindirector, and inside this director there are four backend servers to be balanced. Backend serversserver01toserver04have already been configured earlier and this declaration is only referring to them.
Create a sticky-session pool of servers by using the following code snippet:At first, there's absolutely no difference from theround-robindeclaration to theclientone, but inside yourvcl_recvsubroutine, you'll need to specify what identifies a unique client. Varnish will use, as default, the client IP address, but if you have other services in front of your Varnish Cache (such as a firewall), you'll need to rewrite the value of theclient.identityvariable. In the following example, we'll use theX-Forwarded-Forheader to get the clients' real IP address.sub vcl_recv { set client.identity = req.http.X-Forwarded-For; }
A sticky-session pool is necessary to direct clients to specific parts of your website that requires an HTTP session or authorization, such as shopping cart/checkout and login/logout pages, without breaking their session. Those parts of your website may be critical to your business, and having a sticky-session dedicated cluster can prioritize paying customers while the others are still browsing the products and will not interfere with the checkout process performance.
By using directors to load balance the requests, we can obtain greater service availability and provide paying customers with an improved user experience.
There's more...
Sometimes it's not possible for all the servers to be identical inside a cluster. Some servers may have more available ram or more processors than others, and to balance requests based on the weakest server in the cluster is not the best way to solve this problem, since the higher end servers would be underused.
Weight-based load balancing improves the balance of the system by taking into account a pre-assigned weight for each server as shown in the following code snippet:
